Part 1

Part 1
Power Market Fundamentals
Prologue
1 Why Deregulate?
2 What to Deregulate
3 Pricing Power, Energy, and Capacity
4 Power Supply and Demand
5 What Is Competition?
6 Marginal Cost in a Power Market
7 Structure
8 Architecture
9 Designing and Testing Market Rules
© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Prologue
Is it a fact–or have I dreamt it—that, by means of electricity, the world of matter has become a
great nerve, vibrating thousands of miles in a breathless point of time? Rather, the round globe
is a vast head, a brain, instinct with intelligence! Or, shall we say, it is itself a thought, nothing
but thought, and no longer the substance that we dreamed it?
Nathaniel Hawthorne
The House Of Seven Gables
1851
P OWER SYSTEM ECONOMICS PROVIDES A PRACTICAL INTRODUC-

TION TO POWER-MARKET DESIGN. To assist engineers, lawyers, regulators,
and economists in crossing the boundaries between their fields, it provides the
necessary background in economics and engineering. While Part 1 covers basics,
it provides fresh insights ranging from a streamlined method for calculations, to
the adaptation of economics to the quirks of generation models, to the distinction
between the market structure and market architecture.
Part 2 focuses on the core structure of power markets which determines the basic
character of supply and demand. It encompasses demand-side flaws, short-run
reliability policy and the rigidities of supply. Together these determine the notorious
price-spikes and unstable investment pattern of power markets. Because of its
fundamental nature, this analysis can proceed without reference to locational pricing
or unit commitment which gives less-technical readers access to the most important
and fundamental economics of power markets.
Part 3 discusses the architecture of the day-ahead and real-time markets. This
requires the introduction and analysis of the unit-commitment problem—the
problem of starting and stopping generators economically. To avoid unnecessary
complexity, the other primary problem of power-system economics, network
congestion, is postponed until Part 5. This allows a clearer comparison of the three
fundamental types of power trading: bilateral trading, exchange trading and pool-
based trading.
Part 4 detours from the drive toward an increasingly detailed view of the market
to examine market power. Although best understood in the context of Parts 2 and

PROLOGUE 3
3, it does not consider the implications of the network effects described in Part 5.
It can be read directly after Chapters 1-5 and 1-6 if desired. Part 5 begins with more
than enough power engineering to understand losses and congestion pricing which
are its central topics. The theme of Part 5 is that locational loss prices and conges-
tion prices are nothing more than ordinary (bilateral) competitive market prices.
The problem of market design is not to invent clever new prices but to design a
market that will reliably discover the same prices economics has been suggesting
since Adam Smith.
Power markets deviate from standard economics in two ways: the demand sides
are largely disconnected from the market and the details of supply costs violate
the assumptions of competitive economics. Part 2 focuses on the demand-side flaws
which require a regulatory intervention for a few hours per year to ensure reliability
and efficient investment. Until these flaws are sufficiently reduced they will remain
a great danger to power markets that are poorly designed. Part 2, while basic,
contains the most important new material in the book.
While the aggregate market supply function is perfectly normal, the details of
generation costs violate a basic competitive assumption. Because the violations
are small relative to the size of the market, they may require only a slight adjustment
to power exchange bids. Alternatively, a complex and opaque power pool may be
needed. The contribution of Part 3 is to present this problem in an accessible manner
and to highlight crucial questions that still need answers.
The book tells its story largely through examples. These are highly simplified
but designed to capture important phenomena and display their essential natures.
The key conclusions drawn from the examples are summarized in “Results” and
“Fallacies” which are listed after the table of contents. “Result” is not the best of
terms, but Fallacy has no acceptable antonym. Results are not theorems because
they are not stated rigorously. They are rarely new; most are standard economic
wisdom applied to power markets. They are simply the key points that should be
understood in each area. The Results distill much theory that is not presented, but
the examples reveal the mechanisms at work behind the Results. Fallacies are
treated explicitly to help dispel the handful of popular misconceptions that continu-
ally cause confusion.
READING TO DIFFERENT DEPTHS

Readers who wish to read more deeply on some subjects than on others will find
all chapters organized to facilitate this. Chapters consist of the following parts:

4 PART 1 Power Market Fundamentals
1. Introduction (untitled)
2. Chapter Summary
3. Section Summaries
4. Sections
5. Technical Supplement (for some chapters)
Use the chapter summary of 50 to 200 words to determine if a chapter covers a
topic of interest and to learn its most essential points. For a complete overview of
the chapter’s content, read the Section Summaries. For the importance and context
of the chapter’s focus, read the introduction.
When reading an entire Part, the chapter and section summaries can be skipped,
though they may still provide a useful orientation. The technical supplement usually
contains more difficult mathematics but never more than easy calculus.
READING OUT OF ORDER

Readers with special interests may wish to read chap-
For the Nonlinear Reader ters out of order. The book may also serve as a refer-
A few terms and concepts are crucial and, dependence, with different parts used as they become rele-
ing on the reader's background, frequently misunder- vant. Although each Part of the book has been written
stood. Other terms are specific to this book. These for sequential reading, any of Parts 2 through 5 may
are defined here for the reader who wishes to use be read after Part 1. Even within Parts, chapters in-
the book more as a reference than a text. clude cross references to assist those who read only
a particular chapter. The glossary, which defines all
terms set in bold, will be particularly helpful for those who have skipped previous
chapters.
Skipping chapters or reading them out of order is encouraged, but the reader
will need to understand the concepts discussed below. If these seem difficult, the
reader is advised to begin by reading Part 1. Further clarifications and corrections
can be found at www.stoft.com.
Conventions Specific to this Book

The book follows five conventions that are not customary although they deviate
from custom only to follow conventional economics more closely or more conve-
niently.
Units. Fixed cost and variable cost, as well as the cost of energy, power and
capacity are all properly measured in $/MWh. Duration, though sometimes ex-
pressed in hours per year, is represented in equations by a probability. These units
must be understood before reading the book’s examples. See page 31.
Marginal Cost. The standard definition is used when applicable but is generalized
to cover models that use right-angle supply curves. These assume the supply curve
changes from flat to vertical with an infinitesimal change in output. In such cases,
the left- and right-hand marginal costs are defined to be the cost of the last unit

PROLOGUE 5
produced and the cost to produce the next unit. The marginal-cost range is defined
as the set of values from left-hand to right-hand marginal cost. See page 65.
Variable Cost. Because marginal cost is not defined at full output for a generator
with a standard right-angle supply function, the constant marginal cost before full
output is called the generator’s variable cost. This remains defined even when
marginal cost is not, as it is a property of the supply curve and not dependent on
the generator’s level of output. See page 69.
Scarcity Rent. “Scarcity rent” is not defined in economics texts, yet it is a term
commonly used in power economics. Popular usage attempts to distinguish between
rents earned when supply of all generation is scarce and those earned when only
some types are scarce. This book avoids the ambiguity inherent in such definitions
by defining scarcity rent as equal to the more conventional economic term,
inframarginal rent, which is the area below the market price and above the competi-
tive supply curve. See page 70.
Aggregate Price Spike. The aggregate price spike is defined as the upper portion
of the price duration curve, specifically the region in which price is above the
variable cost of the most-expensive investment-grade peaker (not an old, inefficient
peaker). This cut-off value is well defined only in simple models, but it still provides
some intuition about real markets. Revenue associated with this spike is called the
price-spike revenue and is similar to the popular meaning of scarcity rent. See page
127.
Conventional Terms That Are Sometimes Misinterpreted
Profits. “Profit” is used by economists to mean long-run economic profits, and

these average zero under perfect competition. Short-run profit equals scarcity rent
minus startup and no-load costs. Short-run profits cover fixed costs. See page 58.
Cost Minimization. This refers to the minimization of production cost, not to
the minimization of consumer cost.
Market Clearing. A market has cleared when there are neither offers to buy
output for more than the market-clearing price nor offers to sell output for less than
the market-clearing price. When the market has cleared, supply equals demand.
The market-clearing price need not be the competitive price, an efficient price, a
fair price, or the price set by the system operator.
Efficiency. Efficient production minimizes production cost given the output level.
Efficient trade maximizes total surplus, the sum of consumer surplus and producer
surplus (short-run profits). See page 53.
Spot market. The spot market is only the real-time market, not the day-ahead
market.

Chapter 1-1
Why Deregulate?
The propensity to truck, barter, and exchange one thing for another
. . . is common to all men.
Adam Smith
The Wealth of Nations
1776
I N THE BEGINNING THERE WAS COMPETITION—BRUTAL AND INEFFI-

CIENT. Between 1887 and 1893, twenty-four central station power companies
were established within Chicago alone. With overlapping distribution lines, competi-
tion for customers was fierce and costs were high. In 1898, the same year he was
elected president of the National Electric Light Association, Samuel Insull solved
these problems by acquiring a monopoly over all central-station production in
Chicago. In his historic presidential address to NELA, Insull explained not only
why the electricity business was a “natural monopoly” but why it should be regu-
lated and why this regulation should be at the state level, not the local level. Insull
argued that
exclusive franchises should be coupled with the conditions of
public control, requiring all charges for services fixed by public
bodies to be based on cost plus a reasonable profit.
These ideas shocked his fellow utility executives but led fairly directly to regulatory
laws passed by New York and Wisconsin in 1907 establishing the first two state
utility commissions. Reformers of the Progressive era also lent support to regulation
although they were about equally supportive of municipal power companies.1 Their
intention, to hold down monopoly profits, was at odds with Insull’s desire to keep
profits above the competitive level, but both sides agreed that competition was
inefficient and that providing electricity was a natural monopoly.2
1. See Platt (1991) for information on central station companies and Samuel Insull. The quotation (p. 86)
is from a contemporary account of Insull’s address. Platt describes early competition as follows: “The
Chicago experience of rate wars, distributor duplication, and torn-up streets presented an alternative that
was attractive to virtually no one.” For state commissions and progressives, see Rudolph & Ridley (1986).
2. Smith (1995) relies on Gregg Jarrell to conclude “regulation was a response to the utilities' desire to
protect profits, not a consumerist response to monopoly pricing.” But Knittel (1999) tests causation by
utilities and consumers and finds no significant correlation between profit change and regulation after
correcting Jarrell’s endogeneity problem. This result would be expected from an analysis of profit when
two equal forces have opposite motivations with respect to its level.

CHAPTER 1-1 Why Deregulate? 7
On the scale of an isolated city, provision of electricity is a natural monopoly

and requires regulation or municipal ownership, but as transmission technology
developed, it brought new possibilities for trade and competition. The earliest
electric companies, for instance Brush’s company which lighted New York’s
Broadway in 1880, integrated generation with distribution, and, in fact, sold light,
not electricity. Edison initially did the same, installing the light bulbs in the homes
he lit and charging by the number of bulbs installed. Westinghouse introduced high-
voltage transmission using alternating current (AC) technology to the United States
in 1886, and by 1892 Southern California Edison was operating a 10-kV transmis-
sion line 28 miles in length. This, too, was an integrated part of a full-service utility.
Integrated utilities remained natural monopolies for many years while expansion
of the high-voltage transmission network continued, mainly for purposes of reliabil-
ity. Eventually the entire Eastern United States and Eastern Canada were united
in a single synchronized AC power system. By operating at extremely high voltages,
this system is able to move power over great distances with very little loss, often
less than three percent in a thousand miles.
Regulated, vertically integrated utilities were well established by the time the
transmission system made substantial long-distance trade possible. As a conse-
quence, trade was slow to develop, but the existence of the grid made the de-
integration of the electric industry a possibility.3 Generation could now be split
off to form a separate competitive market, while the remaining parts of the utilities
remained behind as regulated monopolies. By 1990, encouraged by a general trend
toward deregulation, the de-integration trend in electric markets was underway
in a number of countries. Today, more than a dozen semi-deregulated electricity
markets are operating in at least ten countries, with several operating in the United
States.
In spite of this apparent success, many fundamental problems remain. After
ten years of operation, the British market has declared itself a failure and replaced
all of its market rules. In a single year, the California market managed to cost its
customers more than ten years of hoped-for savings. Alberta (Canada) is worried
over the results of its recent auction of generation rights that brought in much less
revenue than planned. New York saw prices spike to over $6,000/MWh in 2000,
and the New England ISO had to close its installed capacity market due to extreme
problems with market power. But initial problems do not prove deregulation is
doomed; some markets are functioning well. A closer look at fundamental argu-
3. See Joskow (2000b, 16) for a similar view of the unimportance of “the demise of natural monopoly
characteristics at the generation level” and the importance of the expansion of the grid, and Ruff (1999)
for an alternative view of the role of the grid and transmission pricing.

ments for and against deregulation may help explain why such mixed results might
be expected.
Chapter Summary 1-1 : Improvements in transmission, rather than changes
in generation technology, have removed the natural monopoly character of the
wholesale power market in most locations. This makes possible the replacement
of regulated generation monopolies with deregulated wholesale power markets.
In principle these can be more efficient than the old-style regulation. In practice,
California has proven bad deregulation to be worse than mediocre regulation, and
England has demonstrated that mediocre deregulation can bring cost-saving
efficiencies to a badly regulated generation monopoly.
In the short run, power-market problems tend to be more dramatic than the
benefits. The problems are primarily the result of two demand-side flaws: the almost
complete failure of customers to respond to relevant price fluctuations, and the
customer’s ability to take power from the grid without a contract. As fundamental
as these are, it is possible to design a workable market around them, but it does
require design as well as extensive and clever regulation. Recent U.S. history has
shown that there are three impediments to such progress: politics, special interests,
and overconfidence. The last is largely due to a dramatic underestimation of the
problem.
Section 1: Conditions for Deregulation. Deregulation requires the market
not be a strong natural monopoly. One view holds that small efficient gas turbines
have overturned the natural monopoly of large coal plants. Yet today’s competitive
suppliers are far larger than any coal plant, so if the size of a large coal plant were
problematic for competition, today’s markets would be uncompetitive.
Section 2: Problems with Regulation. Regulation can provide strong cost-
minimizing incentives and can hold prices down, but it must trade off one against
the other. Competition can do both at once. In practice, regulators hold prices down
near long-run average costs but leave cost-minimizing incentives too weak. The
result is high costs and high prices.
Section 3: The Benefits of Wholesale Competition. Competition provides
full strength cost-minimizing incentives and, at the same time, forces average prices
down toward their minimum. It may also encourage efficient retail prices.
Section 4: The Benefits of Real-time Pricing. Competition may induce real-
time pricing, which will reduce consumption during periods of peak demand. This
will reduce the need for installed capacity and, if extensively adopted, should
provide a net savings of about 2% of retail price. Although this could be achieved

easily under regulation, competition will provide some additional incentives, but
their consequences are still unclear.
Section 5: Problems with Deregulating Electricity. Contemporary electricity
markets have inadequate metering. Consequently it does not make sense for load
to respond to price fluctuations, and bilateral contracts cannot be physically
enforced in real time. As a result demand can and sometimes does exceed supply,
and competitive pricing is impossible at crucial times. These flaws result in high
prices that must be limited, and they provide ideal conditions for the exercise of
market power. Electricity markets are also extremely complex and prone to prob-
lems with local market power due to the inadequacies of the transmission system.
1-1.1 CONDITIONS FOR DEREGULATION

Scale economies make it possible for natural monopolies to produce their output
more cheaply than a competitive market would. A 1-MW power plant is not very
efficient, and there is no way to produce power cheaply on this small a scale. A
10-MW power plant can always do better. Efficiency continues to increase signifi-
cantly to about the 100-MW level but ever more slowly beyond this level. It used
to increase to about 800 MW, and it was once assumed that nuclear plants would
be the most economical and their most efficient size would be even greater. If these
economies of scale continued, the cheapest way to provide California with power
would be to build a 25,000-MW power plant and a few smaller ones to handle load
fluctuations. But a large single power plant could not support competition. A
competitive market necessarily utilizes smaller plants and would therefore have
higher production costs. Consumers would have to pay more if a natural monopoly
is forced to operate as a competitive industry with small-scale plants.
Efficiency gains from the operation of multiple plants are another possible
source of natural monopoly. Even if very large plants are not more efficient, large
generating companies may be. A large company can hire specialists and share parts
and repair crews. If multiplant efficiencies continue to large enough scales, a
competitive market would again be less efficient than a monopolist.4
If a monopolist can produce power at significantly lower cost than the best
competitive market, then deregulation makes little sense. The lack of a natural
monopoly is a prerequisite to successful deregulation, or at least, the condition of
natural monopoly should hold only weakly.
Did Cheap Gas Turbines End the Natural Monopoly?

One popular argument for deregulation claims technical progress has recently
nullified the conditions for a natural monopoly in generation. This view assumes
generation had previously been a natural monopoly because the most efficient size
4. See Joskow and Schmalensee (1983, 54) for a discussion of firm-level economies of scale.

power plant was approaching 1000 MW, and new technologies have made 100-MW
plants almost as efficient.
If this argument were correct, then an 1000-MW supplier must in some sense
be a monopolist, and the market must need suppliers that have capacities smaller
than 1000 MW to be competitive. But in this case, deregulation must certainly have
failed in the United States because every market contains suppliers with capacities
exceeding 1000 MW. Yet no one who suggests small efficient plants are a necessary
condition for competition seems worried by the presence of huge suppliers in the
new markets.
The beliefs that the most efficient size power plant must be quite small, and
that competitive suppliers can own many such plants are contradictory. Most likely
the former is incorrect, at least in markets with peak loads of over 5000 MW.
Fortunately, vast transmission grids have made such large markets the norm. When
small efficient plants are necessary for competition, suppliers with total generating
capacity greater than the most efficient size plant should be prohibited. Greater
threats of natural monopoly conditions come from the economies of multiplant
companies and weaknesses in the power grid that effectively isolate “load pockets”
during peak load conditions.
1-1.2 PROBLEMS WITH REGULATION

The most common argument for deregulation is the inefficiency of regulation. There
can be no quarrel about its inefficiency, but it does not follow that deregulation
will be better. Deregulation is not equivalent to perfect competition which is well
known to be efficient. Electricity markets have their own inefficiencies that need
to be compared with the inefficiencies of regulation.
To date, such comparisons have been largely speculative. The most decisive
answers inevitably are based on the least information. One side claims regulation
is essential because electricity is a basic need. What about housing? That need is
even more basic, and housing is 99% deregulated. The other side claims competition
provides incentives to reduce costs, while regulation does not. What about the many
regulated utilities that have provided reliable power for many years for less than
6¢/kWh? Do they lack all cost-minimizing incentives?
One argument posits that when a regulated utility makes a bad investment,
ratepayers pick up the tab; but, when an unregulated supplier makes a bad invest-
ment, stockholders pick up the tab. This analysis is myopic. Particular losses fall
on the stockholders of particular companies, but the cost of capital takes into
account the probability of such mistakes, and every mistake increases the estimated
probability. Like all costs, the cost of capital is paid for by consumers. Not only
does the cost of capital average in the cost of mistakes, it also adds a risk premium.
To the extent stockholders of regulated utilities are sheltered, they demand less
of a risk premium than do stockholders of unregulated suppliers. With competition,
there may be fewer mistakes, but the mistakes will be paid for by consumers, and
a risk premium will be added.

Regulation has two fundamental problems: (1) it cannot provide a strong

incentive to suppliers as cheaply as can a competitive market, and (2) regulatory
bodies themselves do not have proper incentives. Well-trained regulators could
provide much better regulation. But for government to provide competent regula-
tion, the political process would need to change. The first problem, that of incentives
provided by regulators, is more susceptible to analysis.
The Regulator’s Dilemma

Truly competitive markets do two things at once; they provide full-powered
incentives (1) to hold price down to marginal cost, and (2) to minimize cost.
Regulation can do one or the other but not both. It must always make a trade-off
because suppliers always know the market better than the regulators.5
This trade-off is the core idea of modern regulatory theory. Perfect cost-of-
service (COS) regulation is at one extreme of the regulatory spectrum. It assures
that, no matter what, suppliers will recover all of their costs but no more. This
includes a normal rate of return on their investment. Perfect COS regulation holds
prices down to long-run costs but takes away all incentive to minimize cost.6 If
the suppliers make an innovation that saves a dollar of production costs, the
regulator takes it away and gives it to the customer.
At the other extreme is perfect price-cap regulation. It sets a cap on the supplier’s
price according to some formula that takes account of inflation and technical
progress, and it never changes the formula. Now every dollar saved is kept by the
supplier, so its incentives are just as good as in a competitive market. But it’s
difficult to pick a price-cap formula that can be fixed for twenty years at a time.
A perfect (very-long-term) price cap must always allow prices that are well above
long-run cost to avoid accidentally bankrupting suppliers. Consequently, prices
will be too high.
The reader unfamiliar with the theory of regulation may be tempted to invent
clever ways for the regulator to provide full cost-minimizing incentives while
holding prices down to cost, but all will fail. The inevitability of this trade-off has
been established repeatedly and with great rigor; however, the trade-off can be
improved by improving the regulator’s effective knowledge.7 The main technique
for making the trade-off is to adjust the price cap more or less frequently. Constant
adjustment produces COS regulation while extremely infrequent adjustment
produces pure price-cap regulation. In between, incentives are moderately strong
and prices are moderately low. If the regulator has a fair amount of information,
this trade-off can be quite satisfactory, but it will never equal perfect competition.
5. Suppliers having better information than the regulators is at the root of the regulatory trade-off problem
which, though fundamental, is too complex to discuss here.
6. In reality, cost-of-service regulation does provide incentives in two ways: regulatory lag (see Joskow
2000b) which is discussed shortly, and the threat of disallowed costs.
7. For instance, yardstick regulation compares the performance of one regulated firm with other similar
firms, thus giving the regulator some of the benefit of the other firms’ knowledge without requiring the
regulator to know details (see Tirole 1997).

Regulation in Practice
Competition can hold average prices down to long-run costs while putting full
strength pressure on cost minimization. At best, regulation does a decent job of
both but does neither quite as well as competition. But how does regulation work
in practice?
Regulation tends to err in the direction of driving prices down toward cost. In
fact, most regulators believe this is their entire job and would implement pure COS
regulation if they could. Fortunately, it’s just too much bother to re-adjust rates
continuously, so the result is roughly a price cap that gets reset about every three
years. This inadvertent “regulatory lag” is a major factor in saving COS regulation
from providing no incentive at all. It provides some incentive for cost minimization,
but less than would be provided with an optimal trade-off. Even that is too little
by the standard of a competitive market. In practice, regulation has typically done
a passable job in the United States and could do much better if the effort spent
deregulating were spent improving regulation.
1-1.3 THE BENEFITS OF COMPETITIVE WHOLESALE MARKETS

Competition provides much stronger cost-minimizing incentives than typical “cost-
of-service” regulation and results in suppliers making many kinds of cost-saving
innovations more quickly. These include labor saving techniques, more efficient
repairs, cheaper construction costs on new plants, and wiser investment choices.
Distributed generation is an area in which innovation may be much quicker
under competition than under regulation; cogeneration is one example. Regulated
utilities found such projects extremely awkward at best, so avoided them. A
competitive market easily allows the flexibility that such projects require.
The other advantage of competition is its ability to hold price down to marginal
cost. This is less of an advantage simply because traditional regulation has stressed
this side of the regulatory trade-off. Sometimes price minimization can still be a
significant advantage. Again cogeneration provides an example. Once regulators
decided to encourage it, they needed to price cogenerated power. A formula was
designed with the intention of mimicking a market price. Naturally, political forces
intervened and the result was long-term contracts signed at very high prices (Joskow
2000b). These gave strong, probably much too strong, incentives for cogeneration.
A competitive market can get both incentives and prices right at the same time.
While holding down prices, competition also provides incentives for more
accurate pricing. Because it imposes the real-time wholesale spot price on the
retailer’s marginal purchases, wholesale competition should encourage real-time
pricing for retail customers. This can be done easily by a regulated retailer, but a
competitive retailer should have an added incentive to provide the option of real-
time retail pricing because that would reflect its costs.

1-1.4 THE BENEFITS OF REAL-TIME RATES

One view holds that the main benefit of competition will come from the demand
side of the market more than the supply side. The price spikes of the wholesale
market will be passed on to customers—at least for marginal consumption and will
cause customers to curb their demand when the price is highest and generation is
most costly. This will allow fewer generators to be built and will reduce the total
cost of providing power. A competitive market will pass this savings on to consum-
ers.
Arthur Wright of Brighton, England, invented the real-time meter which Samuel
Insull heard of while visiting his homeland in 1894. He sent Louis Ferguson to
Brighton to study its use. Insull soon replaced most of his meters with Wright real-
time meters and by 1898 was installing them with every new residential hookup
(Platt 1991). Although Ferguson invented the notion of charging according to the
time of peak demand, this was rarely used. Instead customers were charged for peak
demand (a demand charge) and total energy. This combination became known
as a Wright tariff. The purpose of such metering was to improve the system’s load
factor (average load over peak load), and it did so. Central station load curves from
Chicago show that their load factor improved from 30.4% in 1898 to 41.7% eleven
years later. The Wright tariff is not real-time pricing but has a similar, though
generally weaker, effect on peak demand.8
The cost-saving effect of real-time pricing cannot be doubted, but how great
is it? Unfortunately it is much smaller today than it was in 1898, and that may
explain in part why residential customers no longer have real-time meters. Today’s
load factors are typically near 60%, although, in a system like Alberta’s where
regulators have imposed heavy demand charges, the “load factor” can approach
80%. Real-time pricing could do the same. It can never cause load to become
completely constant (100% load factor) because this would put an end to real-time
price fluctuations, and there would no longer be a reason for customers to shift
their consumption off peak. Thus load-shifting caused by real-time pricing is self-
limiting. Assume that the shift would proceed halfway if real-time pricing were
fully implemented. Changing the load factor from 60% to 80% gives about a 25%
reduction in needed generation capacity. This is dramatic, and perhaps a bit optimis-
tic, so for numeric convenience, assume the reduction is only 24%.
With an 80% load factor, most load would be baseload because high real-time
prices reduce peak load. Peaking generators cost roughly half of what an average
generator costs per installed megawatt. The reduction in generation fixed costs
should be in the neighborhood of 12%. Because fuel costs are about as large as
the fixed costs of generation, total wholesale power costs will be reduced only about
6%. (Real-time pricing shifts load to off-peak hours but its effect on total energy
consumption is minimal and could work in either direction.) Wholesale costs are
about 3'8 the cost of retail power, so retail costs are reduced by about 2.25%.
8. For a comparison of demand-charges and real-time rates and an explanation of why real-time rates are
crucial for reducing market power, see Borenstein (2001b).

Peak transmission use often occurs during shoulder hours, not peak hours, so
there is no easy proof of a transmission cost savings. But for simplicity assume
that this savings is 1'3 as great as the saving in generation. That brings total savings
to 3%.
Result 1-1.1 Savings from Real-Time Rates Would Be Small

Fully implemented real-time pricing would improve load factors roughly from
60% to 80%. This would reduce the cost of supply by approximately 2.25% of
retail costs mainly due to savings in peak-load capacity. Additional consumer
costs of accomplishing the load shift would likely more than cancel any transmis-
sion savings for a net savings of about 2% of retail costs.
This 3% reduction in supply cost is offset by a cost increase on the demand

side. Shifting load to off-peak hours is costly for customers as it requires the
purchase of smarter appliances and changes in consumption that they find undesir-
able. In fact customers will shift load up to the point where the marginal cost of
shifting is just as great as the marginal savings on their electricity bill. A rough
estimate puts the total cost to consumers at about 1'3 of the savings and this does
not include the cost of the real-time metering and billing. The net savings from real-
time metering should be in the neighborhood of 2% of total cost of delivered
power.9
A second question about real-time pricing as an argument for deregulation is
why it cannot be done under regulation. Regulated systems have for years computed
real-time system marginal cost. Technically, it would be no problem for a regulated
utility to install real-time meters and charge customers marginal-cost prices.
California is now testing this possibility (Wolak 2001). It may be that regulators
simply lack the will to do this or the ability to carry out the details effectively. On
the other hand, it may take more will and cleverness to implement a competitive
wholesale market than to implement real-time pricing.
1-1.5 PROBLEMS WITH DEREGULATING ELECTRICITY

Electricity is a peculiar product. It is the only product that is consumed continuously
by essentially all customers. In fact it is consumed within a tenth of a second of
its production and less than a tenth of a second of power can be stored as electrical
energy in the system.10 These physical properties result in a product whose marginal
cost of production fluctuates rapidly and, thus, whose delivered cost also fluctuates
rapidly. No other product has a delivered cost that fluctuates nearly this rapidly.
9. Significant savings in transmission and distribution should not be expected because transmission lines
are typically used most heavily during off-peak hours, and both have very large fixed-cost components that
will be unaffected by deregulation.
10. More kinetic energy is stored in rotating generators, much more potential energy can be stored by
pumping water up hill and vastly more energy is stored in local fuel supplies, but all of these stores of
energy must be converted to electrical energy by the process of generation before they can be delivered.

Two Demand-Side Flaws

Although real-time metering began in the late 1800s, it has been discontinued for
residential customers, and almost no industrial or commercial customers see real-
time prices. Consequently, almost no customers respond to the real-time fluctuations
in the delivered cost of power.
Even with this demand-side flaw, the market could operate in reasonably close
accord with economic principles if not for the second demand-side flaw, the ability
of a load to “take power from the grid without a prior contract with a generator”
(Ruff 1999, 28; FERC 2001b, 4). If bilateral contracts could be enforced by
physically cutting off customers who exceed their contracts, the market could
function almost in alignment with the theory of competitive markets. In no other
market is it impossible to physically enforce bilateral contracts on the time scale
of price fluctuations.
Demand-Side Flaw 1: Lack of Metering and Real-Time Billing

Demand-Side Flaw 2: Lack of Real-Time Control of Power Flow to Specific Customers
The first demand-side flaw causes a lack of demand responsiveness to price

or, technically, a lack of demand elasticity. The second demand-side flaw prevents
physical enforcement of bilateral contracts and results in the system operator being
the default supplier in real time.
Because demand responds only minimally to price, the supply and demand curve
may fail to intersect, a market flaw so severe it is not contemplated by any text on
economics.11 The system operator, as default supplier, is forced to set the price,
at least when supply fails to intersect demand. It can also improve the market by
setting price under slightly less dire circumstances. Presently, all power markets
operate like this and will continue to do so until very-short-run demand elasticity
is significantly improved.
These are not just theoretical problems. While the average cost of production
is about $35/MWh, and the maximum cost with new equipment is about $100/MWh,
prices in the $1,000 to $10,000 range have occurred in many markets. All four of
the U.S. markets have formal price caps, and the Midwest market has informal caps
set by system operators who refuse to buy required reserves when the price gets
too high. With the extreme demand inelasticity caused by the first flaw, scarcity
alone would produce high prices, but the flawed demand side coupled with scarcity
also produces ideal conditions for market power which pushes prices still higher
(Joskow 2001a).
Dramatic though these flaws are, it should still be possible to design a well-
functioning market. But it does require design! Deregulating power markets is called
“restructuring” in the United States because the resulting competitive markets have
more federal regulations than the regulated markets they replaced (Borenstein and
Bushnell 2000). In the long run, the demand side of the market should develop
enough price elasticity to clear the market at a finite price. There is a good chance
11. See Jaffe and Felder (1996), Kahn et al. (2001), and Green (1998).

that eventually price spikes will be low enough to need no caps. This change in
market structure should be encouraged from the start. A responsible deregulation
of electricity would first fix the demand-side flaws and then start the market—they
are cheaper to fix than the problems they have already caused.
Complexity and Local Market Power

There are two other fundamental problems with deregulating electricity: complexity
and local market power. A power system is a delicate, single machine that can
extend over millions of square miles. Every generator in the system must be
synchronized to within a hundredth of a second with every other generator in the
AC interconnection. Voltage must be maintained within a 5% limit at thousands
of separate locations. This must be accomplished on a shared facility, half of which
(the grid) must be operated for the common good and half of which (the generators)
are operated for hundreds of different private interests.
Complexity can be overcome by a sufficiently well designed set of market rules,
but the problem of local market power may need to be solved, at least for the
present, by interventionist means. So far, it has been. More than half of the genera-
tors in the California ISO were declared “must run,” meaning sometime during
the year they were crucial to system operation and, therefore, had such extreme
market power there was no choice but to regulate their price. San Francisco and
New York among other cities are “load pockets;” they require more power than
they can import. As a consequence, the two generators in San Francisco would
have extreme market power during peak hours every day if they were not regulated.
These generators are required for their “real” power production, but most must-run
generators are required for their “reactive” power, a concept not well understood
by regulators (see Chapter 5-2).
The most difficult and costly problems with new electricity markets are mainly
matters of market structure as opposed to market architecture (see Chapters 1-7
and 1-8). When this is understood, and demand-side flaws and the problems with
market power and transmission are squarely faced, adequate solutions will probably
be found. Then wholesale power markets should prove superior to regulated
monopoly generation.

Chapter 1-2
What to Deregulate
Nothing is more terrible than activity without insight.
Thomas Carlyle
(1795-1881)
D ELIVERED POWER IS A BUNDLE OF MANY SERVICES. These include

transmission, distribution, frequency control, and voltage support, as well as
generation. The first two deliver the power while the second two maintain power
quality; other services provide reliability.
Each service requires a separate market, and some require several markets. This
raises many questions about which services should be deregulated and which should
not. Even within a market for a single service, one side—either demand or
supply—may need to be regulated while the other side of the market can be deregu-
lated. For instance, the supply of transmission rights must be determined by the
system operator, but the demand side of this market is competitive. In contrast,
the demands for ancillary services are determined by the system operator while
the supply sides of these markets can be competitive.
The most critical service in a regulated or a deregulated power market is that
provided by the system operator. This is a coordination service. For a deregulated
market it typically includes operation of the real-time markets and a day-ahead
market. These provide scheduling and balancing services, but operating these
markets is itself an entirely separate service. While the need for the system operator
service is agreed to by all, the proper extent of that service is the subject of the
central controversy in power market design.
Chapter Summary 1-2 : Many services are required to bring high-quality reliable
power to end users. Each might be provided by free markets, by the state, by
regulated suppliers, or by some hybrid arrangement. Bulk power generation is the
source of nearly half the cost of retail power and is one of the services most easily
provided by a competitive market. Moreover, it seems to offer several possibilities

for significant efficiency improvements. If restructuring gets this much right, it will
reap most of the benefits presently available from deregulation of electricity.
Markets for reserves may be the next most sensible targets for deregulation
because they are so closely tied to the bulk power market. Reserves are provided
by the same generators that provide power. These markets can only be deregulated
on the supply side, and they are quite complicated to design. Retail deregulation
is easier to design, and retailing can be more fully deregulated, but there is much
less to gain. Retail costs are at most 5% of total costs, and the potential savings
are less obvious than in generation because most of the traditional retail services
are irrelevant. This is also the one area in which competition can be expected to
add a significant cost—the cost of marketing.
Section 1: Ancillary Services and the System Operator. The system
operator must keep the system in balance by keeping supply equal to demand. As
many as five markets may be required to accomplish this: one for “regulation”
which works minute by minute, and four to handle larger deviations and emergen-
cies. Collectively, these services, together with a few others, are known as ancillary
services.
Ancillary services benefit the entire market and are either public goods or have
large external effects. Consequently, all of the markets have a fully regulated
demand side, but some can be deregulated on the supply side. The system operator
service, which coordinates these markets and provides the regulated demand for
ancillary services, is a natural monopoly service and can be provided by a nonprofit
or a for-profit entity.
Section 2: Unit Commitment and Congestion Management. Traditionally
(before restructuring), system operators provided two more coordination services:
unit commitment and transmission-congestion management. Unit commitment can
be left to the generators themselves, though this may result in a small decrease in
efficiency. Transmission congestion must be managed, at least in part, by the system
operator.
Section 3: Risk Management and Forward Markets. Both generators and
their customers are risk averse and wish to avoid the fluctuating prices of the spot
market. Not only the hour-to-hour and day-to-day price variations but also the year-
to-year variations in the spot market’s average price are problematic. Forward
energy trading can hedge these risks and needs no more regulation than other
commodity markets.
Section 4: Transmission and Distribution. Delivery of electric power
requires a network of high voltage lines. Because duplicate sets of lines are wasteful,
both distribution and transmission appear to be natural monopolies. There has been

CHAPTER 1-2 What to Deregulate 19
speculation about granting rights to congestion rents to investors in transmission

lines and thereby stimulating a competitive market for such lines, but currently no
jurisdiction plans to rely on such a market.1
Section 5: Retail Competition. Retail service in electricity does not include
distribution of power to the customer. In its present form, it usually does not include
metering or even meter reading. Most retailing consists of financial transactions,
all of which can be provided from an office building. There are relatively few
possibilities for value added. Early on, deregulation advocates claimed that retailers
would provide innovative products (new types of meters and new qualities of
electricity), but now their claims of innovation focus on billing.
The use of green power and time-of-use billing seem to be the most substantive
improvements available in the retail market, but there is little reason to believe these
could not be provided by a regulated market. In fact, green power might be more
easily provided under regulation. Unfortunately, because customers will want or
require the right to switch suppliers, retail competition may make it more difficult
for generators to sign long-term contracts. This could make the wholesale market
more risky and more susceptible to the exercise of market power.
1-2.1 ANCILLARY SERVICES AND THE SYSTEM OPERATOR

The system operator must keep the system in balance, keep the voltage at the right
level, and restart the system when it suffers a complete collapse. The system
operator carries out these basic functions by purchasing what are called “ancillary
services.” These include various types of reserves, voltage support, and black-start
services. Ancillary services are the subject of Chapter 3-2 where they are defined
more carefully and more broadly.
Regulation and Balancing

Any imbalance between supply and demand causes the system frequency to deviate
from the standard (60 Hz in the United States).2 This is problematic for some
appliances; nondigital clocks and phonographs depend on the system frequency,
but large generators depend most on a constant frequency. Changing frequency
makes these enormous machines speed up or slow down, causing added wear and
tear. Consequently, maintaining a precise system frequency is given high priority
1. Markets for transmission rights that cover the cost of congestion already exist but not competitive
markets for transmission lines. Australia and Argentina are experimenting with for-profit transmission
lines. A market for monopoly transco franchises is another possibility (Wilson 1997).
2. Frequency is very precisely maintained as indicated by the following report (NERC 2000, 32). “In late
July 1999 . . . system frequency on the Eastern Interconnection dipped to one of its lowest levels in history
(59.93 Hertz).”

by the system operator. There seems to have been no cost-benefit analysis, but
balancing the system with some accuracy is absolutely necessary.
System frequency is exactly the same for all customers in an interconnection.
For example, Quebec and Florida have exactly the same frequency. As long as any
system operator maintains its frequency, every other system in the Eastern Intercon-
nection will also have exactly the right frequency. As a result, no one has much
interest in trying to maintain frequency. Economics calls this a “free-rider” problem.
It is costly to maintain the system balance, and everyone would prefer to let the
others take care of it. Consequently, some regulatory body must decide to purchase
balancing services and must charge (tax) some market participants to pay for them.
There is no alternative; with a pure free market approach, the system frequency
would be unacceptably unstable. This means the demand for balancing services
must be regulated, but they still can be supplied by a competitive market. Every
system operator must meet a balancing standard set by the North American Electric
Reliability Council (NERC), so it purchases enough balancing services to meet
this standard. Because many different generators can supply the services, there is
a good chance the market will be competitive. But it is not automatic; it depends
on the details of the market.
Self-Provision of Reserves. There are two approaches to the supply of balancing
services. The system operator can purchase the necessary services in a market and
assign the cost to either loads or generators on a pro rata basis, or it can assign each
supplier a fraction of the physical requirement. The latter approach is called “self-
providing” and is often described as a less regulated approach.3 As an example,
consider spinning reserves (spin), a key source of balancing services. Not every
generator will want to provide spin, so if physical requirements are imposed on
suppliers, a market for spin will develop. Those who find it expensive to provide
will buy from those who can provide it more cheaply, and some generators will
“self-provide.”
What is the practical difference between the physical “self-provision” approach
and the financial approach? Imagine that a supplier has been told to supply 100
MWh of spin as part of its contribution to balancing services, and that it “self-
supplies.” If the system operator had instead purchased spin directly and had
charged the supplier for 100 MWh of spin, would the supplier have been worse
off? No. It would have sold its 100 MWh of spin to the market at the spin-market
price, P, for a revenue of 100 × P, and then the system operator would have charged
it 100 × P for its share of the spin requirement. The result is the same. In both cases
the generator provides 100 MWh of spin. In neither case does it have any net cost.
So why the agitation to “self-provide?” There are several possible answers. First,
some may not understand the concept. Others may believe they can more easily
supply poor quality spin when they self-provide. Still others may not be interested
in their own requirement but may wish to provide part of the system-operator
service. In other words, they may want to be the market maker for spin and take
this business away from the system operator. If they can do it more efficiently and
3. See Chao and Wilson (1999a) and Wilson (1999).

with less gaming than the system operator, then they should be allowed to take over
this service.
The problem with “self-providing” is that the services are not provided to the
suppliers themselves but to the system operator. The term is deceptive. There still
must be a physical transaction between the supplier and the system operator. It is
this physical transaction, the verification of the provided reserves, that is difficult.
“Self-provision” is not a way to avoid the regulated nature of the demand side of
the ancillary service market, and it does not make the supply side more competitive.
Voltage Support and Black-Start Capability

Most generators need to take electric power from the grid in order to start them-
selves. Consequently, if the system goes down, they cannot help it restart. Special
generators have the ability to self-start. Like balancing services, black-start services
are a public good and must be purchased by the system operator.
Voltage “sags” when too much “reactive” power is taken out of the system.
Reactive power, unlike the “real” power that lights up incandescent lights, does
not travel very well over power lines. When too much is used locally, for example
by motors, fluorescent ballasts, and transmission lines, the voltage sags locally.
To counteract this, more reactive power must be injected locally by capacitors,
normal generators, or special generators called synchronous condensers.
Reactive power is less of a public good than the other ancillary services. It is
possible, though expensive, to measure its use by individual customers and by the
grid. In principle, these could be charged for their use and there could be a spot
market in reactive power. This would be a complex and expensive market to run
although reactive power is usually very cheap to produce. Also, because of the
difficulty with transmitting reactive power over long distance, there would be far
less competition in the reactive power market.
Voltage support, like other ancillary services, has a free-rider problem unless
all customers are metered for its consumption. Customers lack sufficient incentive
to replace the reactive power they use because the voltage drop caused by one
customer affects many others. Consequently, at a minimum, there is a need for a
regulatory requirement for reactive power purchases. For the present, by far the
simplest approach is for the system operator to buy what is needed and, when
necessary because of market power, to regulate the purchase price.
The System Operator Service

The system operator service, which coordinates the ancillary service markets and
provides the regulated demand for ancillary services, is a monopoly service (Ruff
1999). This monopoly can be either a nonprofit or a for-profit entity. If nonprofit
it is typically called an independent system operator (ISO) in the United States
and can be minimally regulated. Transco is the current U.S. term for a for-profit
system operator, an entity that will probably own the grid and will require extensive
regulation.

Nonprofit “monopolies” are quite different from for-profit monopolies. For

instance, an ISO will have no motive to maximize the rent from congested transmis-
sion lines because it cannot keep that rent as profit. An unregulated for-profit system
operator could have an extremely strong motive to maximize rent by withholding
transmission service. Because such behavior is both unfair and inefficient, the for-
profit system operator will need to be regulated.
One opinion holds that the transmission grid will face competition from distrib-
uted generation (small local generators) and is, therefore, no longer a natural
monopoly. This was the case in 1890, but now such competition is extremely weak.
Imagine a modern city being deprived of its importing transmission lines, and ask
if this would raise its cost of power if it were forced to rely on within-city distributed
generation. It would be enormously expensive. The trend toward increased reliance
on transmission has been apparent for more than a century, and deregulation has
only accelerated this trend. New micro-generator technology has increased the
chances for distributed generation to slow or reverse this trend, but the transmission
system will not lose its natural monopoly character for decades to come. Until then,
transcos must be regulated.
Although ISOs have no motive to extract monopoly rents, they do have a
weakened motive to act efficiently. Their motivation comes mostly from public
scrutiny which is enhanced by the attention of market participants (“stakeholders”)
who have a lot to lose from inefficient ISO operation.
For the present, the system operator service must be provided by a regulated
for-profit monopoly transco or by an lightly regulated nonprofit ISO. Both are poor
choices and may prove to have fairly similar problems. The Transmission Adminis-
trator in Alberta, Canada, a for-profit regulated monopoly, is regulated in such a
way that it behaves like a nonprofit in many respects. For example, it proposed
substituting large incentive payments and charges on generators for the building
of a large new transmission line. While the line may not have been justified (though
now it claims it is), the incentives were structured in such a way that they appeared
to make money for customers. The costs of inducing generation investment in more
costly locations were both hidden and denied. The monopoly transco’s incentive
to optimize transmission was extremely weak to nonexistent. Instead, it exhibited
the same interest as an ISO in gaining favor with the public and stakeholders.
There is one possible way out of this dilemma. Monopolies may sometimes
be forced to compete by granting them a temporary franchise which can be won
through a bidding process. Robert Wilson (1997) has proposed doing this and has
suggested a scheme for providing them with incentives for efficient system opera-
tion. Although this approach shows promise, the technical and political problems
with implementation are still formidable.
1-2.2 UNIT COMMITMENT AND CONGESTION MANAGEMENT

There is nearly universal agreement that the system operator must run the real-time
market, and that the demand for ancillary services must be determined by a central
authority. But the services of unit commitment and congestion management have
caused a great controversy. This is often referred to as the “Poolco” or “nodal

pricing” debate. Professor William Hogan of Harvard has put forward the nodal-
pricing model which has now been accepted by PJM and the New York, and New
England ISOs. California, Australia, Alberta, and England have rejected it in favor
of a less centralized approach.
Hogan (1992, 1995, 1998) specifies a system operator that provides both unit
commitment and congestion management services.4 These two are linked together
more because of historical system operator practice or by choice than by logic. In
fact, the system operator could easily provide only congestion management and
balancing, but not unit commitment.
The Unit Commitment Service

The unit commitment service need not be centrally provided nor even be provided
by a private market. Each generator can provide it for itself. This service simply
tells generators when to turn on, how much to produce at each point in time, and
when to turn off. These decisions can be made privately. The question is, how
efficiently will they be made.
If decisions are private, individual generators will predict the price of energy
during the time under consideration and will then incur the cost of starting up if
they expect to make a profit. When shutting down appears to be more profitable
than running, they will shut down. Economic analysis shows this process works
almost perfectly if price predictions are perfect and there are no costs to starting
and stopping. In real markets, however, neither assumption applies. This raises
two questions: (1) would the system operator predict prices more accurately, and
(2) can the system operator work around the problem of startup costs more effi-
ciently?
The system operator can probably predict market price slightly better than
individual suppliers because it has more information, but this advantage may not
be great. Under nodal pricing, the system operator can predict the day-ahead price
when it schedules, but it is the unknown real-time price that matters.
The inefficiency described by economic theory and caused by “nonconvex”
startup-costs is almost certainly very small.5 (This is discussed at length in Chapter
3-8.) Although startup costs probably do little damage to the market equilibrium,
they make it more difficult to find. This might be a greater source of inefficiency,
particularly if this problem interacts with reliability. Currently, there is little, if any,
evidence that the system operator can save much by performing the unit commit-
ment service. But there is little to lose and perhaps a noticeable gain, especially
in a new market, so the safe course may be to have the system operator provide
the service and make it optional. This is now done by PJM and almost all producers
make use of the service. But the implications are unclear because of indirect
incentives to accept PJM’s service.
4. More recently Hogan (2001b) has explicitly suggested designs that make the unit commitment service
optional.
5. No-load costs, which are greater than startup costs, also cause the production cost function to be
nonconvex, but they apparently cause less trouble.

Transmission-Congestion Management
Congestion management is one of the toughest problems in electricity market
design. Although the costs imposed by congestion in an efficiently run system are
quite low, badly designed congestion pricing can make the system unmanageable.
This was demonstrated by PJM just before the official opening of its market when
it instituted a form of average-cost congestion pricing that resulted in massive
gaming of the pricing rules (Hogan 1999).
Any efficient method of congestion management will charge for the use of
congested lines. If the price charged is set correctly, the demand for the use of the
line will equal its capacity unless the price is zero, and it will be zero only if at zero
price the line is still not fully utilized. Pricing ensures that those who value the line
most get to use it and that the line’s capacity is not wasted. This is the only efficient
way to manage congestion.
Although there is only one set of efficient congestion prices (the prices set by
the “nodal pricing” approach), there are many other approaches that would in theory
give these same prices, or a good approximation of them. Wu and Varaiya proposed
the most extreme alternative to nodal pricing. In their “multi-lateral” approach (Wu
and Varaiya 1995; Varaiya 1996), the system operator would have no knowledge
of the congestion prices. It would simply assign certain parties the right to use lines
in a “reasonable but arbitrary way” that would prevent overuse of the lines. These
parties would then either exercise their rights by using the lines or would sell their
rights to those who valued the lines more. If it worked efficiently as a competitive
market should, the market would then produce the exact same prices for the use
of congested lines as the system operator would compute under nodal pricing.
Congestion prices make money by charging transmission users for a scarce
resource. The revenues collected are called the congestion rent. Because congestion
prices are the same under any efficient system, so is the congestion rent. Under
nodal pricing, the system operator collects the congestion rent. Under other systems,
such as the multi-lateral approach, some private party will collect these rents. This
is at the heart of a controversy. When generators ask to “self-manage” their conges-
tion, they are really asking to “self-collect” as much of the transmission rent as
possible.
Any system of allocating transmission rights can be thought of as selling or
giving away transmission rights. In either case the rights originate with the system
operator and are specified in such a way that the transmission lines are protected
from overuse. The supply of transmission rights must always be regulated.
There is a fundamental choice to be made: should the system operator sell the
transmission rights and use the revenues (congestion rents) to help pay for the cost
of the wires, or should the system operator give the rights away and let private
parties pocket the congestion rents?6 Where physical rights have been purchased
in the past or been acquired by paying for lines or upgrades, grandfathering these
rights, and perhaps converting them to financial rights provides regulatory continuity
6. If the system operator sells transmission rights and collects the congestion rent, it can sell the rights on
a daily or hourly basis in the form of nodal pricing, or it can sell yearly or monthly transmission rights in
periodic auctions, or these can be combined or supplemented with other approaches.

and enhances regulatory credibility. Other schemes for giving away new rights are
usually obscure and arbitrary. Except when there is clear ownership of rights to
begin with, it is preferable for the system operator to create transmission rights,
sell them, and use the resulting revenues to pay for present or future transmission
lines.
1-2.3 RISK MANAGEMENT AND FORWARD MARKETS

The system operator needs to manage the system in real time to keep it physically
secure. It also needs a small lead time for planning (scheduling). This lead time
is generally accepted to be approximately a day, but beyond that, the system
operator has no need to pay attention to the energy market (transmission and
generation capacity is another matter).
Both generators and their customers will want to make long-term arrangements
for the supply of power either in decentralized forward markets or in highly central-
ized futures markets. The futures markets will work best if subject to the normal
regulations imposed on commodities markets, but electricity futures need no special
treatment. These markets should be just as deregulated as any other commodity
market.
Transmission rights also need forward markets. (See Hogan, 1997, for one
market design.) As with congestion management, the system operator may play
a role in the supply of transmission forwards because ultimately it must decide how
much power will be allowed to flow on the various transmission lines. Unfortu-
nately, transmission rights are extremely complex because, in present transmission
systems, it is impossible to choose the path over which power will flow. If there
is a line from A to B and a trader owns only the right to transmit 100 MW over that
line, the trader will not have sufficient rights to make any trade at all. If even 10
MW is injected at A and removed at B, a significant fraction of that power, some-
times more than half, will flow on other lines. Where the power actually flows is
determined by the laws of physics.
Fortunately, designing forward transmission rights as purely financial rights
can simplify the problem of physical flows. A transmission right for 100 MW can
confer on its owner the right to the congestion rent from A to B times 100 MW.
If the congestion rent averaged $10/MWh during peak hours, and the right covers
100 MW of flow for 320 peak hours, the owner receives $320,000. With such rights
the owners can perfectly hedge a transaction from A to B. The markets for such
rights need no special regulation, but they may not be liquid enough to serve the
needs of power traders fully. Much work remains to be done in designing better
hedges for congestion costs.
1-2.4 TRANSMISSION AND DISTRIBUTION

If the congestion rent is paid to the owners of transmission lines, too few lines will
be built. The market for transmission lines cannot be left unregulated, and even
a regulated market may not be up to the job. One way to get lines built is to have

an ISO build them. Another way is to grant one company, a transco, the monopoly
right to all lines in some large geographical area. Like any monopoly of an essential
product, the transco will need regulation.
Several problems need to be considered. Transmission lines can act as both
a substitute for and a complement to generation. New transmission lines increase
competition between suppliers who may therefore oppose them, but they are a
public good because they reduce market power. The siting of a new transmission
line is a highly regulated and contentious process. It is also difficult to assign
individual physical or financial rights to the power grid in such a way that investors
make the appropriate return on their investment.
These complexities make deregulation of the market for transmission lines
impossible. The distribution system seems even more difficult to deregulate, and
so far there have been few, if any, proposals to do so.
1-2.5 RETAIL COMPETITION

The push to deregulate generation was clearly predicated on reducing the cost of
generation which accounts for nearly one half the cost of power. Wholesale compe-
tition could save a lot of money; retail competition needs a different rationale. When
the costs of the electricity industry are analyzed, they are traditionally divided into
three major categories: generation, transmission, and distribution—retail is not
mentioned. Retail costs could be cut in half, and no one would notice as they are
only a small fraction of distribution costs.7
What is retailing? It is not distributing power at the local level or even hooking
up individual customers. It is typically only financial transactions and sending out
bills; occasionally it involves meter reading. Generally, a retailer buys wholesale
power, signs up retail customers, and sends out bills. Although an individual retailer
may manage to purchase power cheaply, on average a retailer will pay the average
cost of wholesale power. Also, there is no reason to believe that competition on
the demand side of this market will reduce the cost on the supply side. There may
be room to cut billing costs, but there are other motives at work in the push for retail
competition.
The impetus for retail competition comes primarily from two sources: those
who believe they can profit by being retailers, and big commercial and industrial
customers. Some of them believe they are smarter or more desirable customers
and so can cut a better deal on their own. These motivations, though strong, do
not translate into politically persuasive arguments, so more theoretical explanations
are proffered. These fall into three categories: customer choice, innovative products,
and price competition.
Customer Choice
Although those pushing hardest for retail competition are not particularly “green,”
their main example of customer choice is that of green power. This could have been
7. According to Joskow (2000a, 29), the total cost of retailing services amounts to between 3.3% and
4.7% of total retail revenue. A reduction by half would at most amount to a savings of 2.4%.

provided quite easily under regulation, but it wasn’t, so the green choice may be
a benefit of retail competition. It allows customers to pay a premium for their energy
and then makes sure some of that premium finds its way to generators that use
renewable energy sources.
One disadvantage of retail competition is that it makes verifying the reality of
this product much more difficult. Physically, green power is identical in all respects
to coal power. The consumer cannot verify the product without the help of some
third party to audit the suppliers. The source of power is more difficult to monitor
when there are many competitive green and semi-green private companies instead
of just one regulated utility.
Another choice, often suggested but so far not implemented, is the choice of
reliability level. Clearly, some customers have a much greater need for reliability,
and this is the one way in which electricity service can differ between customers.
It is not possible to provide a more stable frequency or better voltage support on
a customer by customer basis, but some reliability (avoidance of some blackouts)
can be provided individually. This would require installing remotely controlled
individual circuit breakers. Unfortunately this is currently quite impractical. Part
of the problem is that those with a strong financial interest in this service are those
who want more reliability. This cannot be provided by installing a breaker for them;
it must be provided by installing breakers for a great many others who would be
sacrificed in an emergency to provide reliability for those who would pay for it.
Innovative Products
One view holds that competition in retail electricity will spawn innovative new
electrical products the way competition in telecommunications has spawned new
types of phones and phone services.8 But so far AC power has shown no prospects
of becoming wireless. Power engineering went through a period of rapid innovation
between 1870 and 1910 that was similar to the current innovation in telecommunica-
tions, and someday it may again. Technology has a life of its own. Competition
may spur some development of new generating technologies, but the basic AC outlet
will be around for a while.
Those who promised dramatic new physical innovations in retail power five
years ago have now shifted their focus from new kinds of electricity meters and
smart appliances to innovative billing systems. Would they have held telecommuni-
cations up as an example of the benefits of deregulation if the last fifteen years had
produced only more complex phone bills while telephones remained rotary dial?
The argument for the value of billing innovations is a curious one. In England,
electricity deregulation has brought some of the same aggravation to electricity
shopping that U.S. customers experience in shopping for phone service. Littlechild
argues that the fact that customers spend so much effort comparing rates proves
they must benefit from doing so.9 Perhaps customers draw a different conclusion
8. Though competition has spurred innovation in telecom, if competition had arrived forty years earlier,
we would not have had vacuum-tube cell phones in the 1950s. Much of what is taken to be the miracle of
telecom competition is just normal technological progress.
9. See Littlechild (2000, 11).

than economists because they understand their efforts are as much to avoid bad
rates that were previously not a danger as to find good rates that were previously
not available.10 In any case, other than real-time pricing, it is hard to imagine how
better electricity bills can bring much improvement to the life of the average
customer.
Price Competition
The latest rationale for retail competition is the benefit of price competition. Even
if customers don’t think fancy electric bills are as useful as cell phones, cheaper
bills would certainly prove popular. Littlechild (2000, p. 9) pushes price competition
as the major benefit of retail competition and lists two ways this can happen, first,
by “reducing the costs of retailing.” Unfortunately, as already noted, the costs of
retailing electricity under regulation are exceptionally small. Even with massive
improvements in efficiency, customers will probably not notice a difference, and
it is quite possible that marketing costs will more than offset any efficiency gains.
Littlechild also mentions a reduction in wholesale costs caused by “improving
wholesale power procurement.” Where does this savings come from: better genera-
tors, better operation, or a reduction in market power? With or without retail
competition, generators still keep every dollar they save, no more and no less, so
retail competition cannot possibly improve the cost-minimizing incentives of
generators. With costs the same, the only room for savings is a reduction in profit.
Because the industry cannot survive on below normal profits, this can only come
about if there is some excess profit at the outset.
So retail competition can only lower wholesale costs by reducing the market
power of wholesalers. But market power on the supply side of the wholesale market
would normally be reduced by an increase in competition on the supply side, not
by increased competition on the buyer’s (retailer’s) side. In fact more competition
on the buyers’ side means less monopsony power to counteract the monopoly power
of the suppliers.
There appears to be only one small chance for retail competition to inhibit
wholesale market power. If retailers sign long-term contracts with wholesalers,
the contracts will inhibit the wholesalers’ market power in the spot market. Unfortu-
nately, competitive retailers have a more, not less, difficult time signing long-term
contracts than the monopoly utility-distribution companies (UDCs) because their
customer base is less stable. The customers of the UDC can’t leave if there is no
retail competition, but customers of competitive retailers generally can leave on
fairly short notice and will leave if they do not like the current price. This makes
it difficult for competitive retailers to sign long-term power contracts, especially
for the supply of residential customers. If the retailer buys its power with a long-
term contract and sells it to customers under short-term contracts, it takes an
enormous risk. If the market price goes down in the future, the retailer must either
10. Recently, and long after this section was written, this phenomenon was documented in the PJM market
by Michael Rothkopf, who was the recipient of an offer to pay 38% above the going rate with only extra
cancellation penalties as compensation. In California deceptive practices similar to those found in phone-
rate competition were quickly spotted in the retail power market.

sell at a loss to its residential customers or lose them and resell its power at a loss
to someone else.
The net result seems to be that retail competition offers no benefits in reducing
wholesale market power. As it will not bring down the costs of generation, it seems
to hold little promise of improving wholesale performance. The slim hope that price
competition will save more on billing costs than it spends on marketing is a flimsy
basis for such a large experiment.

Chapter 1-3
Pricing Power, Energy, and Capacity
It is not too much to expect that our children will enjoy in their homes
electricity too cheap to meter.
Lewis L. Strauss
Chairman, Atomic Energy Commission
1954
P OWER IS THE RATE OF FLOW OF ENERGY. Similarly, generating capacity,

the ability to produce power is itself a flow. A megawatt (MW) of capacity is worth
little if it lasts only a minute just as a MW of power delivered for only a minute
is worth little. But a MW of power or capacity that flows for a year is quite valuable.
The price of both power and energy can be measured in $/MWh, and since
capacity is a flow like power and measured in MW, like power, it is priced like
power, in $/MWh. Many find this confusing, but an examination of screening curves
shows that this is traditional (as well as necessary). Since fixed costs are mainly
the cost of capacity they are measured in $/MWh and can be added to variable costs
to find total cost in $/MWh.
When generation cost data are presented, capacity cost is usually stated in $/kW.
This is the cost of the flow of capacity produced by a generator over its lifetime,
so the true (but unstated) units are $/kW-lifetime. This cost provides useful informa-
tion but only for the purpose of finding fixed costs that can be expressed in $/MWh.
No other useful economic computation can be performed with the “overnight” cost
of capacity given in $/kW because they cannot be compared with other costs until
“levelized.” While the U.S. Department of Energy sometimes computes these
economically useful (levelized) fixed costs, it never publishes them. Instead it
combines them with variable costs and reports total levelized energy costs.1 This
is the result of a widespread lack of understanding of the nature of capacity costs.
Confusion over units causes too many different units to be used, and this requires
unnecessary and sometimes impossible conversions. This chapter shows how to
make almost all relevant economic calculations by expressing almost all prices and
1. In Tables 14 through 17 of one such report (DOE 1998a) the useful (amortized) fixed costs are not
reported, and the fixed O&M costs are reported in $/kW which may be an amortized value reported with
the wrong units or, if the units are correct, may represent a misguided conversion of an amortized cost to
an “overnight” cost.

CHAPTER 1-3 Pricing Power, Energy, and Capacity 31
costs in dollars per megawatt-hour ($/MWh). The remainder of the book confirms
this by working every example in these units.
Chapter Summary 1-3 : Energy is measured in MWh, while power and capacity
are measured in MW. All three are priced in $/MWh, as are fixed and variable costs.
Other units with the same dimensions (money divided by energy) may be used,
but this book will use only $/MWh. Screening curves plot average cost as a function
of capacity factor. The slope of the curve is variable cost, and the intercept is fixed
cost. The average cost (ACK) plotted in these graphs is not the average cost of using
a megawatt-hour of energy produced at a certain capacity factor but rather the
average cost of a megawatt-hour of generating capacity. Because the equation for
a screening curve is used through the book, understanding this distinction is crucial.
Section 1: Measuring Power and Energy. Power
Working Summary is the flow of energy and is measured in watts (W),
Readers wishing to gain only a working knowledge kilowatts (kW), megawatts (MW), or gigawatts (GW).
of measurement units for use in later chapters
should understand the following. Energy is an accumulation of power over a period of
Quantity Quantity units Price Units time. For instance, a kilowatt flowing for one hour
Energy MWh $/MWh delivers a kilowatt-hour (kWh) of energy. The price
Power MW $/MWh of both energy and power is expressed in $/MWh. It
Capacity MW $/MWh can also be expressed in “mills,” short for “milli-dol-
lars per kilowatt hour,” with 1 mill equal to $1/MWh.
Cost Symbol Cost Units
Fixed FC $/MWh Section 2: Measuring Capacity. Capacity is the
Variable VC $/MWh
potential to deliver power and is measured in mega-
watts. Like power, it is a flow.
Average ACK = FC + cf × VC $/MWh
Average ACE = FC ' cf + VC $/MWh Section 3: Pricing Capacity. “Overnight” capacity
costs are measured in $/kW and so cannot be added
Ratio Symbol Units
to or averaged with variable costs to find which gener-
Capacity factor cf none
ator could more cheaply serve load of a specific dura-
Duration D none tion. Screening curves plot the annual revenue require-
Notes: Energy is a static amount while power and ment (ARR) of a generator as a function of the genera-
capacity are rates of flow. The average cost of using
capacity, ACK, depends on the capacity factor, cf ,
tion’s capacity factor. Fixed cost (FC) is the value of
which is the fraction of time the capacity is used. The ARR for a capacity factor of zero. Since ARR is mea-
average cost of energy, ACE, produced by a specific sured in $/kWy, the same must be true of fixed cost.
generator also depends on cf . Dividing FC by 8.76 converts it to $/MWh, a more
convenient set of units. Considering the rental cost of
capacity makes these units seem more natural.
To avoid confusion when using screening curves and their associated algebra,
the distinction between the average cost of capacity (ACK ) and the average cost
of energy (ACE ) should be kept in mind. Traditional screening curves graph ACK.

1-3.1 MEASURING POWER AND ENERGY
Power Versus Energy

Power is the rate of flow of energy. This is true for any form of energy, not just
electricity. If you wish to boil a cup of water you need a quantity of energy to get
the job done, about 30 watt-hours. Any specific power level, say a thousand watts
(kilowatt, or kW), may or may not make you a cup of
Unit Arithmetic tea depending on how long the power continues to
Units—kilowatts, hours, and dollars—follow the flow. A typical microwave oven delivers power at a
normal laws of arithmetic. But it must be understood rate of about 1 kW (not 1 kW per hour). If it heats your
that a kWh means a (kW × h) and a $ per hour water for one second, the water will receive power at
means a ($/h). the rate of one thousand watts, but it will gain very
Also note that “8760 hours per year” has the value little energy and it will not make tea. Two minutes in
of 1, because it equals (8760 h) ' (1 year), and the micro-wave will deliver the necessary energy, 1'30
(8760 h) = (1 year). of a kWh.
As an example, $100/kWy = Confusion arises because it is more common to
$100 1000 kW 1 year have the time unit in the measurement of a flow than
× × in the measurement of a quantity. Thus if you want to
kW× year 1 MW 8760 h fill your gas tank, you buy a quantity of 15 gallons of
which reduces to $11.42/MWh. gasoline, and that flows into your tank at the rate of
5 gallons per minute. But if you need a quantity of
electric energy, that would be 30 watt-hours, and it would be delivered at the rate
of 1000 watts.2 Because a watt-hour is a unit of energy, it would make sense to
speak of delivering 1000 watt-hours per hour, but that just boils down to a rate of
1000 watts (1 kW) because a watt-hour per hour means watts times hours divided
by hours, and the hours cancel out.
The Price of Power and Energy

Because power is a flow, its total cost is measured in dollars per hour, not dollars.
The total cost of a certain quantity of energy is measured in dollars. Consequently
the price (per unit cost) of power is measured in dollars per hour per MW of power
flow, while the price of energy is measured in dollars per MWh. But these units
are the same:
(dollars per hour) per MW = ($/h)/MW = $/MWh
so the units for the price of power are the same as for the price of energy.
Typically the price of retail energy is about 8¢/kWh.3 At that price, the price
of power would be 8 cents/hour for a kilowatt of power flow, which is the same.
These units are convenient for home use but are inconveniently small for bulk power
systems. Consequently this book will use megawatts (millions of watts) instead
2. Watts per hour has units of watts divided by hours and has no use in the present context.
3. Average revenue per kilowatt hour to ultimate residential consumers was 8.06¢/kWh, according to DOE
(2001c, Table 53).

of kilowatts. The same energy price can be re-expressed as $80/MWh. When

discussing large markets and annual energy use, power may be measured in
gigawatts (GW, or billions of watts) and energy in terawatt hours (TWh, or trillions
of watt hours).
Another commonly used unit is the mill, short for “milli-dollar,” or 1'1000 of a
dollar. This unit might seem particularly inappropriate for wholesale markets, but
it is commonly used to compensate for using the kW which is also inappropriately
small. Together these give rise to “milli-dollars per kilowatt-hour,” often incorrectly
shortened to “mills.” Scaling both the numerator and denominator up by 1000 has
no effect on the numeric value and converts milli-dollars to dollars and kilowatts
to megawatts. So 80 mills/kWh is identical to $80/MWh.
1-3.2 MEASURING GENERATION CAPACITY

The size of a generator is measured by the maximum flow of power it can produce
and therefore is measured in MW. The capacity to produce a flow of power is best
conceptualized as a flow just as a MW of power is a flow of energy.4
In principle one could define an amount of capacity related to the flow of
capacity as energy is related to power, but this is not necessary. Moreover, it is likely
to cause confusion because when applied to a generator, it would aggregate a flow
of capacity over many years without any discounting. For these reasons, the idea
of a capacity amount, different from a capacity flow, will not be introduced or
utilized.
Having found that capacity, like power, is a flow measured in MW, it is natural
to ask if it is priced in $/MWh as is power. Most would say no, but it is best to look
to its use in solving real economic problems before drawing this conclusion.
Consider the problem of choosing which generator can most cheaply serve a load
of a particular duration. The long tradition of solving this problem by using
“screening curves” will provide the key to this puzzle.
1-3.3 PRICING GENERATION CAPACITY
The “Overnight” Cost of Capacity

A generator has an “overnight cost” which is typically given in $/kW. For example,
the overnight cost of a coal plant might be $1,050/kW, so a 1000 MW plant would
cost $1,050 million. In economic terms, this is the present-value cost of the plant;
it would have to be paid as a lump sum up front to pay completely for its construc-
tion.
A conventional gas-turbine generator (GT) would have an overnight cost closer
to $350/kW. Although the GT is three times cheaper than the coal plant, for some
4. The flow of available capacity is interrupted during generator outages, but the flow of installed capacity
is continuous. This chapter ignores the difference and assumes that the flow of capacity from a generator
is continuous and constant.

purposes the coal plant is the more inexpensive choice. Fuel costs must always
be taken into account when evaluating the choice of generators. Coal plants are
built because their cost of fuel per unit of energy output is less. Assume coal costs
only $10/MWh of energy produced, while the cost of fuel for a GT comes to
$35/MWh. Now which plant is cheaper?
More information is needed. The comparison depends on how much the plant
will be used, and that depends on the load it will serve. For concreteness, assume
that the load has a duration of 25% (2190 hours/year) so the plant serving it will
have a capacity factor of 25%. Now, which plant is cheaper?
Focusing on only the basics, the problem seems workable. The overnight cost
captures the fixed cost of generation, and the fuel cost per unit of output captures
the variable cost. Duration gives a sufficient description of the load. But the
problem is still impossible to solve because the fixed cost of capacity has been
measured in the wrong units. Overnight costs measured in $/kW cannot be added
to fuel costs measured in $/kWh. This would produce nonsense.
Fallacy 1-3.1 Fixed and Variable Costs Are Measured in Different Units
Because capacity is usually paid all at once, while fuel is paid for over time,
variable costs but not fixed costs should include a time dimension.
When units have the same “dimensions,” they differ only by a scale factor (a
pure number). Different quantities having units of the same dimension can be added.
For example, 1 MWh can be added to 100 kWh to get 1100 kWh (or 1.1 MWh).
But quantities whose units have different dimensions cannot be added. This is the
meaning of the famous saying, “you can’t add apples and oranges.” For example,
1 MW cannot be added to 1 MWh. Engineers and physicists pay close attention
to mismatched units because they always signal deeper trouble. Any calculation
that involves adding MW and MWh simply does not make sense.
Identifying Fixed Costs on Screening Curves

Screening curves, shown in Figure 1-3.1, are used to compare generation costs by
taking account of the three factors of our present problem: fixed cost, variable costs
and load duration (which determines the generator’s capacity factor). Necessarily,
they provide guidance on the proper units for fixed costs. Traditionally, these curves
plot “annual revenue requirement per kW” (ARR) as a function of capacity factor
(cf ). The generator’s capacity factor is its percentage utilization which is deter-
mined by the load’s duration.5
Traditionally, the variable cost component of ARR is computed by taking the
fuel cost expressed in $/MWh and converting to $/kWy.6 The result is $87.60/kWy
for a coal plant and $306.60/kWy for a GT. This assumes full-time operation, so
5. Duration is measured as a percentage (see Chapter 1-4), so if all load served has the same duration, the
capacity factor equals load duration. If the load has a range of durations, these must be averaged.
6. For simplicity, this assumes that fuel is the only variable cost. Operation and maintenance include an
additional variable cost component which should be expressed in $/kWy.

Figure 1-3.1
Use of screening curves
to select a generator.
to find the variable component for any particular cf , these must be multiplied by
cf , 25% in the case of the present example.
The overnight cost of capacity is more problematic. A coal plant with an
overnight cost of $1,000/kW does not cost $1,000/kWy. This would imply a plant
life-time of one year and a discount rate of zero. The correct fixed-cost component
of ARR is the overnight cost amortized (“levelized”) over the life of the plant. This
is equivalent to computing home mortgage payments based on a mortgage that lasts
the life of the house. Obviously a discount rate (interest rate) is involved. The
formula for amortization is
r ⋅ OC r ⋅ OC
FC = − rT
≈ (1-3.1)
1− e 1 − 1 / (1 + r ) T
Notice that fixed cost (FC ) depends only on overnight cost (OC ), the discount rate
(r, in % per year) and the life of the plant (T, in years).7
Table 1-3.1 Technology Costs
VC VC OC FC FC
Corrected in Technology
( /MWh) ( /kWy) ( /kW) ( /kWy) ( /MWh)
published version. Gas turbine $35 $87.60 $350 $109.96 $12.21
Coal $10 $306.60 $1050 $40.48 $4.62
Fixed-costs are based on r = 0.1 and on T = 20 for gas turbines and 40 for coal plants. Equation 1-3.1
gives fixed costs in $/kWh which are then converted to $/MWh by dividing by 8.76.
FC is a constant flow of cost that when added to VC gives ARR, the annual
revenue requirement per kW of generation capacity. Of course this assumes a
capacity factor of 1. If cf is less, VC will be reduced proportionally, but FC is
unaffected because capacity must be paid for whether used or not. That is why FC
is termed the fixed cost. The formula for ARR is
Screening Curve: ARR = FC + cf × VC
7. Using monthly instead of annual compounding in the second formula greatly improves its accuracy as
an approximation. To do this, change r to r ' 12 in the denominator and T to 12 T.

for ARR to be valued in $/kWy, both FC and cf × VC must also be valued in $/kWy.
As these are the traditional units for ARR, the traditional units for fixed cost must
also be $/kWy. These units have the same dimension as $/MWh and any quantity
expressed in $/kWy can be converted to $/MWh by dividing by 8.76.
Variable cost is naturally expressed in $/MWh, so capacity factor, cf , must be
a pure number (dimensionless), otherwise, cf × VC would not have the same units
as ARR. This is correct; a capacity factor is just the fraction of a generator’s potential
output that is actually produced. It is actual energy output divided by potential
energy output, so the energy units cancel.
Result 1-3.2 Energy, Power, and Capacity Are Priced in $/MWh

Although power is measured in MW and energy in MWh, both are priced in
$/MWh. Like power, generating capacity is a flow measured in MW and conse-
quently is also priced in $/MWh.
The Rental Cost of Capacity

Fixed costs are the costs of generation capacity. It may be argued that buying a
generator is buying capacity and that generators are measured in MW, not in MWh.
This is only partially true. If a 1 MW gas-turbine generator is worth $350,000, does
this mean 1 MW of capacity is worth $350,000? No, the gas turbine is worth that
only because it has a certain expected lifetime. An identical but older gas turbine
is worth less, even though it has the same 1 MW capacity. Thus the price of capacity
always involves a time dimension, either explicitly or implicitly.
Measuring capacity in MW indicates that capacity is being considered a flow.
A 100-MW generator delivers a 100-MW flow of capacity for some unspecified
period of time. That flow must be paid for by a flow of money—so many dollars
per hour. This corresponds to a rental cost. If a generator is rented, the cost of
renting will be so much per hour, or per day, or per year. If this is scaled by the
generator’s capacity, for easy comparison with the rental rate of other generators,
then it is natural to express the rental cost of a generator in $/h per MW, or equiva-
lently in $/MWh.
The above screening curve analysis can be summarized as saying that generation
capacity costs should be expressed as a rental rate and not as a one-time (overnight)
purchase price. Rental rates naturally have the same units as variable costs and so
make total and average cost calculations convenient.
Two Kinds of Average Cost: Avoiding Confusion

The cost of operating a generator with a specific capacity factor can be read from
a screening curve. Traditionally this cost is expressed in $/kWy and called an annual
cost. Although a kWy has the dimensions of energy, this cost is not the annual cost
of energy produced by the plant! A screening curve shows the average cost of using
the coal plant’s capacity.

Figure 1-3.2
Capacity-cost based and
energy-cost based
screening curves.
With the price of energy always expressed as an hourly cost, it is more conve-
nient to divide the annual cost, in $/kWy, by 8.76 and arrive at an average cost per
hour for the year expressed in $/MWh. Like the annual cost, this hourly average
cost is still not a cost of energy produced but the cost of using capacity. This book
will always report capacity, energy and power costs in $/MWh for ease of compari-
son and to make addition of costs and cost averaging possible.
Although screening curves plot the average cost of capacity use, the average
cost of energy produced is also interesting and could be used to construct hyperbolic
screening curves. A pair of these is shown on the right side of Figure 1-3.2; they
are nonlinear (hyperbolic), but they still intersect at exactly the capacity factor at
which one plant becomes more economical than the other. The equations for the
linear and hyperbolic screening curves are closely related and are shown in Figure
1-3.2.
The average cost of capacity (ACK) used with capacity factor cf is the fixed cost
of using that capacity, plus cf times the variable cost of producing energy. If
cf = 1'3 , then one third of the variable cost of maximum potential energy output
must be added to the constant fixed cost which increases the average cost per unit
of capacity by cf × VC.
The average cost of energy (ACE) when the genera-
Screening Curves tor runs with capacity factor cf is the variable cost of
A traditional screening curve plots average cost as producing that energy, plus the fixed cost of the capac-
a function of capacity factor. When using the screen- ity divided by cf . If cf = '3 , then fixed costs must be
1
ing curve equation, this average cost can be con- spread over only '3 of the total possible energy output,
1
fused with the average cost of the energy produced so they are multiplied by 3 (divided by cf ) before being
by a certain type of generator. Instead, it is the added to VC.
average cost of using a unit of capacity. A load slice is a horizontal strip cut from a load-
Screening curves could have been defined using duration curve (see Chapter 1-4.1). Depending on its
the average cost of energy. Then as the capacity
average duration it will be served by some particular
factor approached zero the average cost would
approach infinity. The average-energy-cost equation
technology, baseload, midload, or peaking. Any given
is used to analyze market equilibria in later chapters, load slice is defined by a capacity, Kslice, which is the
but nonstandard screening curves are never used. height of the slice, and an average duration, D. It also
has a total energy requirement, E = D × Kslice. To serve

this load, generation capacity of Kslice must be installed and must run with a capacity
factor of cf = D. Having selected a technology, one can compute the average cost
per MW-of-capacity of serving the load, ACK, and the average cost per MWh-of-
energy of serving the load, ACE. The total cost of serving load is then given by both
Kslice × ACK and E × ACE. Because Kslice and E are both fixed, choosing the technol-
ogy that minimizes either ACK or ACE will minimize the total cost of energy. This
is why either the traditional or the hyperbolic screening curves can be used.
These relationships can be summarized as follows. For a particular load slice
served by generators with fixed cost, FC, and variable cost, VC, the average cost
of capacity and energy can be found as follows:
Capacity: ACK = FC + cf × VC = FC + D × VC (1-3.1)
Energy: ACE = FC ' cf + VC = FC ' D + VC (1-3.2)
The capacity factor of the generator, cf , equals the average duration of the load, D.
No one uses hyperbolic screening curves, but when an average cost is computed
for a specific technology, say by DOE, a value for ACE (not ACK) is always com-
puted. Typically, DOE might report the overnight cost (OC), some information
about fuel costs, and a value for ACE based on technical capacity factor (cf ).8 In
other words, DOE reports some technologically determined value on the technol-
ogy’s hyperbolic screening curve.
ACK is used to determine the optimal durations of various generation technolo-
gies, and from these durations the optimal investment in these technologies. Since
competitive markets optimize technology, ACK is also used to determine competitive
outcomes. Either ACE or ACK may be used to compare the cost of peak energy with
the value of lost load, depending on whether peaker costs are equated to the value
of lost load or the average hourly cost of lost load (see Chapter 2-2 and 2-3). ACE
is also well suited to DOE’s interest in alternative technologies—nuclear, wind,
solar, and so on. These have in common capacity factors which, even in a market
environment, are not affected by normal variations in market structure but are
instead technologically determined because their variable costs are almost always
below the market price. They run whenever they are physically able, so their
capacity factor is determined by their technical capability. The economics of an
alternative technology can be assessed by comparing its ACE with the market’s
average price.
Standard technology generators have capacity factors determined by the market
and not just by their technology. In this case, the duration of the load they serve,
which determines their capacity factor, needs to be determined from their fixed
and variable costs along with those of other technologies. This is done with screen-
ing curve analysis, or with algebra based on screening curves. Traditional linear
screening curves prove simplest. These curves and their associated algebra will
be used throughout the book, as is the formula for ACE.
8. See DOE (1998a) Tables 14–17. In a table of cost characteristics of new generating technologies (DOE
2001a, Table 43) OC is given, but not FC. In a slide labeled “Electricity Generation Costs,” DOE (2001b)
reports the capital costs in mills/kWh. As the title indicates, these are FC ' cf , for cf determined by
technology-based capacity factors, and so are components of ACK, as advertised in the title and thus points
on the hyperbolic screening function.

Table 1-3.2 Fixed and Variable Cost of Generation

Overnight Fixed Fuel Variable
Capacity Cost Cost Cost Heat rate Cost
Type of Generator $/kW $/MWh $/MBtu Btu/kWh $/MWh
Advanced nuclear 1729 23.88 0.40 10,400 4.16
Coal 1021 14.10 1.25 9,419 11.77
Wind 919 13.85 — — 0
Advanced combined cycle 533 7.36 3.00 6,927 20.78
Combustion turbine 315 4.75 3.00 11,467 34.40
* Overnight capacity cost and heat rates are from DOE (2001a), Table 43. Plants not labeled “advanced” are “conventional.” Rental capacity
costs are computed from overnight costs, a discount rate of 12% and assumed plant lifetimes of 40 years except for wind and gas turbines which
are assumed to be 20 years. For simplicity, operation and maintenance costs are ignored.
1-3.4 TECHNICAL SUPPLEMENT
Checking Fixed-Cost Units with the Amortization Formula

As a final check on the units of fixed costs, the amortization formula can be
analyzed. Interest rates (e.g., 10% per year) has the dimension of “per unit time,”
and T has the dimension of time, so rT is dimensionless, that is, a pure number.
This is necessary for compatibility with “1” in the denominator. In the numerator,
OC has the traditional units of $/kW and “r” again has the dimension of 1 ' time,
so r × OC has the dimensions of OC per unit time, for example, $/kW per year.
If overnight cost is measured in $/kW and interest is given in percent per year, fixed
cost must be measured in $/kWy.
Fixed and Variable Costs for Different Technologies

Table 1-3.1 computes fixed and variable costs for five types of generators as an
example of converting overnight cost to fixed costs. The listed values of FC and
VC are exactly the values needed to draw screening curves and choose the most
efficient plant to serve loads of any duration. For example, the cost of serving load
of duration D with an advanced combined-cycle plant is
ACK = (7.36 + 20.78 D) $/MWh.
To convert this to the more traditional units of $/kWy, both values should be
multiplied by (1 M ' 1000 k)(8760 h/y) or 8.76. (Note that, including units, this
is just multiplication by 1 since 1 M = 1000 k and 8760 h = 1 y.)
To avoid having ACK appear to have the same units as variable costs, its units
are often stated as “$/kW per year” which translates to $/kW/year. But just as
x ' y ' z = x ' (y × z ), so $/kW/year equals $/kW-year which is denoted by $/kWy.
The phrase “$/kW per year” is correct, but it means no more and no less than
$/kWy, which has the dimensions of dollars per energy.

Chapter 1-4
Power Supply and Demand
And when the Rain has wet the Kite and Twine, so that it can conduct the Electric Fire freely,
you will find it stream out plentifully from the Key on the Approach of your Knuckle.
Benjamin Franklin
1752
T HE PHYSICAL ASPECTS OF SUPPLY AND DEMAND PLAY A PROMINENT

ROLE IN POWER MARKETS. Shifts in demand, not associated with price, play
a role in all markets, but in power markets they often receive attention to the
exclusion of price. This is not simply the result of regulatory pricing; even with
market prices, demand shifts will play a key role in determining the mix of produc-
tion technologies. In this way hourly demand fluctuations determine key long-run
characteristics of supply.
Because electric power cannot be stored, production always equals consumption,
so the difference between supply and demand cannot be indicated by flows of
power. Neither is the instantaneous difference indicated by contracts since real-time
demand is determined by customers physically taking power. The short-run
supply–demand balance is indicated by voltage and, especially, frequency. This
unusual market structure requires some elementary background in system physics.
More detail is provided in Chapters 5-1 and 5-2.
Chapter Summary 1-4 : Load duration curves are still relevant in unregulated
markets, but their role in analysis is more subtle because their shape is affected
by price and its correlation with load. They can still be used with screening curves
to check an equilibrium, but to predict an equilibrium they must be used in combina-
tion with price elasticity.
Power production always equals consumption (counting losses as part of
consumption) which makes it impossible to assess the supply–demand balance by
observing quantities or quantity flows. Instead, frequency is the proper indicator
of system-wide balance, and net unscheduled flows between regions are used to
share the responsibility of maintaining this balance.

CHAPTER 1-4 Power Supply and Demand 41
Section 1: Describing the Demand for Power. A year’s worth of hourly

fluctuations can be usefully summarized by a load-duration curve that plots demand
against duration, the fraction of the year during which demand is at or above a
certain level. It can also be thought of as the probability of finding load above a
certain level.
If customers are charged real-time prices, peak demand will be reduced, allowing
a reduction in generating capacity. The result will be a load-duration curve with
its peak cut off horizontally.
Section 2: Screening Curves and Long-Run Equilibrium. If the screening
curves of the available technologies are drawn on the same graph, their intersections
determine capacity factors that mark technology boundaries. By mapping these
capacity factors to durations and then to the load-duration curve, the optimal
capacities for these technologies can be read off the vertical axis.
This technique can be used to partially confirm a market equilibrium but not
to find one. In a market, price affects the shape of the load-duration curve, so it
cannot be taken as given until the equilibrium is known.
Section 3: Frequency, Voltage, and Clearing the Market. When consumers
turn on ten 100-W light bulbs, they are demanding 1000 W of power, and if
generators supply only 900 W, the system will not be “in balance.” In spite of this,
the power supplied will exactly equal the power consumed (ignoring losses). This
equality of power flows is caused by a decrease in voltage sufficient to cause the
100 W bulbs to use only 90 W of power. For motors the same effect is caused by
a drop in frequency. Because voltage is automatically adjusted at substations,
frequency is the main balancer of power inflows and outflows.
The United States is divided into three AC interconnections: the Western,
Eastern, and Texas. The system frequency is constant throughout each AC intercon-
nection, which means that a change in frequency cannot be used to locate a
supply–demand imbalance. Instead, net power flows are tracked out of each control
area and compared with scheduled power flows. This allows the imbalance to be
located.
1-4.1 DESCRIBING THE DEMAND FOR POWER

Traditionally the demand for power has been described by a load-duration curve
that measures the number of hours per year the total load is at or above any given
level of demand. An example is shown in Figure 1-4.1. Total demand (load) is a
demand for a flow of power and is measured in MW. Although the load-duration
curve describes completely the total time spent at each load level, it does not include
information about the sequence of these levels. The same load-duration curve can

Figure 1-4.1
A load-duration curve.
be produced by wide daily swings in demand and little seasonal variation or by

wide seasonal variation and limited daily swings.
The introduction of a market adds the dimension of price. Economists often
represent demand by a demand curve which expresses demand solely as a function
of price. Nonprice fluctuations of the type captured by a load-duration curve are
referred to as shifts in the demand curve and are generally not described in detail.
But electricity is different because it is not storable, so peak demand must be
satisfied by production from generators that are used as little as 1% of the time.
Such generators, peakers, are built with technology that differs markedly from
that used for baseload generators which run most of the time and are stopped only
rarely. As a result, power markets face the problem of determining how much
generation capacity should be built using each type of technology, for example,
coal-fire steam turbines or gas-fired combustion turbines (gas turbines). This
explains the unusual importance of demand shifts and consequently of load-duration
curves in power markets.1
Load-Duration Curves
A load-duration curve can be constructed for a given region (or for any collection
of loads) by measuring the total load at hourly intervals for each of the 8760 hours
in a year, sorting them, and graphing them starting with the highest load. The result
is a curve that slopes downward from the maximum load in the peak hour, hour
1, to the minimum load, baseload, in the most off-peak hour, hour 8760 (see Figure
1-4.1).
Duration is traditionally measured in hours per year, but both hours and years
are measures of time, so duration is dimensionless, which means it can be expressed
as a pure number, a ratio, or percentage. To convert from units of hours per year
(h/year) to a pure number, simply multiply by 1 in the form (1 year) ' (8760 h).
Duration has a natural interpretation as the probability that load will be at or above
a certain level. To use this interpretation pick a load level, say 35 GW, and using
the load-duration curve, find the corresponding duration, 20% in this case. This
1. Service industries such as restaurants and airlines often have demand fluctuations which cause similar
problem because their output is not storable, but they tend to use the same technology on and off peak.

indicates that load is 35 GW or greater 20% of the time. Put another way, the
probability of load being 35 GW or greater in a randomly selected hour is 20%.
This interpretation is most convenient.
The Price-Elasticity of Demand

Presently, demand is almost completely unresponsive to price in most power
markets because wholesale price fluctuations are not usually passed on to retail
customers. Often retail prices remain under some form of price regulation, but
competitive retailers have also been slow to implement real-time pricing. In the
longer run, retail prices do change, sometimes seasonally. In the long run a 10%
increase in the price of power will cause approximately a 10% reduction in the use
of power.2 This is not a very accurate approximation, but the long-run response
to a 10% increase in price is likely to be found between 5% and 15% and is certainly
not zero. Economists term this price sensitivity a price elasticity of demand, which
is often shortened to demand elasticity. If a 10% change in price causes a 5%, 10%,
or 15% change in demand, the elasticity is said to be 0.5, 1.0, or 1.5, respectively.
(Technically, demand elasticities are negative, but this book will follow the common
convention of re-defining them to be positive.)
Real-Time Pricing and the Load-Duration Curve

Under regulation, residential load usually faces a price that fluctuates very little
while commercial and industrial load often face time-of-use (TOU) pricing or
demand charges. Time-of-use prices are designed to be high when demand is high,
but the approximation is crude as they are set years in advance. Consequently they
miss the crucial weather-driven demand fluctuations that cause most problematic
supply shortages. Demand charges are no more accurate as they are based on
individual demand peaks, not system peaks. Coincident-peak charges improve on
this by charging customers for their use at the time of the system peak, but these
are less common.
Because supply is fairly constant, the market is tightest when demand is highest.
Consequently, high wholesale prices correspond well with high demand. If these
real-time prices are passed through to customers, then retail prices will track load
fairly well. Although real-time prices work best, all four pricing techniques, TOU,
demand, coincident-peak, and real-time, tend to raise prices when demand is highest
and reduce prices when demand is lowest. This results in lowering the peak of the
load-duration curve and raising the low end of the curve.
If load faced real-time prices, the need for generation capacity might be reduced
to the point where the load-duration curve under regulation had a duration of, say,
10%. Then, between 0% and 10% duration, supply and demand would be balanced
by price. Instead of having generation follow load, load would be held constant
by price at the level of installed capacity. In the lowest duration hours, price would
2. There is no natural definition of short- and long-run demand elasticity, which can be defined usefully
over any time horizon from five minutes to twenty years. This text will use short-run elasticity to mean
something on the order of one day and long-run to mean about five years.

Figure 1-4.2
The effect of price
elasticity on load
duration.
need to be very high to reduce demand to this level. By fluctuating sufficiently,

price would control demand and produce a flat-topped load-duration curve with
a maximum load just equal to generating capacity as shown in Figure 1-4.2.
1-4.2 SCREENING CURVES AND LONG-RUN EQUILIBRIUM

When demand is inelastic or when it faces a fixed price so that the load-duration
curve is fixed, this curve can be used to find the optimal mix of generation technolo-
gies. The technique was developed for a regulated power system in which price
and the load-duration curve are often fixed, but it is still useful for understanding
certain aspects of competitive markets.
It assumes that fixed and variable costs adequately describe generators. These
are used to draw screening curves for each technology on a single graph as shown
in Figure 1-4.3. The intersections of these curves determine capacity factors that
separate the regions in which the different technologies are optimal. These capacity
factors equal the load durations that determine the boundaries between load that
is served by one technology and the next. The screening curves in the figure are
taken from Figure 1-3.1 and intersect at a capacity factor, cf , of approximately 30%.
Consequently all load with a duration greater than 30%, or about 2600 hours, should
be served by coal plants, while load of lesser duration should be served by gas
turbines. The arrow in the figure shows how the needed capacity of coal plants
can be read from the load-duration curve. The optimal GT capacity is found by
subtracting this from maximum load which is the total necessary capacity. (Forced
outages and operating reserve margins are considered in Parts 2 and 3.)
If customers face the wholesale market price through real-time pricing, this
technique cannot be used because the load-duration curve depends on price, and
price depends on the choice of technology, and the choice of technology depends,
as just described, on the load-duration curve. This circularity is in no way contradic-
tory, but it makes it difficult to find the competitive market equilibrium. Not only
is calculation more difficult, but, also, the elasticity of demand must be known.
In spite of this circularity, the traditional technique can be used to partially
confirm a long-run equilibrium. The load-duration curve observed in a market

Figure 1-4.3
Using screening curves
to find the optimal mix of
technologies.
includes the effect of price on demand. When it is used along with the screening
curves of the available technologies, the traditional method should predict the mix
of technologies observed in the market if the market is in long-run equilibrium.
In practice many complications must be overcome.
1-4.3 FREQUENCY, VOLTAGE, AND CLEARING THE MARKET

So far, this chapter has considered how to describe demand and how to find the
optimal mix of technology to supply it. This section considers the physical details
of the supply–demand balance in real time. At any instant, customers are using
power, generators are producing it, and the amount produced is exactly equal to
the amount consumed. Some may object to the word “exactly,” but the discrepancy
is at least a thousand times smaller than anyone’s ability to measure it and is entirely
irrelevant. The determination of the supply–demand balance depends on electrical
phenomena more subtle than the concepts of quantity and quantity flow.
Losses
In real networks, a few percent of the power consumed is consumed by the network.
This consumption should be considered part of demand even though it serves no
end used. With this convention, the system can be viewed as maintaining a perfect
balance between supply and consumption (including losses) at all times and between
supply and demand whenever customers are getting the power they want.

Convention Loss Provision is Not Considered Part of Supply or Demand

Losses will be considered as a service paid for by traders and provided separately
from the trading arrangement. Consequently, from a trading point of view, power
flows can be viewed as lossless.
Supply precisely equals consumption because there is no storage of power in

transmission systems.3 But if supply equals consumption regardless of price, what
signal should be used for price adjustment? How can demand be observed to be
either greater than or less than supply? A mismatch between supply and demand
is signaled not by power flows but by frequency and voltage. When they are below
their target values, demand exceeds supply and vice versa.
Frequency and Voltage

Power systems attempt to maintain a constant frequency, the rate at which alternat-
ing current alternates. In the United States, alternating current (AC) reverses
direction twice, thus completing one cycle and returning to its original direction,
60 times per second. The frequency of AC in the United States is therefore said
to be 60 cycles per second, also known as 60 Hertz or 60 Hz. In many countries
the target frequency is 50 Hz.
Voltage is the amount of electrical pressure that pushes current through electrical
appliances such as lights and motors. As with frequency, power systems have a
certain target voltage that they attempt to maintain. In the United States the target
residential voltage is about 120 volts. In some countries the target voltage is about
twice as high. When an unprotected 120-V appliance is plugged into a 240-V outlet,
the extra electrical pressure (voltage) causes twice as much current to flow through
the appliance. This causes the appliance to use four times as much power (power
is voltage times current) and the appliance typically burns out. The important point
for this section is that as voltage increases, most appliances use more power, and
as voltage decreases most appliances use less power.
Imagine a system with ten generators operating at full throttle supplying ten
thousand homes with lights burning and motors running. If one generator goes off
line, two things happen. The system voltage and frequence both decrease. Both
cause electrical appliances to use less power. This effect has been described for
voltage, but for more complex reasons most motors use less power when the system
frequency declines. The decline in voltage and frequency is produced automatically
by the physics of the entire system including all loads and generators. It happens
to the exact extent necessary to balance inflow (supply) and outflow (consumption).
If this did not happen, a law of physics, just as fundamental as the law of gravity,
would be violated.
Although nothing can prevent the combined drop in frequency and voltage when
generation is reduced and load maintained, it is possible to influence the relative
extent to which each decreases. In fact, the system has automatic controls that do
3. More precisely, the amount stored is minuscule and cannot be utilized for trade.

just that. At substations, where very high transmission voltages are reduced to the
lower, but still high, distribution voltage, there are automatically controlled trans-
formers. These adjust so that the distribution voltage remains relatively constant
even when the transmission voltage drops. Because of these devices, more of the
power flow adjustment that accompanies the loss of a generator is accomplished
through frequency reduction than through voltage reduction. Nonetheless both can
and do happen.
Frequency and Interconnections

An interconnection is a portion of the power grid connected by AC power lines.
The three interconnections in the United States—the Eastern Interconnection,
the Western Interconnection, and most of the Great State of Texas—each maintain
a uniform frequency. Frequencies in Maine and Texas bear no particular relationship
to each other, but the AC voltage in Maine stays right in step with the AC voltage
in Florida, night and day, year in and year out. The frequency in every utility in
an entire interconnection is exactly the same.4 If one utility has a problem, they
all have a problem.
These three interconnections are connected to each other by a number of small
lines, but they are separate interconnections because the connecting lines are all
DC lines. No AC power flows between them. On DC lines, the electrical current
flows in only one direction; it does not alternate directions. Thus DC lines have
no frequency and as a consequence need not (and cannot) be synchronized with
the power flow of an AC interconnection. This allows trade between two different
interconnections that are not synchronized with each other.
The Signal for Price Adjustment

When a generator breaks down unexpectedly (a forced outage) and supply de-
creases, demand is then greater than supply, even though consumption still precisely
equals supply. Consumption is less than demand because of rationing. A consumer
with a 100-W light turned on is demanding 100 watts of power. During a brown
out, however, 100 watts are not supplied to the bulb as power to the bulb is rationed
by the suppliers. This rationing is not due to deliberate action but is a consequence
of system physics which automatically lowers the voltage and frequency. For
simplicity, in the remainder of this section, rationing will be discussed as if it
happened solely through frequency reduction, as this is generally considered to
be the predominant effect.
A drop in frequency below the target level of 60 Hz is a clear and accurate
indication that demand exceeds supply for the interconnection as a whole. Similarly,
any frequency above 60 Hz indicates that supply exceeds demand. In other words,
more than 100 watts are being delivered to 100-W motors. This extra power is
4. If the frequency difference between Maine and Florida were 0.001 Hz, for one minute, it would cause
an accumulated phase change between the two states of 22 degrees. This would lead to dramatic changes
in power flow. Thus while the frequency lock between utilities is not exact, it is extremely tight, and there
can be no persistent frequency difference. One utility cannot have a problem unless they all do.

generally unwanted because appliances are built and operated on the assumption
that power will be delivered at a frequency of 60 Hz.
Because frequency indicates the discrepancies between supply and demand,
frequency is the right guide for interconnection-wide price adjustment. When
frequency is high, price should be reduced; when frequency is low, price should
be raised. This is the classic adjustment process for keeping supply equal to demand.
Definition Demand
The demand for power is the amount of power that would be consumed if system
frequency and voltage were equal to their target values for all consumers. Note
that shed load is included as part of demand. This is an economic definition and
contradicts the engineering definition provided by North American Electric
Reliability Council (NERC). (Often “load” is used to mean demand.)
Result 1-4.1 Supply Equals Consumption but May Not Equal Demand
As in all markets, demand is the amount customers would buy at the market price
were supply available. If voltage or frequency is low, customers consume less
power than they would like so supply is less than demand.
As always, the real world adds one more layer of complexity. The frequency
in every power market in an interconnection is exactly the same. Thus frequency
reveals nothing about the supply and demand conditions in any particular market
but only about the aggregate supply and demand conditions of the entire intercon-
nection. Consequently individual markets cannot rely on the frequency alone to
determine their price adjustments.
NERC defines another control variable that takes account of both frequency
and the net excess flow out of a trading region (the net interchange). The net excess
outflow is the actual outflow minus the scheduled outflow. An excess outflow is
a strong signal that supply is greater than demand in the trading region. If the
frequency is high in the interconnection this is a weak signal of excess supply in
any particular market. These two signals are combined to form a single indicator
of excess supply for each market. The indicator is called the area control error, or
ACE. Control areas are required to keep their ACE near zero, and they do. ACE
is the main indicator of the supply–demand balance in every control area in the
United States and when there is a market, it is the signal that determines whether
the price will be increased or decreased by the system operator.

Chapter 1-5
What Is Competition?
The rich, . . . in spite of their natural selfishness and rapacity, . . . though the sole end which they
propose . . . be the gratification of their own vain and insatiable desires, they divide with the poor
the produce of all their improvements. They are led by an invisible hand to make nearly the same
distribution of the necessaries of life, which would have been made, had the earth been divided
into equal portions among all its inhabitants, and thus without intending it, without knowing
it, advance the interest of the society.
Adam Smith
The Theory of Moral Sentiments
1759
C OMPETITION IS LEAST POPULAR WITH THE COMPETITORS. Every

supplier wants to raise the market price, just as every buyer wants to lower it.
Perfect competition frustrates both intentions.
Some commodity markets provide almost perfect competition; eventually power
markets may work almost as well. But designing such markets is difficult. Economic
competition is not like competition in sports, which may be considered perfect when
there are just two powerful and equal competitors. Economists consider competition
to be perfect when every competitor is small enough (atomistic is the term used)
to have no discernable influence against the “invisible hand” of the market.
Adam Smith guessed intuitively that a perfectly competitive market, in the
economic sense, would produce an outcome that is in some way ideal. Many
difficulties can cause a market to fall short of this ideal, but even a market that is
only “workably competitive” can provide a powerful force for efficiency and
innovation.
Power markets should be designed to be as competitive as possible but that
requires an understanding of how competition works and what interferes with it.
On its surface, competition is a simple process driven, as Adam Smith noted, by
selfishness and rapacity; but the invisible hand works in subtle ways that are often
misunderstood. Those unfamiliar with these subtleties often conclude that suppliers
are either going broke or making a fortune. This chapter explains the mechanisms
that keep supply and demand in balance while coordinating production and con-
sumption to produce the promised efficient outcome.

Chapter Summary 1-5 : The plan of deregulation is to achieve efficiency

through competition. Economics guarantees this result provided the market reaches
a classic competitive equilibrium. This requires at least three conditions to be met:
price taking suppliers, public knowledge of the market price, and well-behaved
production costs. Although production costs seem problematic to many, they cause
little trouble, and deregulation will probably succeed if markets are designed for
maximum competition and transparent prices.
Section 1: Competition Means More than Struggle. The dictionary defines
competition as “a struggle with others for victory or supremacy,” but economics
does not. Designing markets to be “competitive” in the dictionary’s sports-oriented
sense produces poor designs about which little can be predicted. Economic competi-
tion requires many competitors on each side of the market and results in a lack of
market power and “price taking” behavior.
Section 2: Efficient Markets and the Invisible Hand. The central result
of economics states that competition leads to efficiency. But to achieve short-run
efficiency, competitive behavior must be supplemented with well-behaved costs
and good information. Long-run efficiency requires free-entry of new competitors
as well. Efficiency means that total surplus, the sum of profit and consumer surplus,
is maximized.
Section 3: Short- and Long-Run Equilibrium Dynamics. Price and quantity
adjustments, usually by suppliers, lead the market to equilibrium. In a competitive
market, suppliers adjust output until marginal cost equals the market price and adjust
price until the market clears (supply equals demand). They are price takers because
when considering what quantity to produce they take the market price as given;
that is, they assume it will remain unchanged if they change their output.
A long-run competitive equilibrium is brought about by investment in productive
capacity. Profit (which means long-run economic profit) is revenue minus costs,
and cost includes a normal return on capital (investment). Thus, zero economic
profit provides a normal return on investment. If economic profit is positive and
the market competitive, new suppliers will enter. In this way profit is brought down
to zero under competition, but this is enough to cover all fixed costs and a normal
risk premium.
Section 4: Why is Competition Good for Consumers? Competition mini-
mizes long-run costs and pays suppliers only enough to cover these minimum costs.
Although it is possible to depress price in the short run, it is not possible to pay
less on average than minimum long-run average cost.

CHAPTER 1-5 What Is Competition? 51
1-5.1 COMPETITION MEANS MORE THAN “STRUGGLE”

The dictionary defines competition as “a struggle with others for victory or suprem-
acy.” This definition is based on sports, not economics, but is quite influential with
regulators and politicians. Consequently, when economics says “competition is
desirable,” this is often interpreted to mean that struggle among market players
is desirable. There is a grain of truth to this interpretation, but it misses the main
point.
The popular view judges competition mainly on fairness, so market power on
the supply side is not a problem provided that demand is similarly endowed.
Competition is now in vogue with many regulators, and many who have spent a
lifetime passing judgment on the fairness of prices have taken up the call to “let
the market do it.” They see their new job as making sure the new markets are fair,
that “the playing field is level.” They believe it is only necessary to ensure the
struggle between market players is fair. Because economics promises that competi-
tive markets will be efficient, a good outcome is thought to be assured.
The economic promise of efficiency is not predicated on a fair struggle. Two
fairly matched “competitors” do not approximate what economics means by
competition. For example, economics makes no guarantee that pitting a monopoly
transco against an equally powerful load aggregator will produce an even moder-
ately acceptable outcome. Economics cannot predict the outcome of this kind of
“competition” and would view this as a very poorly structured market.
The economist’s notion of competition refers to competition among suppliers
or among demanders but not between suppliers and demanders. Competition is
not a struggle between those who want a higher price and those who want a lower
price. The process of economic competition between many small suppliers works
by suppliers undercutting each other’s price in order to take away the others’
customers. This drives the price down to the marginal cost of production but no
lower because at lower prices suppliers would lose money. If supply-side competi-
tion is stiff enough, the market price will be pinned to the marginal-cost floor. This
is the meaning of perfect competition.
When suppliers face such stiff competition that they cannot affect the market
price and must simply accept it and sell all they can sell profitably at this price,
they are said to be price takers. This is the principle requirement for a market to
be perfectly competitive and is the primary assumption on which economic claims
of market efficiency rest.
Generally it takes many competitors, none of which have a large market share,
to produce perfect competition in the economic sense.1 If there are any large
suppliers they are likely to have the ability to profitably raise price. In this case
they are not price takers and are said to have market power. They know they can
affect the supply–demand balance by reducing their output and thereby drive up
the price enough to increase their profit.
1. Under the special and uncommon conditions of Bertrand competition, two competitors are enough.

1-5.2 THE EFFICIENCY OF PERFECT COMPETITION

One economic result is, without doubt, the most prominent in all of economics.
It is the point made by Adam Smith in the Wealth of Nations:
... he intends only his own gain, and he is in this, as in many other
cases, led by an invisible hand to promote an end which was no
part of his intention.2
Vague as this may be, Adam Smith, and later Leon Walras, are correctly credited
with developing the notion that competitive markets harness the profit motive to
produce an efficient and socially useful outcome. This Efficient-Competition Result
has been re-examined many times and modern proofs have resulted in “Nobel
Prizes” for Kenneth Arrow (1972) and Gerard Debreu in (1983).3
Short-Run Competition
The Efficient-Competition Result has limitations. It does not mean that every free
market is efficient, or even that every free market in which suppliers are price takers
is efficient. Because of these limitations, economics has carefully defined both
competition and efficiency and has added two more concepts: well-behaved cost
functions and good information. The modern Efficient-Competition Result can
be summarized as follows:
This result can also be summarized as “a competitive equilibrium is efficient.”4

The three conditions listed above under “competition,” are necessary to guarantee
that the market will reach a competitive equilibrium. If suppliers have small
enough market shares, they will not have the power to change the market price and
profit from doing so, and they will take price as given. This is called acting competi-
tively, but it does not guarantee a competitive equilibrium. First, such an equilibrium
must exist and second, traders must be able to find it. If costs are not well be-
haved—and startup and no-load costs are not—there will be no equilibrium. If
traders lack adequate information, including publicly known prices, they may not
2. Smith is often quoted as saying that a market is “guided as if by an invisible hand.” But a full text
search of The Wealth of Nations reveals only this one use of “invisible hand.” In the same year Smith’s
book was published, George Washington observed, in his first inaugural address that the “Invisible Hand”
(of God) had guided the United States to victory. In fact, both Smith and Washington viewed the invisible
hand, God, and the forces of nature as being nearly synonymous.
3. Economists do not get authentic Nobel prizes. The prize in economics is given by the Bank of Sweden,
not by the Nobel Foundation.
4. This discussion is necessarily far from rigorous and is meant only to convey a general understanding
of the most important concepts. See Mas-Colell et al. (1995) starting on p. 308.

find the competitive equilibrium which consists of an optimal set of trades. (Prob-
lems with ill-behaved, nonconvex, costs and lack of information are discussed in
Chapter 3-8.)
Definitions Perfect Competition

Agents act competitively, have well-behaved costs and good information, and free
entry is brings the economic profit level to zero.
Act Competitively
To take the market price as given (be a price taker).
Well-Behaved Costs
Short-run marginal cost increases with output and the average cost of production
stops decreasing when a supplier’s size reaches a moderate level.
Good Information
Market prices are publicly known.
Long-Run Competition
A short-run competitive equilibrium is (short-run) efficient; it makes the best use
of presently available productive resources. A long-run competitive equilibrium
guarantees that the right investments in productive capacity have been made but
requires that the three short-run conditions be met and adds two new ones. Produc-
tion costs must not possess the conditions for a natural monopoly (see Section 1-
1.1), and competitors must be able to enter the market freely.5 With free entry,
if there are above-normal profits to be made, new suppliers will enter which will
reduce the level of profits. In this way free entry ensures that profits will not be
above normal. A normal profit level is the key characteristic of a long-run competi-
tive equilibrium. Barriers to entry is the term used to describe market characteris-
tics that prevent free entry.
Efficiency and Total Surplus
Almost every proposed market design is declared efficient, but in economics the
term has a specific meaning. The simplest meaning applies to productive efficiency
which means that what is being produced is being produced at the least possible
cost. Minimizing cost is often the most difficult part of the market designer’s
problem, so this meaning is generally sufficient.
When not qualified as productive efficiency, efficiency includes both the supply
and demand sides of the market.6 Efficiency means (1) the output is produced by
the cheapest suppliers, (2) it is consumed by those most willing to pay for it, and
(3) the right amount is produced. These three can be combined into a single criterion
by using the concept of consumer surplus.
5. See Mas-Colell et al. (1995, 334).

6. The term “allocative efficiency” is almost universally used to mean demand-side efficiency. But this
is not the meaning found in economics dictionaries; which include both sides of the market and do not
distinguish it from “efficiency.”

Figure 1-5.2
Total surplus equals the
area between the demand
curve and the marginal
cost curve.
Definitions Productive Efficiency

Production costs have been minimized given total production.
Efficiency
Total surplus has been maximized. This automatically includes minimizing the
cost of what is produced and maximizing the value of what is consumed, as well
as producing and consuming the right amount.
A consumer’s demand curve measures how much the consumer would pay for
the first kilowatt-hour consumed, and the second, and so on. Generally the more
consumed, the less would be paid for the next kilowatt-hour. Because the initial
kilowatt-hours are so valuable, the total value of consumption is generally much
greater than the amount paid. The difference between the maximum a consumer
would pay as revealed by the consumer’s demand curve and what the consumer
actually does pay is the consumer’s surplus.
Profit is analogous to consumer surplus and is often called producer surplus.
It is total revenue minus total cost, while consumer surplus is total value to consum-
ers, V, minus total consumer cost, CC. (V is sometimes called gross consumer
surplus, and is the area under the demand curve for all consumption.) Both V and
CC are measured in dollars. If the sum of profit and consumer surplus is maximized,
the market is efficient, and all three of the above criteria follow.
When profit (R ! C ) and consumer surplus (V ! CC ) are added, the consumer
costs (CC ) and producer revenue (R ) cancel because they are the same. The result
is total surplus (V ! C ), consumer value minus producer cost. Consequently
efficiency is the same as maximizing total surplus.
The Efficient-Competition Result
Result 1-5.1 Competitive Prices Are Short- and Long-Run Efficient
If productions costs are well behaved so competitive prices exist, these prices will
induce short-run (dispatch) efficiency and long-run (investment) efficiency.

Note that the Efficient-Competition Result does not say that a market’s prices will
be competitive if costs are well behaved. That conclusion requires the lack of market
power and good information. The Efficient-Competition Result says that if market
prices are competitive then supply and demand will be efficient. Assuring competi-
tive prices is the main problem addressed by this book.
Problems Caused by Production Costs

The Efficient-Competition Result depends on well-behaved production costs and
these cannot be designed. Either efficient generators have well-behaved costs or
they do not. If they do not, and costs are sufficiently problematic, then a standard
competitive market design cannot be depended on to provide an efficient outcome.
Cost problems present the most fundamental challenge. Three different problems
receive attention: (1) nonconvex operating costs, (2) the fixed costs of investment,
and (3) total production costs that decrease up to very large scales of production.
The third problem is the problem of natural monopoly and was discussed in Chapter
1-1.
The second problem is the subject of the most frequent misconception, an
extremely pessimistic one. It holds that ordinary fixed costs are sufficient to disrupt
a competitive market. This pessimism about competition is often accompanied by
optimism about less-competitive free markets. The belief is that if the market can
avoid the problems of a competitive equilibrium free-market forces will produce
a good outcome. Perhaps, it is argued, if the market is not monitored too closely,
generators will exercise market power and thereby earn enough to cover fixed costs
and keep themselves in business. Chapter 2-1 shows that fixed costs are not a
problem for competition and the proposed noncompetitive remedy is unnecessary
and detrimental.
The problem of nonconvex operating costs is the most difficult. The cost of
starting a generator makes generation costs nonconvex because it makes it cheaper
per kWh to produce 2 kWh than to produce 1kWh. This causes the market to lack
a competitive equilibrium and could easily cause inefficiency in the dispatch of
an otherwise competitive market. To circumvent this problem, some markets use
a unit-commitment auction that attempts to replace a standard “classic” competitive
equilibrium with a different equilibrium which is still efficient. Chapter 3-9 dis-
cusses this problem in detail and suggests it may be of minor importance and that
a standard competitive market might still provide a very high level of efficiency.
These conclusions are part of a larger pattern. Competitive markets are difficult
to design, and none of the three basic requirements of efficiency can be achieved
to perfection. But if the two controllable ones, price taking behavior and good
information, are well approximated, a very efficient market will result. The quirks
of a competitive equilibrium are not much of a problem, but designing a competitive
market requires a great deal of care.

1-5.3 SHORT- AND LONG-RUN EQUILIBRIUM DYNAMICS

Markets are never in equilibrium, but economics focuses primarily on their equilib-
rium behavior. The ocean is never in equilibrium, yet it is always found at the lowest
elevations where physics predicts its equilibrium to be. In equilibrium the ocean
would have no waves. Although markets, like oceans, have “waves,” they too
usually stay near their equilibrium. An equilibrium may change over time as the
globe warms and the ice caps melt, but this does not prove the equilibrium uninter-
esting. A market’s equilibrium is a useful guide to its behavior, even though the
market is never exactly in equilibrium.
Both the supply and the demand side of the market
The Meaning of adjust their behavior in order to produce a market
Short Run and Long Run equilibrium, but competitive economics is primarily
7
These concepts do not, as is often supposed, refer concerned with the supply side. This section explores
to specific periods of time but instead refer to the the forces that push the supply side of a market toward
completion of particular market adjustment pro- a competitive equilibrium.8
cesses. “In the short run” indicates that adjustments
in the capital stock (the collection of power plants)
are being ignored, but adjustments in the output of The Short-Run Equilibrium
existing plants are being considered.
Marginal cost is the cost of producing one more unit
The phrase “in the long run” indicates adjustments
of output, one more kilowatt-hour. It is also approxi-
in the capital stock are not only being considered but
are assumed to have come to completion. This is a mately the savings from producing one less kWh. In
useful abstraction. If the market has not recently this section and the next, these are assumed to be so
suffered an unexpected shock, it should be near a close together that no distinction is necessary, which
state of long-run equilibrium because business is typically the case. Chapter 1-6 pays a great deal of
spends a great deal of effort attempting to discern attention to the special case where these are different.
future conditions, and for the last five years, today In a competitive market suppliers are price takers.
was the future. They cannot change the market price profitably, so
Of course mistakes are made and markets are they consider it fixed. Price taking also means they can
never in exact long-run equilibrium. But mistakes are
sell all they want at the market price, but they cannot
as often optimistic as pessimistic, and consequently
a long-run analysis is about right on average. How- sell anything at a higher price. Most markets are not
ever, a newly-created market is more likely than perfectly competitive, and suppliers find that at a
most to be far from its equilibrium. higher price they sell less but more than nothing. This
will be ignored as the present purpose is to analyze
how a market would work if it were perfectly competitive.
A short-run competitive equilibrium determines a market price and a market
quantity traded. To bring the market into equilibrium, two dynamic adjustment
mechanisms are needed: (1) a price adjustment and (2) a quantity adjustment. In
most markets suppliers adjust both, although in some, buyers set the price.
7. Demand-side management concerns itself primarily with information problems on the demand side of
the market. These problems also deserve attention.
8. For a more complete treatment of the microeconomics of competition presented with power markets
in mind, see Rothwell and Gomez (2002).

Quantity Adjustment. A price taking supplier will increase output if its marginal
cost, MC, is less than the market price, P, and will decrease its output if MC > P.
Its profit increases by (P ! MC ) for every unit produced when P is higher than MC
and decreases by (MC ! P ) when P is lower.
Price Adjustment. Whenever demand exceeds supply, suppliers raise their prices,
and whenever supply exceeds demand, they lower prices.
Equilibrium. The quantity adjustment dynamic causes the marginal cost to equal
the market price in a competitive market. The price adjustment dynamic causes
the quantity supplied to equal the quantity demanded. When supply equals demand,
the market is said to have cleared, and the price that accomplishes this is called
the market-clearing price, or the equilibrium price, or, for a competitive market,
the competitive price. Together the two adjustment mechanisms bring a competitive
market to a competitive equilibrium.
Price Taking vs. Price Adjustment. Notice that

Price Taking and Price Adjusting “price taking” suppliers adjust their prices in order to
Suppliers typically name their price. For example, clear the market. This is not a contradiction. “Price
most retailers put price tags on their wares, and taking” is something that happens in the quantity-ad-
customers pay those prices. So how can suppliers justment dynamic but not in the price-adjustment dy-
be “price takers?”
namic. Price takers “take the price as given when com-
Because “price-taking” has a specialized mean-
puting their profit-maximizing output quantity.” This
ing, suppliers can be both price takers and price
adjusters at the same time. Suppliers take price as means they assume that their choice of output will not
given when deciding how much to produce and affect the price they receive for it.
adjust their price if they notice excess supply or The quantity dynamic, which causes MC to equal
demand in the market. the market price, acts as a coordinating mechanism
among suppliers because there is only one market
price. This is why public knowledge of the market price is a key assumption of
the Efficient-Competition Result. Because all suppliers have the same marginal
cost in the competitive equilibrium, no money can be saved by having one produce
more and another less. This is what makes production efficient.
Some will object to this result on the grounds that coal plants have lower
marginal costs than gas turbines even in a competitive equilibrium. This objection
is based on a misunderstanding of the definition of “marginal cost,” which will
be explained in the following chapter.
The Marginal-Cost Pricing Result

Result 1-5.2 Competitive Suppliers Set Output So That MC = P
A competitive producer sets output to the level at which marginal cost equals the
market price, whether or not that is the competitive price. This maximizes profit. (MC = P
for all suppliers.)

The Long-Run Equilibrium

The process of long-run competition involves investing in plant and equipment,
not simply changing the output of existing plants. This dynamic requires a definition
of profit. Profit is of course revenue minus cost, but economics defines costs more
broadly than does business. Economics, and this book, define cost to include a
normal rate of return on all investment. This rate of return is defined to include
a risk premium. If a supplier covers its costs, it automatically earns a normal rate
of return, including an appropriate risk premium, on its entire investment. Under
this definition of “normal,” a business that earns more is considered to be worth
investing in, and a business that earns less is not. A normal business investment,
therefore, has revenues that exactly cover all its costs in the economist’s sense.
Because profit equals revenue minus cost, a normally profitable supplier earns
zero profit.
Definition (Economic) Profit

Revenue minus total cost, where total cost includes a normal, risk-adjusted, return
on investment. The normal (economic) profit level is zero. (Business defines a
normal return on equity to be profit, while economics defines it as covering the
cost of equity.)
Short-Run Profit
Revenues minus short-run costs which include variable, startup and no-load costs.
The “profit function,” defined in Chapter 2-7, computes short-run profits.
As defined, profit is synonymous with long-run profit which is different from

short-run profit which does not include the cost of capital; that is, it does not include
any return on investment. Consequently, short-run profit is expected to be positive
on average so these profits can cover the fixed cost of capital.
Result 1-5.3a Under Competition, Average Economic Profit Is Zero

In a long-run competitive equilibrium, the possibility of entry and exit guarantees
that profits will be normal, which is to say zero.
Result 1-5.3b Under Competition, Fixed Costs Are Covered
When profit is zero, all costs are covered including fixed costs, so in the long run,
competition guarantees that fixed costs will be covered.
Result 1-5.3c A Supplier with a Unique Advantage Can Do Better
If a supplier has access to limited cheaper inputs (hydro-power or geothermal
energy), it will have greater profits. If the advantage is unlimited, it has a natural
monopoly.
If the expected market price is so low that a supplier cannot enter the market
and cover all costs, no supplier will enter. More specifically, if a new generation
unit cannot cover all costs, no new units will be built. The result will be a gradually
diminishing supply of generation (due to retirements of old plants) in the face of

gradually increasing demand. This tightening of the market will cause the price
to rise, and eventually price will be high enough to cover all costs.
Similarly, if price is so high that costs are more than covered, suppliers will
build new generating units. This will increase supply and cause the price to fall.
The result of this long-run dynamic is that the profit in any competitive market
returns to the normal level of profit (zero) in the long-run competitive equilibrium.
1-5.4 WHY IS COMPETITION GOOD FOR CONSUMERS?

In the long-run producers cover their fixed costs, and in the short run total surplus
is maximized, but what consumers want is a low price. Does competition provide
the lowest possible price?
Not in the short run. In the short run, it is possible to design market rules which
lower the market price without reducing supply. This is difficult but possible. But
at a lower price producers will not cover their fixed costs. This will make future
investors think twice. The result will be a risk-premium added to the cost of capital
and future production will be more costly than it would have been had cost been
left at the competitive level.
Competition does not guarantee the lowest possible price at any point in time.
Instead it guarantees that suppliers will just cover the long-run total costs and no
more. It also guarantees that the cheapest suppliers will be the ones producing.
Together these mean production costs (including the long-run cost of invested
capital) are minimized and producers are paid only enough to cover their cost. This
implies that the long-run average cost to consumers is also minimized. No market
design regulated or unregulated can induce suppliers to sell below cost on average.
Competition minimizes long-run average costs of production and long-run average
costs to consumers.

Chapter 1-6
Marginal Cost in a Power Market
The trouble with the world is not that people know too little,
but that they know so many things that ain't so.
Mark Twain
(1835–1910)
S IMPLIFIED DIAGRAMS OF GENERATION SUPPLY CURVES HAVE CON-

FUSED THE DISCUSSION OF MARGINAL COST. Typically, these supply curves
are diagrammed to show a constant marginal cost up to the point of maximum
generation. Then marginal cost becomes infinite without taking on intermediate
values. Typically it jumps from about $30 to infinity with only an infinitesimal
increase in output. Mathematics calls such a jump a discontinuity. In fact, the curve
would be discontinuous if it jumped only from $30 to $40.
The definition of marginal cost does not apply only to the points of discontinuity.
Hence it does not apply to a right-angle supply curve at the point of full output,
neither does it apply to the points of a market supply curve at which it jumps from
one generators marginal cost to the next. Unfortunately market equilibria sometimes
occur at such points, and concerns over market power often focus on them. Attempts
to apply the standard definition at these points can produce confusing and erroneous
results.
Fortunately, the definition is based on mathematics that generalizes naturally
to discontinuous curves. Applying this generalization to the textbook definition
clears up the confusion and restores the economic results that otherwise appear
to fail in power markets. For example, in power markets, as in all other markets,
the competitive price is never greater than the marginal cost of production.
Chapter Summary 1-6 : Individual supply curves are often constructed with
an abrupt end that causes the market supply curve to have abrupt steps. The standard
marginal-cost definition does not apply at such points. Instead, left- and right-hand
marginal costs should be used to define the marginal-cost range. Then the competi-
tive price, which remains well defined, will always lie within that range. A market
price exceeding the marginal-cost range indicates market power.

CHAPTER 1-6 Marginal Cost in a Power Market 61
Section 1: The Role of Marginal Cost. Marginal costs play a key role in cost-
based power auctions because they help determine the competitive price. They also
play a key role in analyzing market power and gain their importance by defining
the competitive supply curve for individual generators. To find the market (aggre-
gate) supply curve, individual supply curves are summed horizontally.
Section 2: Marginal-Cost Fallacies. In power-market analysis, marginal cost
is often defined as the cost of the last unit produced, but this definition is found
in no economics text. A second fallacy asserts that when marginal cost is ambigu-
ous, the competitive price is ambiguous. Together these lead to a variety of errone-
ous conclusions, such as “the competitive price is above marginal cost,” and “the
competitive price is ambiguous.”
Section 3: The Definition of Marginal Cost. When a marginal-cost curve
is discontinuous (has a sudden jump), marginal cost can be specified only within
a range at the points of discontinuity. This range extends from the left-hand to the
right-hand marginal cost at the point under consideration. For all points where the
curve is continuous, the range is a single point equal to the standard marginal cost.
Section 4: Marginal Cost Results. The competitive price is within the
marginal-cost range of every competitive generator and within the marginal-cost
range of the market. If even one supplier has a supply curve that is continuous at
the market price, the market supply curve is continuous at that price and the
competitive price is equal to the standard marginal cost which is well defined. In
any case, the competitive price is the price at which the supply and demand curves
intersect.
Section 5: Working with Marginal Costs. This book assumes that supply
curves have extremely large but finite slopes rather than the infinite slopes fre-
quently assumed. This is a more realistic assumption and has no practical conse-
quences, but it has the simplifying property of making marginal cost well defined
and the marginal cost of all operating competitive generators equal to the market
price.
Section 6: Scarcity Rent. Scarcity rent is revenue less variable cost and is
needed to cover startup and fixed costs. Economics refers to this as “inframarginal
rent,” and has no separate definition of a scarcity rent. A folk-definition defines
scarcity rent as actual revenue minus the maximum revenue that is collected just
before the system runs completely out of capacity. Used with a stylized model, this
definition has some appeal, but when applied to real systems it is highly ambiguous
and misleading.

Figure 1-6.1
Adding individual supply
curves horizontally to
find the market supply
curve. If B is continuous,
A + B is also.
1-6.1 THE ROLE OF MARGINAL COST

Marginal cost plays a key role in the economic theory that proves a competitive
market is efficient, but there are also two practical uses of marginal cost that
increase its importance in a power market. First, many power markets rely on a
central day-ahead auction in which generators submit individual supply curves and
the system operator uses these to determine the market price. Because price should
equal marginal cost in an efficient market, the auction rules should be informed
by a coherent theory of marginal cost. Second, many power markets suffer from
potential market-power problems which cause the market price to diverge from
marginal cost. Market monitors need to understand this divergence.
Although the competitive market price usually equals the marginal cost of
production, it is not determined by that alone. At times marginal cost is ambiguous,
yet the competitive price is not. Then, marginal value (to customers) plays the
decisive role. The competitive price is determined by the intersection of the market’s
supply and demand curves. Marginal cost determines only the supply curve.
A supply curve can be thought of as answering the question, How much would
a generator produce if the market price were $P/MWh? As explained in Section
1-5.3, price-taking suppliers adjust output until marginal cost equals the market
price. As a consequence, if Q is the quantity supplied at a given price P, then P
must equal the marginal cost. Thus a price-taker’s supply curve and marginal cost
curve are the same.
The market’s supply curve, also called the aggregate supply curve, is found
by summing horizontally all of the individual generators’ supply curves. For a given
price, the quantity supplied by each generator is read horizontally from each
individual supply curve and these quantities are summed to find the market supply.
This quantity is plotted at the given price, as shown in Figure 1-6.1.
Notice that because one generator has a continuous supply curve (no vertical
section) the market has a continuous supply curve. Notice also that when both
generators are operating and have defined marginal costs, they have the same
marginal cost. Section 1-6.3 generalizes this by showing that every operating
generator either has a marginal cost equal to the market price or has a marginal-cost
range that includes the market price.

1-6.2 MARGINAL-COST FALLACIES
Discontinuous Supply Curves

Individual supply curves are almost always drawn as “hockey sticks.” That is, they
are drawn with a slight upward slope (or as flat) until they reach the capacity limit
of the generator and then they are drawn as perfectly vertical (see curve A, Figure
1-6.1). Textbook supply curves usually have a slope that increases gradually (See
curve B, Figure 1-6.1). Curves without a vertical segment are called continuous.
Unfortunately, a generator’s supply curve, as typically drawn, takes an infinite
upward leap when it reaches full output (which is the most common output level
for an operating generator). At this point, marginal cost is not smooth but jumps
from say $30/MWh to infinity with only an infinitesimal change in output.
The smoothness of textbook supply curves plays a crucial role in keeping the
textbook definition of marginal cost simple, and this has led to mistakes and
confusion. Eliminating the confusion requires the introduction of a carefully
constructed definition which applies to the discontinuous supply curves used in
power-market analysis. With this definition of marginal cost, all standard economic
results are found to apply to power markets. Once this is understood, the problem-
atic supply curves can be analyzed correctly with a simple rule of thumb. This
provides guidance when setting the market price in a cost-base auction and when
determining whether market power has been exercised.
Fallacies
Two basic fallacies underlie a series of misconceptions surrounding competitive
pricing and market power. These are (1) the Marginal-Cost Fallacy and (2) the
Ambiguous-Price Fallacy. Both of these will be illustrated using Figure 1-6.2, which
shows a normal demand curve and a supply curve that is constant at $30/MWh
up to an output of 10 GW, the capacity limit of all available generation.
The Marginal-Cost Fallacy takes two forms. The simple form asserts that
marginal cost at Q = 10,000 MW is $30/MWh in Figure 1-6.2. The subtle form
asserts that nothing can be said about the marginal cost at this output level. Some
of the conclusions drawn from these assertions are as follows:
1. The competitive price is $30/MWh, and the market should be designed to
hold prices down to this level.
2. The competitive price is $30/MWh, and this is too low to cover fixed costs,
so marginal-cost prices are inappropriate for power markets.
3. Scarcity rents are needed to raise prices above marginal-cost-based prices.
4. Market power is necessary to raise prices to an appropriate level.
5. The competitive price cannot be determined.

Figure 1-6.2
A normal market
equilibrium for an
abnormal supply curve.
All of these conclusions assume that there is some problem with standard
economics caused by the supply curve coming to an abrupt end instead of turning
up smoothly as it does in undergraduate texts. In fact, economic theory has no
difficulty with this example, and all of the above conclusions are false. Consider
a competitive market, with many suppliers and many customers, described by the
curves in Figure 1-6.2. What if the price in this market were $30/MWh? At this
price, the demand curve shows an excess demand of about 4 GW. Some customers
trying to buy more power are willing to pay up to $70/MWh for another MW of
supply. They will find a supplier and offer to pay considerably more than $30, and
the supplier will accept. This shows that the competitive price is above $30/MWh.
The story will be repeated many times, with different values, until the market price
reaches $70/MWh. At that price every supplier will produce at full output, so the
supply will be 10 GW, and demand will be 10 GW. At any higher price demand
would fall short of supply, so the price would fall, and at any lower price, demand
would exceed supply, so the price would rise. There is nothing unusual about this
equilibrium; it is the classic story of how price clears a market by equating supply
and demand.
The Marginal-Cost Fallacy
Fallacy 1-6.1 Marginal Cost Equals the Cost of the Last Unit Produced
Marginal cost equals the savings from producing less even when this is different
from the cost of producing more.
(Subtle Version)
Nothing can be said about marginal cost at the point where a supply curve ends
or jumps from one level to another.
But shouldn’t price equal marginal cost? In this example, all that can be said
is that marginal cost is greater than $30/MWh. So there is no contradiction between
price and marginal cost, but they cannot be proven to be equal. The desire to pin
down marginal cost precisely seems to arise from a belief that competitive suppliers
should set price equal to marginal cost and thereby determine the market price.
But this logic is backwards. As explained in Section 1-5.3, suppliers set price to

clear the market and set quantity to bring marginal cost in line with price. In this
example, the market-clearing forces of supply and demand determine price unam-
biguously, and although marginal cost is ambiguous, it is greater than $30/MWh
which is enough to determine supply unambiguously. Everything of practical
importance is precisely determined.
The Ambiguous-Price Fallacy

Fallacy 1-6.2 When Marginal Cost Is Ambiguous, so Is the Competitive Price
Competitive suppliers set price equal to marginal cost; thus when marginal cost
is hard to determine, the competitive price is hard to determine.
Having analyzed the example, the preceding list of incorrect conclusions can
be restated in their.
1. The competitive price is not $30/MWh, and the market design should not
hold price to this level.
2. The competitive price is high enough to contribute significantly to fixed
cost recovery.
3. No mysterious “scarcity rent” need be added to the marginal cost of physical
production.
4. Market power is not needed if the market is allowed to clear.
5. The competitive price is $70/MWh.
1-6.3 THE DEFINITION OF MARGINAL COST

The above discussion is accurate but informal. Because of the controversy in this
area, it is helpful to formalize the concepts used in analyzing supply curves with
discontinuities or abrupt terminations.
The MIT Dictionary of Modern Economics (1992) defines marginal cost as “the
extra cost of producing an extra unit of output.” Paul Samuelson (1973, 451) defines
marginal cost more cautiously as the “cost of producing one extra unit more (or
less).” The “or less” is important. The assumption behind this definition is that
producing one more unit of output would cost exactly as much as producing one
less unit would save. This is true for the continuous marginal-cost curves of
textbook economics but not for the discontinuous curves used by power-market
analysts. To discuss the marginal cost of a discontinuous supply curve, the definition
must be extended to include the points of discontinuity where the cost to produce
an extra unit is distinctly greater than the savings from producing one less.
Left- and Right-Hand Marginal Costs

In the example of Figure 1-6.2, the marginal cost of production goes from $30 on
the left of 10 GW to infinity on the right of 10 GW. This is a double complication.
Not only does marginal cost change abruptly, it becomes infinite. The present

Figure 1-6.3
Right- and left-hand
marginal costs.
definitions can be illustrated more clearly with a less pathological marginal-cost

curve.
Figure 1-6.3 shows the total cost curve and the marginal cost curve of a simple
market. The discontinuity is the jump in marginal cost at the 10 GW output level.
To the left of 10 GW the marginal cost is $20/MWh, while to the right it is
$40/MWh. But what is the marginal cost precisely at 10 GW? It is undefined, but,
as every textbook would confirm, the answer is not MC = $20/MWh.
To formalize this definition, it is useful to consider the mathematics of the total
cost curve shown at the left of Figure 1-6.3. To the left of 10 GW, its derivative
(slope) is $20/MWh, while to the right its slope is $40/MWh. But the mathematical
definition of a derivative breaks down at 10 GW, and since marginal cost is just
the derivative of total cost, the definition of marginal cost also breaks down at this
point. Mathematics does define two very useful quantities at the 10-GW point, the
left-hand derivative (slope) and the right-hand derivative (Courant 1937, 199–201).
These are, of course, $20 and $40/MWh, respectively. Because marginal cost is
just the derivative, it is natural to define left-hand marginal cost (MCLH) as the
left-hand derivative, and right-hand marginal cost (MCRH) as the right-hand
derivative. Other points along the total cost curve also have left and right-hand
derivatives, and these are just equal to the normal derivative. Similarly, MCLH and
MCRH are normally equal to each other and equal to standard marginal cost, MC.
The marginal-cost range, MCR, is defined as the range of values between and
including MCLH and MCRH. This definition is motivated by the idea that marginal
cost cannot be pinned down at a point of discontinuity but can reasonably be said
to lie somewhere between the savings from producing one less and the cost of
producing one more unit of output.

Definitions Left-hand marginal cost (MCLH )

The savings from producing one less unit of output.
Right-hand marginal cost (MCRH )
The cost of producing one more unit of output. When this is impossible, MCRH
equals infinity.
The marginal-cost range (MCR )
The set of values between and including MCLH and MCRH.
1-6.4 MARGINAL COST RESULTS
Refining the Marginal-Cost Pricing Result

In Figure 1-6.2, the MCLH at 10 GW is $30/MWh, but what is the MCRH? It is
tempting to say it is undefined, but again mathematics provides a more useful
answer. The MCRH at 10 GW is infinite. This definition is both mathematically
sound and useful because it allows a simple rewriting of the standard economic
results concerning marginal costs.
Result 1-6.1 Competitive Suppliers Set Output so MCLH <– P <– MCRH
A competitive producer sets output to a level at which its marginal-cost range,
MCR, contains the market price, P, whether or not that is the competitive price.
First, a price-taking supplier will decrease output as long as P < MCLH because
producing one less unit will save MCLH and cost only P in lost revenues. Thus, the
savings is greater than the cost. Similarly, if MCRH < P, the supplier will increase
output. Thus whenever P lies outside the range between left- and right-hand
marginal costs, the supplier will adjust output. When the range is below P, output
is increased, which raises the range and vice versa when MCR is above P. As a
result, the marginal-cost range will end up encompassing P.
This means that in a competitive market, price will never exceed marginal cost;
this would violate basic economics. Technically, P > MC can never be proven true
in a competitive market.1 Competitive price will always be less than or equal to
left-hand marginal cost, and there is no need for it exceed this value for fixed cost
recovery.2
1. Even those who best understand these concepts sometimes add to the confusion. “Thus in the absence
of market power by any seller in the market, price may still exceed the marginal production costs of all
facilities producing output in the market at that time.” (Borenstein 1999, 3) “. . . the price of electricity
has to rise above its short-run marginal cost from time to time, or peaking capacity would never cover its
fixed costs.” (Green 1998, 4).
2. Part 3 discusses “nonconvex costs,” complexities of the production cost function that require deviations
from marginal cost. Essentially this means that startup costs and other short-run, avoidable costs must be
covered by price.

Figure 1-6.4
The smallest possible
change in the supply
curve of Figure 1-6.2
restores all normal
economic properties.
This result can be extended from a single producer to the whole market. The
MCRH of the market is the least cost of producing one more unit, so it is the mini-
mum of the individual marginal-cost ranges. Similarly, the market MCLH is the
maximum individual MCLH. In a competitive market, every supplier is a price taker
and adjusts its output until P is within its marginal-cost range. Thus P is less than
or equal to every individual MCRH, so it is less than or equal to the MCRH of the
supplier with the lowest MCRH, which is the MCRH of the market. Similarly, P is
greater than or equal to the MCLH of the market.
The range from the market MCLH to the market MCRH is contained within the
marginal-cost ranges of each individual supplier. If even one supplier in the market
has MCLH = MCRH, the market will also have this property. In other words, if even
one supplier has a well defined-marginal cost at the market price, then the market
itself has a well-defined marginal cost.
The System-Marginal-Cost Pricing Result

Result 1-6.2 Competitive Price Equals System Marginal Cost
In a competitive market, price is within the marginal-cost range of every generator
supplying power. It is thus within the market’s marginal-cost range. If even one
operating supplier has a continuous marginal cost curve, the competitive price
actually equals marginal cost as defined by the aggregate supply curve.
Finding the Competitive Price

Fortunately the above results are needed only for untangling the current confusions
over marginal cost. They demonstrate, among other things, that price does not
exceed marginal cost in a competitive power market.
Fortunately, these results are not needed to find the competitive equilibrium,
which is determined, as in any other market, by the intersection of the supply and
demand curves. This is most easily seen by smoothing out one of the problematic
supply curves very slightly.
Standard economic theory applies once the vertical segments have been removed
from the cost curves. This can be done with an arbitrarily small change in its shape.

As shown in Figure 1-6.4, giving the marginal cost curve a nearly, but not perfectly,
vertical slope makes no noticeable difference to any economic result. And this is
how it should be. Economics should not and does not depend on splitting hairs.
Notice that in the finitely-sloped model, price really does equal marginal cost at
the intersection of the two curves. The price and quantity dynamics of a market
with the vertical supply curve will be essentially the same as those of the continuous
market. In this example, at a price different from $70 and a quantity different from
10 GW, the markets have essentially identical gaps between supply and demand
and between price and marginal cost. So they adjust price and quantity in the same
way.
Result 1-6.3 Supply Intersects Demand at the Competitive Price

To find the competitive price and the marginal cost, draw the supply and demand
curves, including the vertical parts of the supply, curve if any. The intersection
of supply and demand determines “MC,” P, and Q.
This demonstrates that the standard method of finding the competitive equilib-
rium works even when the marginal cost curves have infinite slopes. Of course
if the slope is infinite at the intersection of supply and demand, marginal cost will
be technically undefined. Yet pretending that the true marginal cost is determined
by this simple short cut will never give the wrong answer to any real-world question.
1-6.5 WORKING WITH MARGINAL COSTS

Discussing left- and right-hand marginal costs and the marginal-cost range is
cumbersome and unnecessary. If every vertical segment of a marginal cost curve
is replaced with a nonvertical but extremely steep segment, the new curve will be
continuous and will not jump from one value to another. Such a change may or
may not improve its accuracy, but in either case it will make no detectable differ-
ence to any economic prediction of consequence.
This book will tacitly assume all supply curves and marginal cost curves that
are depicted as having vertical segments actually have extremely steep but finite
slopes. In other words, all marginal-cost curves are assumed to be continuous.
Consequently, marginal cost is always a well-defined single value.
For example, a supply curve that is constant at $30/MWh up to a
maximum output of 500 MW can be replaced with one that is identical
up to 500 MW and then slopes upward linearly reaching a value of
$30,000/MWh at an output of 500.001 MW. No measurement, however
careful, could discern the difference. Yet this supply curve, being
continuous, has a well-defined value (marginal cost) at every level of
output.
In fact most, if not all generators, have continuous marginal cost
curves. Typically, they have an emergency operating range above
their nominal maximum output level and are willing to produce in this
region if well paid or coerced. Most generators in PJM include such

an emergency operating range in their bids and the total capacity available in this
range is 1,900 MW out of a total installed capacity of about 60,000 MW.3 As long
as there is one such generator in a market, the market’s marginal cost curve is
continuous. Real markets always have well-defined marginal costs and the competi-
tive price equals that marginal cost. The difficulties resolved in this chapter only
matter for the “simplified” diagrams used by power-market analysts.
This book also will use the same simplified diagrams but without taking the
vertical segments literally. Such supply curves will have constant marginal costs
up to the nominal “maximum” output level, but above that marginal costs will
increase rapidly. If the supply curve is flat at $30 but the market price is $50, the
generator’s marginal cost will be $50 and it will produce on the steeply sloped
segment. When referring to such a generator, it is both wrong and confusing to say
its marginal cost is $30 as is the custom. To avoid this confusion, the marginal cost
of a generator’s supply curve to the left of the “maximum” output level will be
termed its variable cost. This is not entirely standard, but it is in keeping with the
term’s normal usage which refers to all costs that vary with the output level.
1-6.6 SCARCITY RENT

“Scarcity rent” has no formal economic definition but many popular
meanings.4 Although several are useful, most do not lend themselves
to careful analysis. However, one essential economic concept comes
close to the popular meaning. Scarcity rent will be defined as revenue
minus variable cost.5 Economics refers to scarcity rent as inframarginal
rent.
In the figure at the left, when demand is described by D1, both
generators are producing at full output, and load would be willing to pay
either generator more than its variable cost of production if it would
produce more. In this sense they are both scarce and both earn scarcity
rents.
With demand reduced to D2, as shown in the lower half of the figure,
generators of type G2 have excess capacity and are no longer scarce and
earn no scarcity rent; their variable costs equal the market price. Genera-
tors of type G1 are still scarce because load would be more than willing
to pay their variable cost if they would produce more. If G2 had a
variable cost of $1,000/MWh so that G1 were earning a rent of, say,
$950/MWh because G1 could not satisfy the entire load, G1 would
commonly be seen as in scarce supply. The above definition coincides
with an important concept of economics and with the common meaning of scarcity.
3. Personal communication from Joe Bowring, head of PJM’s Market Monitoring Unit, January 7, 2002.
4. Samuelson (1973, 623) comes close to using the term when he says “Competitively determined rents
are the results of a natural scarcity.” His definition of such rents is the long-run analog of the short-run
definition of scarcity rent given here.
5. This is greater than short-run profit by the amount of startup costs and no-load costs which will be
ignored until Part 3.

Figure 1-6.5
Folk-definition of
scarcity rent.
A Folk Definition of Scarcity Rent

Sometimes, in the field of power system economics, scarcity rent is defined
as actual revenue less the highest revenue earned before total generation becomes
scarce. This might be called a “folk definition.” The notion is that until the system
runs out of capacity, price increases are due to increases in marginal cost, but after
that point they are driven up by ever increasing scarcity. In an idealized model,
this definition has some appeal.
Definition Scarcity Rent

Revenue minus variable operating cost (which do not include startup costs and
no-load costs).
Say there are only ten types of generators on the market, and call the one with
the highest variable cost the peaker. Next assume that there are no out-of-date
generators with higher variable costs installed in the system. Finally assume that
no installed generator has an emergency operating range in which its marginal
costs increase dramatically as it increases its output beyond its normal rating. With
these assumptions, peakers will earn enough to cover more than variable cost only
when the system runs out of capacity. In other words, peakers can cover their fixed
costs only from scarcity rents but not from any nonscarcity inframarginal rents.
All other generators cover their fixed costs from a combination of scarcity and
nonscarcity rents. The left half of Figure 1-6.5 illustrates this property of an ideal-
ized supply curve.
The folk definition has the advantage of allowing the following types of state-
ments which seem designed to segregate scarcity conditions from the normal
operating conditions of the market.
1. Scarcity rents pay capital costs of units that run infrequently.
2. In the long-run competitive equilibrium, scarcity rents are just high enough
to cover the fixed costs of peakers.
3. Scarcity rents are paid only infrequently.

This appears to ratify the view that power markets are qualitatively different in their
cost structure and consequently cannot be analyzed with the standard marginal-cost
apparatus.
In the idealized model, these statements are true, although they give the impres-
sion that scarcity rents are mainly or wholly associated with peakers. In fact, under
the folk definition, every type of generator receives the same amount of scarcity
rent per MWh. In addition, the average scarcity rent in $/MWh does not equal the
fixed cost of peakers but is greater by a factor of one over the duration of the
peaker’s use, something that is not easily determined.
Two problems with this definition make it unworkable in a real market. First,
there are likely to be old generators on the system with variable costs greater than
the most expensive new generator that would be built (the peaker). In this case
scarcity will not set in until the old generator is at full output. This will expand the
nonscarcity rents and shrink the scarcity rents to the point where they no longer
cover the fixed costs of a peaker. Second, there will be some (probably many)
generators with marginal cost curves that continue on up to some very high but
ill-defined value. This will reduce scarcity rents to some negligible and indetermin-
able value. Proving scarcity rents exist requires proving price is above the point
where the supply curve becomes absolutely vertical; absolutes are notoriously hard
to prove.
Because of these shortcomings and the limited usefulness of the folk definition,
this book will use only the definition given above that coincides with “inframarginal
rents,” a term that has proven itself useful in economics. This is in keeping with
the chapter’s general view that generation cost functions present no new problems
of consequence and require only a minimal expansion of the definition of marginal
cost and then only to deal with the stylized mathematics of discontinuous cost
functions.

A Marginal-Cost Example
P Four suppliers can each produce 100 MW but no more.

P Each supplier has constant marginal cost (MC) up to this limit.
P Marginal costs and demand are as shown in the figure.
If demand is given by D1,

1. The competitive price is $60/MWh.
2. Any higher price indicates market power.
3. If the market is competitive, no supplier has MC < $60/MWh.
If demand is given by D2 and the suppliers are price takers,

1. The market price (P ) will be $100/MWh.
2. No generator will have a marginal cost of less than $100/MWh.
3. No market power is exercised at this price.
4. P is greater than the cost of the last unit produced ($60/MWh).
In both cases the marginal-cost rule for competition is

MCLH # P # MCRH *
This is sufficient to determine the competitive market price and output.
* MCLH is the savings from producing one unit less. MCRH is the cost of producing
one unit more and is considered arbitrarily high, or infinite, if another unit cannot
be produced.

Chapter 1-7
Market Structure
The work I have set before me is this . . . how to get rid of the evils of competition while retaining
its advantages.
Alfred Marshall
(1842-1924)
P OOR MARKET STRUCTURE POSES THE GREATEST THREAT TO THE

HEALTH OF POWER MARKETS. “Structure” refers to properties of the market
closely tied to technology and ownership. The classic structural measure is a
concentration index for the ownership of production capacity. The cost structure
of an industry, another component of market structure, describes both the costs
of generation and the costs of transmission.
Most aspects of market structure are difficult to alter and some, such as the high
fixed costs of coal-fired generation, are impossible. But power markets contain
some unusual technology-based arrangements that can easily be altered or that
require administrative decisions regarding their operation. These arrangements are
part of the market structure and require design just as do the architectural compo-
nents described in the next chapter.
The notion of market structure developed as part of the “structure-conduct-
performance” paradigm of industrial organization in the early 1950s. The present
discussion, however, is based on the structure-architecture-rules classification of
market-design problems presented by Chao and Wilson (1999a) and Wilson (1999).
The present chapter extends their definition of structure, particularly in the direction
of administered reliability policies.
Chapter Summary 1-7 : Market structure has a decisive impact on market power
and investment. The second demand-side flaw, the ability of users to take power
from the grid in real time without a contract (see Section 1-1.5), makes structural
intervention necessary. Regulators must trade-off price spikes against involuntary
load shedding, thereby largely determining the incentives for investment in genera-
tion capacity.

CHAPTER 1-7 Market Structure 75
Regulators will continue to control transmission investment either directly or

through incentives. The strength of the transmission grid affects the market’s
competitiveness. Demand elasticity can be greatly enhanced by improved metering
and real-time billing, which can dramatically reduce market power and improve
the stability of generation investment. Long-term power contracts and supply
concentration also play a key role in controlling market power.
Because market structure is typically difficult to affect, it is usually ignored
by policy makers. The power market is unusual in this respect, partly because it
is new, flawed, and utilizes the grid, a regulated shared asset. The present lack of
attention to market structure in the United States is producing unnecessarily high
price spikes, boom-bust investment cycle and problems with market power during
the short-supply phase of the cycle. Part 2 discusses the structural design issues
most crucial to the solution of these problems.
Section 1: Reliability Requirements. The system operator buys energy and
various grades of operating reserves to balance the system and to provide reliability.
The number and type of submarkets used for this purpose are a matter of market
architecture, but the reserve requirements and price limits imposed administratively
are matters of market structure.
The structure of reliability requirements determine not only short-term reliability,
but the height and frequency of price spikes and therefore long-run investment in
generation and long-run reliability. These consequences of market structure are
often overlooked, so the design of the reliability structure is often inappropriate,
sometimes with serious consequences.
Section 2: Transmission. The transmission grid determines a significant part
of the cost structure of the wholesale power industry. Investment in wires and the
structure of access charges both have significant impacts on long-distance trade
and thus on the market’s competitiveness. This is an aspect of market structure
which can and must be influenced by policy.
Section 3: Effective Demand Elasticity. Demand appears to be inelastic
because it is not given real-time price signals. It would be more elastic if customers
had a reason to purchase and use the equipment necessary for responding to price
changes. Policy can easily influence two key aspects of market structure, meters
and billing, that would greatly increase effective demand elasticity.1 This would
reduce the necessary investment in peak generating capacity, but more importantly,
it would curb market power.
1. When considering the wholesale market, retail billing is part of structure. When considering the retail
market it is simply a market outcome.

Section 4: Long-Term Contracts. The structure of the spot market includes

the extent of long-term obligations by suppliers.2 In new power markets, policy
can influence this structural component. It can require vesting contracts for newly
divested generation and limit the amount of divestiture so obligations to loads at
regulated prices are retained. Increasing the extent of long-term obligations increases
the competitiveness of the spot market.
Section 5: Supply Concentration. HHI, the classic structural index, measures
the concentration of the ownership of productive capacity. While this aspect of
market structure significantly affects market power, it can be difficult to change,
though it is easily influenced in new markets by divestiture requirements.
1-7.1 RELIABILITY REQUIREMENTS

The Second Demand-Side Flaw. Electricity customers can take power in real
time without a contract and cause other customers to be blacked out, although in
most cases, they will suffer no disruption of their own. (See Section 1-1.5.) Rotating
blackouts are implemented without regard to contracts or consumption levels. At
such times, the system operator is faced with a difficult choice; it can pay even
more for power or it can blackout more customers. The choice is easily made in
favor of reliability when the price is within ten times the long-run average, but when
it increases to 100 times normal, the correct choice is less obvious. At some price,
every system operator chooses to interrupt customers rather than pay the price.
Operating Reserve Requirements. Deliberate interruptions of service are rare
events, but their possibility has an enormous impact on the market. To avoid them,
power systems buy several kinds of operating reserves, generators that are paid
to be ready to provide power at a moment’s notice. Together these reserves amount
to approximately 10% of load at any given time.
Normally the effect of operating reserves on market price is modest. When
available capacity exceeds load by 10% or more, a competitive market will hold
the price of power down to approximately the variable cost of the most expensive
generator producing power. In the normal operating range, this is under $100/MWh.
In a year when system load never exceeds the normal range, prices are modest in
spite of the administered operating reserve requirement. In fact prices are so low
that generators cannot cover their fixed costs. If this situation were to continue year
after year, no new generators would be built, and this would be the right outcome.
Price Spikes and Investment. As consumption increases, supply becomes
tighter, and the system operator finds it impossible to maintain a 10% operating
reserve margin at all times. When reserves run short, the system operator offers
to pay more either for power or reserves and this drives up the price of energy. This
dynamic is the source of the high prices that induce investment. The heart of this
process is the administrative decision about how much to pay depending on how
2. When considering the market for long-term contracts, this is an outcome and not part of the market’s
structure.

low operating reserves are. This links concerns about reliability with the incentive
to invest.
The high prices paid by the system operator to meet reliability requirements
control all high prices in the power market. Customers can choose between forward
purchases and letting the system operator buy power for them in real time, and they
choose the cheaper alternative. This holds the price in all forward markets down
to the price that the system operator charges for real-time purchases, and it charges
only as much as it pays. So the price paid by the system operator sets a limit on
the price paid in all markets.3 During the intervals when its requirements for
operating reserves are not met, the system operator’s pricing policy controls the
price spikes and high forward prices that induce investment.
Structure vs. Architecture of the Balancing Market. The balancing market
keeps supply and demand in balance until the system operator is forced to balance
the system by shedding load. This market must be administered by the system
operator, but it may include a sizable bilateral component. It may be integrated with
the markets for operating reserves, or these may be separate. These are questions
of market architecture. At a more detailed level, there are innumerable choices
concerning market rules.
The market rules and architecture do not determine the height and duration of
real-time price spikes, nor how closely voltage and frequency will be maintained,
nor the chance that the system will not recover from an unexpected generation
failure.4 These fundamentals are determined by the structure of the balancing
market. The rules and architecture determine how efficiently trades are organized,
who gets their transaction terminated when reliability is threatened, and how closely
prices approximate the competitive level.
The structure of the balancing market is in part determined by the interconnec-
tion’s reliability authority (NERC) and in part by local design. It is also influenced
by the regional regulatory authority (FERC) when it caps real-time prices. The
accuracy of balancing and short-term reliability are largely determined by the
structure of the balancing market. Less obviously, price spikes, generation invest-
ment and long-term reliability are also largely determined by the balancing market.
From this perspective, current balancing market structures appear haphazard and
inappropriate. Part 2 examines these problems and presents methods for designing
a better structure.
1-7.2 TRANSMISSION
Market structure includes the arrangement and capacity of power lines. Insufficient
capacity can cause bottlenecks and local market power while additional capacity
can expand the size of the market and reduce market power. The effect of transmis-
sion on market power makes transmission expansion more valuable in a competitive
than in a regulated market.
3. The proposition that a price limit in the real-time market effectively caps the forward markets was well
tested by the California Power Exchange.
4. This assumes that the rules are at least modestly functional.

While this structural component can only be changed slowly, it is one whose
design cannot be ignored. Unless there is a reversal of the 100-year trend, new lines
will be needed for reliability and economy as demand for power continues to grow.
These will be built either at the direction of market designers working for the system
operator or by for-profit transcos whose incentives, determined by regulators,
govern their choice of transmission upgrades.
So far, this structural component has received the most attention, and the design
issues have proven to be complex and contentious. Australia, New Zealand,
Argentina, Alberta, California and Britain have all provided examples of transmis-
sion design issues that have been resolved with only partial success. Alberta’s case
is particularly interesting because it has highlighted the trade-offs between transmis-
sion and generator location. Alberta’s for-profit Transmission Administrator chose
not to upgrade Alberta’s major transmission path and instead offered long-term
incentives to new generation investment in locations that would alleviate the need
for more transmission. This was quite successful, but two years later the TA strongly
favored the now unnecessary upgrade in order to facilitate exports from northern
Alberta to the United States.
Alberta’s transmission issues did not involve the impact of transmission on
market power, but in both Australia and California, this has been a prominent
concern. Due to their historic rivalry, California’s two largest investor-owned
utilities are connected by inefficiently small power lines (Path 15). As a conse-
quence, these frequently inhibit trade, causing increased generation costs and
increased market power. Besides such major bottlenecks, there are dozens of smaller
ones causing infrequent but locally severe market power. Over half of the generators
in California are regulated at times, supposedly because of these transmission
limitations.5 While a must-run classification is sometimes only an excuse for a
profitable regulatory “must-run” contract, many cases reflect real transmission
constraints.
1-7.3 EFFECTIVE DEMAND ELASTICITY

Perhaps the most dramatic structural problem of power markets is the almost
complete lack of demand response to fluctuations in the wholesale price. It is
conceptually dramatic because it sometimes prevents the intersection of the market’s
supply and demand curves, a flaw so fundamental it is not addressed in any econom-
ics text. Its consequences grab headlines when California suffers blackouts and
New York prices surge above $6,000/MWh. It is the flaw that makes market power
a major issue in so many power markets that are otherwise well structured. But
what is puzzling is that the remedy is straightforward and would pay for itself. That
such a cheaply fixed fundamental flaw has been consistently overlooked is due
in part to confusion over the supposed costliness of real-time rates. But it is also
part of the pattern of ignoring structural problems in favor of architectural problems
and disputes over market rules.
5. “Must-run” generators are paid regulated prices, which are sometimes quite high, when they are
required to run due to transmission constraints.

When the cost of delivered power is $1,000/MWh instead of the normal

$35/MWh, most customers will save only $35 by consuming a megawatt less. They
do not respond to the delivered cost of power because they receive no credit for
their response. Their meters do not track their usage in real time. The problem is
not so much low demand elasticity as lack of pricing. In addition, because customers
do not face marginal-cost prices, most have developed little capacity to respond.
This problem can be addressed either indirectly through architecture or directly
as a structural design problem. The indirect approach is to add competitive retail
markets to the mix of wholesale markets in the hope that they will change the
wholesale market structure by installing more real-time meters and implementing
real-time rates. The evidence so far is not encouraging. One direct approach was
initiated in California in the spring of 2001 in response to dramatic market failures,
many of which were greatly exacerbated by this demand-side flaw.6
The direct approach requires installation of more real-time meters and implemen-
tation of real-time retail rates. The former is trivial. California extended real-time
metering of load from 8 to 13 GW in just a few months at a cost of $25M, the
amount of money one utility was losing per day at the height of the crisis. Real-time
rates would be simple if customers were not risk averse, but they are, and so it
requires care to implement an appropriately hedged pricing scheme. California has
adopted a complex administrative approach to hedging that requires documentation
of a customer’s past usage patterns and a series of continuing corrections whenever
plants and equipment are expanded or changed for reasons not related to energy
conservation.
To hedge a customer’s bill under real-time rates, the customer’s usage should
be divided into two parts, a “base” usage and deviations from the base. The base
can be defined in many ways, and this determines the complexity of the plan and
the properties of the hedge. By defining the base as the average of the customer
class, or in some other mechanical way, instead of relying on the past usage of each
customer, the plan can be greatly simplified. Also, the extent of hedging can be
varied. For instance the pricing plan can guarantee that customer bills will be
immune to monthly changes in the average real-time wholesale prices. Although
this prevents customers from reacting immediately to changes in the long-run price
level, they still feel the full effect of hourly price changes (scaled to remove the
change in the average monthly price).7
The essential point is that effective demand elasticity is a crucial structural
parameter that is easily affected by policy, and there are many opportunities for
clever market design. Typical values of the parameter are so unfavorable to market
structure and improvement is so manageable that no power market should be started
without implementing a major real-time rate program.
6. This demand-side flaw was largely responsible for California’s market power problems and blackouts.
This fact is often ignored because the flaw, being structural, is considered akin to a law of nature and thus
automatically exempted from blame.
7. This result is based on the following real-time rates. Let Q(t) be the “base” profile that would be
charged the flat monthly rate. A real-time customer that actually uses R(t) is charged at the flat rate times
[the real-time cost of R(t) divided by the real-time cost of Q(t)].

1-7.4 LONG-TERM CONTRACTS

Long-term contracts and regulatory obligations to serve load are an often overlooked
aspect of market structure. These can greatly increase competitiveness in the spot
market. A generator that has sold 90% of its power forward has only 1'10 the
incentive to raise the price in the spot market as an identical generator that has sold
nothing forward.
If a utility sells half of its generating plants, it will become short of generation,
need to buy power, and have an incentive to hold down the market price. But,
without long-term contracts, the nonutility generators (NUGs) that bought these
plants will have a strong incentive to raise the price. Because the NUGs have more
ability to raise the price than the utility has to hold it down, the net result can be
a significant increase in market power. This happened in the California market.8
To avoid this problem, it is both customary and advisable when plants are
divested to require that the purchasers sign vesting contracts to sell most of the
output back to the utility for an extended period of time.9 This hedges the utility
(preventing a California-style bankruptcy) and dramatically reduces the market
power of the suppliers who bought the divested generation. The price of the long-
term contract can be indexed to fuel costs and inflation but should not be indexed
to the spot market wholesale price. Such vesting contracts were used extensively
in Australia and have served to keep wholesale prices low under almost all market
conditions.
Vesting contracts can dramatically improve market structure at the time of
divestiture, but in the long run, the permanent market structure and design will
determine the equilibrium level of long-term contracts. This is a topic of great
importance that is in need of research.
1-7.5 SUPPLY CONCENTRATION

Several indexes have been invented to measure supplier concentration; the best
known is the Herfindahl-Hirschman Index (HHI). (See Chapters 4-3 and 4-5.)
These indexes are of little use for predicting market power because concentration
is only one of several important determinants, but this problem with HHI does not
reduce the importance of concentration.
The two most effective methods of controlling concentration are divestiture
restrictions and limitations on mergers. Utilities are frequently required to limit
the amount of capacity sold to any single investor. For example, purchasers might
be forbidden to own more than 5% of the total capacity inside a market’s territory.
In small markets, the purchasers of particularly large plants might need an exemp-
tion from this requirement simply because a single plant exceeds the 5% limit.
8. See Borenstein, Bushnell, and Wolak (2000), the California Independent System Operator (2000),
Joskow and Kahn (2001a and 2001b), and Wolak, Nordhaus and Shapiro (2000).
9. The vesting contract should not specify which plants are providing the power, only that the new owner
of the plant will provide the power. This has the desired effect on market power without unnecessarily
restricting plant operation. See Wolak (2000) for the effectiveness of vesting contracts.

Note that the stricter the limit, the less valuable the plants will be. Exercising
market power is more difficult for those with a smaller market share. Because a
lower limit on market share reduces the profitability of generators, it reduces the
price they will fetch when sold. Consequently, the utility required to divest will
oppose any reduction in such a limit. It will argue that economies of scale or scope
require a purchaser to own a large amount of capacity in order to operate a plant
efficiently. These can be important effects, particularly when several plants are
located on a single site, but disinterested expert opinions should be sought on the
extent of such economies. Many suppliers own very little capacity in a market. Most
economies from multiple plant ownership can be achieved by owning plants in
different markets.10
10. Alberta, a relatively small market, has attempted to avoid the conflict between small concentration
ratios and efficiency by selling twenty-year contracts for a plant’s control and output, while leaving the
plant’s operation and ownership in the hands of the utilities. It remains to be seen how well this has
worked.

Chapter 1-8
Market Architecture
The whole world may be looked upon as a vast general market made up of diverse special markets
where social wealth is bought and sold. Our task then is to discover the laws to which these
purchases and sales tend to conform automatically. To this end, we shall suppose that the market
is perfectly competitive, just as in pure mechanics we suppose, to start with, that machines are
perfectly frictionless.
Leon Walras
Elements of Pure Economics
1874
A MARKET ’ S ARCHITECTURE IS A MAP OF ITS COMPONENT

“SUBMARKETS.” This map includes the type of each market and the linkages
between them.1 The submarkets of a power market include the wholesale spot
market, wholesale forward markets, and markets for ancillary services. “Market
type” classifies markets as, for example, bilateral, private exchange, or pool.
Linkages between submarkets may be implicit price relationships caused by
arbitrage or explicit rules linking rights purchased in one market to activity in
another.
Architecture should be specified before rules are written, but it is often necessary
to test the architecture during the design process, and this requires a rough specifica-
tion of the rules. Architectural design must also consider the market structure in
which it is embedded, which may inhibit the proper function of some designs.
Market design should not be rigidly compartmentalized, yet it is useful to consider
the market’s architecture apart from the details of the rules and the limitations of
market structure.
Chapter Summary 1-8 : A market design or analysis project concerns a collec-
tion of “submarkets” which are collectively referred to as the “entire market.” (Both
will often be referred to simply as markets.) Deciding which submarkets should
be created for a power market is the first step in architectural design. Section 1-8.1
briefly discusses day-ahead and real-time energy markets and transmission-rights
markets as a prelude to Part 3 which examines these choices in more depth.
Private submarkets range from disorganized to highly centralized, and each has
its advantages. There is no simple rule for choosing between types of submarkets
1. This chapter owes a great debt to Wilson (1999) and Chao and Wilson (1999a), though it is not
intended as a summary of their views.

CHAPTER 1-8 Market Architecture 83
once one has been included in the design list. By bringing the entire market into
focus and specifying the market’s architecture, the designer can take account of
linkages between markets and decide on the types of submarkets. The architectural
approach to design helps with understanding interactions and relationships between
submarkets and thus with avoiding the proverbial “Chinese menu” approach.
Linkages are the heart of market architecture, but the science of linkages is not
well developed. Timing, location, and arbitrage are the keys to most naturally arising
(implicit) linkages, while explicit linkages are limited only by the imagination of
the designer. Unfortunately many of these have unpleasant side effects. A careful
examination of the market’s architecture is the best antidote to inappropriately
designed linkages.
Section 1: Listing the Submarkets. The architecture of the entire market
includes the list of both designed and naturally occurring submarkets. Many
controversies surround the questions of which particular submarkets to include.
For example, a day-ahead centralized energy market may or may not be included,
and some suggest that the only designed markets to include are those for transmis-
sion rights and ancillary services.
Besides the designed submarkets, others already exist or will arise naturally.
If these play an important role in the functioning of the entire market, they are part
of its architecture.
Section 2: Market Types: Bilateral Through Pools. Bilateral markets can
be, in order of increasing centralization, search, bulletin-board, or brokered markets,
while mediated markets can be dealer markets, exchanges, or pools. Public central-
ized markets tend to have certain advantages over private markets that are decentral-
ized: lower transaction costs, quicker transactions, greater transparency of price,
and easier monitoring. Decentralized private markets are more flexible while
centralized private markets are similar to their public counterparts.
Energy exchanges accept only bids that, according to their bid-in values, at least
break even. Pools accept some apparently losing bids. Accepted bids that would
otherwise lose money are compensated with a “make-whole” side payment. Ex-
changes do not make side payments. Pools typically use much more complex bids,
though exchanges can also use multipart bids.
Section 3: Market Linkages. Implicit linkages are most often produced by
arbitrage, the most important example being the arbitrage-induced equality between
a forward price for delivery at time T and the expected spot price at that time. The
selection and arrangement of submarkets can take advantage of implicit links as
when ancillary-service markets are sequenced so that excess supply in one spills

into the next. Alternatively, multiproduct markets can be designed with explicit
linkages, sometimes increasing efficiency but also adding complexity.
1-8.1 LISTING THE SUBMARKETS

The term market is used in many ways and has no strict definition. Tirole (1997)
explains that it should not be so narrow as to encompass only a specific product
produced at a specific location, nor should it be “the entire economy.” He concludes
that there is no simple recipe. In practice, the market designer must choose a
definition to suit the problem at hand by relying on common sense rather than
theory.
At least two categories of market are needed. First the designer must define
the scope of the design problem which will be termed the entire market. “Power
market” in the present book refers to such a market. Depending on context this could
include only the wholesale market or the retail market as well. This choice demon-
strates the need for a second market concept. An entire market typically includes
components that are themselves markets. These will all be called submarkets. The
distinction between an entire market and a submarket is relative, not absolute. In
a different context it could be useful to view the ancillary services market as an
entire market with submarkets instead of as a submarket of the entire power market.
An entire market may comprise many or few submarkets depending on its degree
of vertical integration or unbundling. In other words, the intermediate products
used to produce the final products of the entire market can be produced internally
by the producer that uses them or can be purchased in a submarket. The first step
in mapping an entire market’s architecture is to decide on the list of submarkets.
This is a key step in the design, requiring careful consideration, and it can be highly
contentious. Currently, there is no consensus, even within the most informed circles,
as to the best collection of submarkets from which to construct a power market.
Just as for a market, there is no simple recipe for the definition of a submarket.
If a day-ahead energy market contains two zones, then different suppliers will sell
into each zone and there will be two prices. Clearly there are two products: energy
delivered to zone 1 and energy delivered to zone 2. Should these be considered
different submarkets? That may prove convenient, but when there are 500 locations
with different prices it will be necessary to count it as a single multiproduct
submarket.
Conversely, PJM’s day-ahead (DA) market sells energy and transmission rights.
Although the prices of these are strongly linked, they are such different products
they should be considered as supplied by two distinct but closely linked markets.
Whether the transmission-rights market belongs on the list is distinct from the
question of the energy market.
For a given project, some submarkets need to be designed and some do not.
The energy futures market at Palo Verde was not designed as part of the California
restructuring process though it is an important part of that power market. Generally,
private markets will not be part of a design project of the type contemplated, though,
when centralized, these markets are privately designed. Submarkets that are de-

signed as part of the public design process contemplated here are by definition
public markets and are almost always centralized. Submarkets that play a significant
role in an entire market should be included in the list of markets even if not designed
as part of the current project because they are a vital part of the market’s architec-
ture. Linkages between designed and naturally occurring or pre-existing submarkets
are crucial to the health of the entire market.
A Public Day-Ahead Energy Submarket

The question is not whether a DA energy market belongs on the list. If the system
operator does not provide one, a private market will develop. But from the market
designer’s perspective, the question of whether the system operator will run a DA
energy market is crucial. For this reason the public and private DA energy
submarkets should be considered distinct. If a public DA market is added to the
list, there will still be the question, discussed in Section 1-8.2, of what type of
market it should be.
If a centralized DA energy market is left off the list, a decentralized market must
certainly be included because one will develop and play an important role in the
entire market. Even with a centralized DA market, a private bilateral market is likely
to develop and should be included.
Energy vs. Transmission Submarkets

Perhaps the most fundamental controversy concerning which submarkets belong
in a power market concerns the question of whether the system operator should
operate an energy market or a transmission-rights market. One view holds that the
system operator is only needed to operate the grid and sell rights to its use, but it
should minimize its role in the market and refrain from trading or pricing energy.
A centralized transmission-rights market belongs on the list, but a centralized energy
market does not. The extreme version of this view would eliminate not only the
centralized DA energy market just discussed but also a centralized real-time (RT)
energy market.
The opposing view holds that such an architecture, plausible in a simplified
theoretical world, is wholly impractical. At least in real time, the system operator
needs to buy and sell energy directly and needs to set different prices for energy
provided at different locations. A real-time locational energy market should be on
the list. The extreme version of this view holds that public DA and RT transmission-
rights markets are not needed.
Reasons for Including a Submarket

An entire market typically consists of a set of closely related end-product markets
and the intermediate-product markets that feed into them. For a power market, the
end product markets might be only a single wholesale electricity market in a
particular region. Wholesale markets include all forward, futures, and options
markets. The difficult task is to decide which public markets should be created.

While there are no clear-cut rules, several possible motivations should be consid-
ered:
1. Nondiscriminatory access.
2. Completeness (trading of a product not otherwise traded).
3. Reduced trading costs.
4. A publically known price.
5. Transparent operation.
Nondiscriminatory access may be guaranteed by the governance structure of
a public market and is usually of most importance to small consumers and produc-
ers.2 Completeness is most relevant in the case of public goods such as reliability
services. It may also apply to natural-monopoly goods and services such as the unit-
commitment service. A public centralized market, such as an exchange, will
typically have much lower trading costs than a private decentralized market but
not necessarily lower than a private exchange.
A publically known price serves two different purposes. First, it is a required
assumption of the Efficient-Competition Result and is quite helpful to traders in
making efficient trades.3 Second it can be used as a benchmark for other transac-
tions, both regulated and private, such as settling financial futures. Transparent
operation is essential when market power is a potential problem and needs to be
monitored.
There are also drawbacks to public markets, all of which seem related to the
lack of proper incentives for regulators. A public market may offer products that
are not well designed or are too limited, or transactions costs may be higher than
necessary. These possibilities argue both for higher quality public institutions and
for the substitution of private ones.
1-8.2 MARKET TYPES: BILATERAL THROUGH POOLS

There are two basic ways to arrange trades between buyers and sellers. They can
trade directly, one buyer and one seller making a “bilateral” trade, or suppliers can
sell their product to an intermediary who sells it to end-use customers. Both bilateral
and mediated markets come in several types with bilateral markets usually less
organized but with some overlap in this regard.
Entire markets often use a mixture of types. For example, the used car market
is a mixture of direct search, bulletin board, and dealer markets. The New York
2. Motivations 1, 3, 4, and 5 are discussed in Chao and Wilson (1999a, 29).

3. See the statement of the first welfare theorem in Mas-Collel et al. (1995, 308).

Stock Exchange (NYSE) uses an auction, although when the market is thin, it
becomes a dealer market. The term “pool” has a special meaning with regard to
power markets. For years, utilities in some regions have organized their production
in power pools, some of which used a centralized dispatch. In a deregulated market,
a pool is an exchange in which the supply bids are complex, and the system operator
carries out a complex calculation to select and pay the winners.
Some markets work better as one type and some as another. In the heat of debate,
those favoring bilateral markets often imply that exchanges are in some way like
central planning, socialism, or even communism, but these analogies contribute
little. Often the right answer is for an entire market to utilize both approaches side-
by-side. The long-term energy market utilizes a bilateral forward market that trades
individualized forward contracts and centralized futures exchanges that trade
standardized futures contracts. The transaction cost of trading in the forward market
is greater but provides flexibility while trading in the futures market provides no
flexibility in contract form but is inexpensive.
Bilateral markets can be either direct search markets or brokered markets and
can be more or less centralized. If the market is brokered, as is the housing market,
the brokers do not actually buy or sell in the market but are paid a commission for
arranging a trade. Some forward energy markets, thanks to the Internet, are now
organized as bulletin-board markets which are just a partially centralized variety
of a direct-search market.
With the exception of the bulletin-board approach, bilateral markets need little
design. They require an enforcement mechanism for complex contracts, but this
is provided by the pre-existing legal framework.
Bilateral Markets, Dealers, Exchanges, and Auctions

In bilateral markets buyers and sellers trade directly, although this is typically
facilitated by a broker. Such markets are extremely flexible as the trading parties
can specify any contract terms they desire, but this flexibility comes at a price.
Negotiating and writing contracts is expensive. Assessing the credit worthiness
of one’s counter party is also expensive and risky. For these reasons there is a great
advantage to moving toward more standardized and centralized trading when this
is made possible by the volume of trade.
A dealer market is the most rudimentary type of mediated market.4 Unlike a
broker, a dealer trades for his own account, and usually maintains an inventory.
He buys the product and holds it before reselling. There is no brokerage fee, but
at any point in time the dealer buys for a price that is lower than the price he sells
for. This difference is called the spread.
An exchange provides security for traders by acting as the counter party to all
trades, eliminating traders’ concerns over creditworthiness. Exchanges utilize
auctions and are sometimes called auction markets; the NYSE is an example. As
Bodie et al. (1996, 24) point out, “An advantage of auction markets over dealer
markets is that one need not search to find the best price for a good.” Because an
4. The terms “direct search market,” “brokered market,” “dealer market,” and “auction market” are
defined in Bodie et al. (1996, 24).

exchange interposes itself between buyers and sellers, the two halves of the market
can operate independently, although they are linked by what is called a double
auction. (Auctions are a traditional method of implementing a competitive market
and are discussed in Chapter 1-9.)
An exchange can have a number of advantages over a bilateral market. It can
reduce trading costs, increase competition, and produce a publically observable
price. Depending on design and circumstances, it can also facilitate collusion and
generally provides less flexibility than a bilateral market. Power marketers often
favor bilateral markets because without an exchange there is more room to earn
commissions as brokers and to appropriate the spread when they act as dealers.
Because exchanges are inflexible, they can operate much faster than bilateral
markets. Stock exchanges routinely execute trades in under five minutes while
bilateral markets take hours to weeks. In power markets, speed is crucial. Catastro-
phes can happen in seconds and system operators often need to exercise minute-by-
minute control. Because of their speed, exchanges can operate much nearer to real
time than can bilateral markets. This makes them the obvious choice for the real-
time market. Weeks in advance, bilateral markets and dealer markets may play a
larger role than exchanges. In between, there is much room for disagreement over
which is better.
Exchanges vs. Pools

If there is to be a public day-ahead market, should it be a pool or an exchange?
PJM, NYISO, and ISO-NE all adopted pools, but CA ISO adopted a combination
of public exchange (the Power Exchange) and private exchanges and dealers (the
other “scheduling coordinators”). Although the California market has performed
disastrously, this probably has little to do with its architecture and everything to
do with its structure. No conclusion regarding exchanges can be drawn from this
evidence.
California’s Power Exchange was an exchange because it did not use the “make-
whole” side-payments which characterize a pool. As is typical of exchanges, it used
simple bids. These expressed only an energy quantity and price which meant
generators could not take account of their startup costs and no-load costs directly
within the bid format. Consequently, they had to manipulate or “game” their bids
in some way to avoid a loss. Chao and Wilson (1999, 48) discuss such a circum-
stance as follows:
Gaming strategies are inherent in any design that requires trad-
ers to manipulate their bids in order to take account of factors
that the bid format does not allow them to express directly.
In order to avoid this problem, more complex bids, perhaps two or three part bids,
are needed. Alternatively a pool could be used.
Pools are defined by the existence of side payments. Generators bid their
marginal cost and certain other costs and limitation into the pool which computes
a price and a set of accepted bids. Some accepted bids are found to lose money
because the pool price is not enough higher than their marginal cost to cover their

other bid-in costs. The pool makes up for this be granting accepted bidders, that
would otherwise lose money, a side payment that makes them whole. Because some
apparently losing bids are accepted, losing bidders have no way to verify whether
their bid was correctly rejected. Typically pools utilize very complex bids which
attempt to comprise a complete economic description of the generation process,
but this is not necessary. A pool could be designed with two-part bids.
Although the principle of designing bid formats so that they do not require
“manipulation” is a useful one, it need not be carried to extremes. Another principle
states that, in a competitive market, competitive forces will induce bidders to
represent true costs as accurately as possible within the bid framework. Thus in
the Power Exchange, bidders included in their bids expected startup and no-load
costs as well as their marginal costs. Chapter 3-9 shows that even in an exchange
with one-part bids, where considerable manipulation is necessary, a competitive
market will do a remarkably efficient job of dispatching generation. The “gaming”
exhibited in a competitive market with an inappropriate bid format is largely
beneficial and should probably be described by another name.
1-8.3 MARKET LINKAGES

Market linkages, aside from arbitrage, have no standard classification or nomencla-
ture. Nonetheless they are tremendously important to the functioning of the entire
market. Linkages can be either explicit or implicit. For instance, the requirement
to purchase a transmission right, in order to inject and withdraw power, is an explicit
linkage between the market for transmission rights and the bilateral market for
wholesale energy. An implicit linkage causes the price in a forward energy market
to approximate the expected price in the spot market during the forward’s delivery
period. There is no rule that enforces this relationship, only the discipline of
arbitrage. Implicit links are not designed, but they are an important part of the
architecture and must be reckoned with.
Sometimes when explicit linkages are needed, it indicates that two markets
should be merged into a multiproduct market. Implicit linkages occur naturally and
are usually helpful; explicit linkages are helpful when they reflect real costs. They
are frequently harmful when they reflect a preconceived notion of how the market
should operate.
Because power markets are geographically distributed, many of their submarkets
are multiproduct markets and contain vast arrays of internal linkages. When the
transmission system is congested (or if losses are charged for, as they should be)
energy at location A is technically a different product from energy at location B.
This is not just an academic definition. A completely unregulated bilateral market
will price energy differently at the two locations. Consequently, an energy market
is a multiproduct market with internal linkages between the products.
Just as there are spacial linkages, there are also temporal linkages. Market
architecture establishes the temporal order of markets, and this order causes implicit
linkages to develop between the markets. Both temporal and spacial linkages
pervade the architecture of power markets.

The Arbitrage Linkage of Forward to Spot Prices

In a well-arbitraged market the forward price for delivery at time T will equal the
expected spot price at time T. Because the market designer will find this common
linkage extremely useful in understanding and predicting the behavior of entire
markets, it is explained in some detail here.
There are two types of arbitrage (Tirole 1997, 134). The first, and more
commonly recognized type, involves the transfer of a commodity from a high-priced
location to a low-priced location. The relevant type for this analysis involves the
transfer of demand from a high-priced product to a low-priced product. In the
present case, customers can transfer their demand between forward purchases and
spot purchases.
Say a customer knows it will need a certain quantity of power at a future time
T. The customer can either buy an energy future now, at known price FT, to be
delivered at time T, or can wait and buy the power in the spot market at an as yet
unknown price PT. Which is preferable? Payment on the future can be largely
postponed until delivery, so interest has a negligible effect. Generally a known price
is preferred to an unknown price, but the main effect is that FT will be preferred
if PT is expected to be higher and vice versa. If E(PT) is the currently expected spot
price at time T, this can be summarized as
FT < E(PT) causes futures to be preferred.
FT > E(PT) causes spot purchases to be preferred.
The result is that whenever FT is lower, demand for futures will increase and FT
will increase, while, if FT is higher, demand for futures will decrease, with the result
that FT = E(PT).
Result 1-8.1 The Forward Price Is the Expected Future Spot Price
The price of a future or forward specified for delivery at time T is approximately
equal to the expected spot price at time T.
This relationship is not exact. Both buyers and sellers tend to prefer the certainty
of knowing the futures price. If the buyer’s preference is stronger, FT may be greater
than E(PT), and vice versa. These preferences will depend somewhat on the possibil-
ities for diversifying the risks involved, but altogether these effects are too subtle
and too unpredictable to be of interest to the power-market designer. For practical
design purposes, FT can be expected to equal E(PT).
One immediate consequence is that if the real-time (spot) price is capped at
$500, then the day-ahead price will not rise above $500. Of course if there are
penalties for trading in the spot market, these must be taken into account. This effect
was well documented in California, and it provides a mechanism for capping
markets that are not accessible to the system operator.
Besides being a valuable tool for understanding the implications of system
architecture, this result can provide practical insight. Seeing current high spot market
prices and low future prices, i.e., FT < P0, many concluded that it was nearly always

much cheaper to buy forwards than to wait for the inevitable high spot price. In
contradiction of Result 1-8.1, they concluded that FT < E(PT), and probably much
less. This was one of two factors which led to California’s huge purchases of
forwards in the spring of 2001. It now appears that purchasing all of California’s
power requirements for 2004 from an extremely thin market during a power
shortage, was not a sure bet. Pre-announcing a determination to buy no matter what
and using inexperienced traders compounded the problem. Most likely, California
did manage to create a temporary exception to the forward arbitrage linkage,
FT = E(PT), but in the opposite direction from what motivated the purchases: FT
was probably much greater in California during the spring of 2001 than a rational
expectation of future spot prices.
Locational Price Linkages

Consider what happens if transmission rights are auctioned independently, as they
often are. Suppose the right from A to B and from B to C and from A to C are all
bid for separately, and the system operator sells them at the lowest set of prices
that clears the market, given feasibility constraints on the rights. This would give
three prices, PAB, PBC, and PCA, but if the market is working efficiently, these prices
will be tightly linked and in particular will sum to zero. Arbitrage is also key to
this result, but the point is not the result itself but the fact that spacial linkages of
energy prices are strong and important.5 This is just one of a number of such results.
Cascading Markets
Generators can provide energy or they can provide reserves, an option to buy energy
when more is needed. Various qualities of reserves are graded by the quickness
and sureness of their response. This classification provides an unambiguous
ordering by value with the best quality of reserve always preferred to the second
best, and so on. Suppose there are three such products called R1 (best), R2, and
R3, and the system operator requires a certain amount of each.
If no linkage is made between the markets for R1, R2, and R3, they will take
place simultaneously but separately. In this case any excess capacity in one market
cannot flow into the others to help lower the price in those markets, so the markets
should be linked, at least by conducting them in sequence. If the R3 market were
conducted first, any excess of R2 might not be bid into the R3 market because it
would not yet be known that it was in excess. To achieve a cascade, the markets
should be cleared starting with the highest quality. Any surplus R1 reserves could
bid into the R2 market and so on. This increases the efficiency of the markets
relative to the absence of a cascade or one set up in reverse. Even with a forward
cascade these markets will not perform optimally. If there is a shortage in the market
for R3 and in total, but not for the first two markets, the price in the R3 market could
exceed the other two prices. In this case high-quality reserve units would hold out
for a chance at the third market, which could result in an inefficient use of reserves.
5. This is explained in 3-1.3 and can be proven from that facts that in general PAB = !PBA and
PAB + PBC = PAC .

Efficiency can be increased by collapsing the three into a single multiproduct

market with strong internal linkages. Such an arrangement will be simpler for the
suppliers but more complex for the system operator. The reserve market can also
be integrated with the energy market. Every new level of integration brings new
complexity to the market clearing mechanism and reduces its transparency, but
it has the potential to increase efficiency. Making the proper trade-off between
efficiency and internal complexity, with its attendant opportunities for design error
and gaming, is a controversial and unresolved issue.

Chapter 1-9
Designing and Testing Market Rules
Genius is one per cent inspiration, ninety-nine per cent perspiration.
Thomas Edison
c. 1903
If Edison had a needle to find in a haystack he would proceed . . . to examine straw after straw.
A little theory and calculation would have saved him ninety percent of his labor.
Nikola Tesla
New York Times
1931
U NTESTED MARKET DESIGNS CAUSE REAL-WORLD MARKET FAIL-

URES. Suppliers are quick to take advantage of design flaws, especially those that
pay $9,999/MWh for a product that is worth less than $5/MWh.1 Currently, many
if not most, market designs are implemented without any explicit testing.
Although the most serious market flaws typically arise from structural problems,
while architectural problems rank second in importance, problems with rules are
the most numerous and their cost can be impressive. The design of rules is more
art than science, but economics offers two guiding principles: mimic the outcome
of a classically competitive market, and design markets so competitors find it
profitable to bid honestly. Simplicity is another virtue well worth pursuing but
notoriously difficult to define.
Chapter Summary 1-9 : In a pay-as-bid auction, a coal plant bidding its variable
cost of $12/MWh would be paid $12/MWh, while in a single-price auction it would
be paid the system marginal cost which might be $100/MWh. In this case many
would object to paying the $100 competitive price to the “inexpensive” coal plant
and seek to improve on the competitive model. Pay-as-bid is one suggestion. The
result is gaming and, probably, a very modest decrease in price and a modest
decrease in efficiency. Ironically, if pay-as-bid succeeded as its advocates hope,
it would put an end to investment in baseload and midload plants. In the long run
this would dramatically raise the cost of power. The pay-as-bid fallacy illustrates
the topics of the first three sections: the danger in attempting to subvert competition,
the benefits of “incentive compatible” design, and the relevance of auction theory.
1. One of several design flaws that produced this outcome was prohibiting the California ISO from
substituting a cheaper better product for a more expensive poorer product (Wolak, 1999). Also see Brien
(1999).

That testing is the key to successful design is well understood by engineers until
they design markets instead of equipment. It is not well understood by policy makers
or economists, and the results are predictable. Rigorous testing, though worthwhile,
is expensive so a simple “bottom-line” test should always be conducted first. This
only requires building the simplest relevant model computing the cheapest possible
production costs, and then computing the costs under the proposed design. If these
are much different, reject the design. This test cannot prove a design will work,
but it can save the cost of a rigorous test or a real-world failure.
Section 1: Design for Competitive Prices. Competitive prices sometimes
include a scarcity rent much greater than needed to cover the concurrent fixed cost
payment. Frequently this inspires attempts, such as FERC’s advocacy of pay-as-bid
auctions, to redesign the market to pay a price below the competitive level. If such
a scheme were to succeed, it would cause reduced and distorted investment.
Fortunately most such schemes are largely subverted by market forces.
Section 2: Design to Prevent Gaming. Rules that induce truth telling are
called “incentive compatible” and often provide good market designs. Some bidding
rules force suppliers to submit bids that do not reflect their true costs; such rules
induce gaming. The pay-as-bid auction design is an example. Gaming usually causes
inefficiency, the importance of which needs to be evaluated on a case-by-case basis.
Section 3: Auctions. Exchanges and pools use auctions to determine market-
clearing prices. The four main auction types all produce the same revenue when
bidders are buying for their own use or selling their own product. There is a slight
efficiency advantage for a “second-price” auction which is incentive compatible.
Recent work in auction theory shows that these results do not generalize to the
multi-unit auctions with elastic and uncertain demand characteristic of power
markets. In this setting, results are ambiguous, but pay-as-bid auctions tend to inhibit
market power, sometimes at the cost of reducing welfare.
Section 4: Testing Market Rules. Every market design should undergo at
least minimal testing before use. A minimal “bottom-line” test consists of three
steps: (1) model the market with and without the design in enough detail to compute
the design’s impact on production costs, (2) find the minimum possible cost of
delivered power, and (3) find the cost of delivered power when the market operates
under the proposed rules. If the design raises costs significantly, it fails the test.
Such a test cannot prove that the design will work well in the real world, but it often
shows it will fail under even ideal conditions, a useful, if disappointing, result.
Section 5: Technical Supplement—Example of a “Bottom-Line” Test.
A proposed charge for transmission access is tested and found to induce generation
investment in a pattern that increases the total cost of delivered energy. Since the

CHAPTER 1-9 Designing and Testing Market Rules 95
proposed design fails to perform under even the simplest network conditions, it
could have been rejected without expensive testing.
1-9.1 DESIGN FOR COMPETITIVE PRICES

Markets should be designed to produce competitive prices, but especially in power
markets, competitive prices sometimes appear disconcertingly high. While most
designers remain loyal to the ideal of competitive prices, many decide to redefine
them to be lower at times, and some decide that they are just not right.
Recently FERC proposed a pay-as-bid auction design in the hopes of holding
prices below their competitive level.2 Several early proposals for the California
market also had this intent. Even the PJM market contains elements of this flaw.
These initiatives reflect a basic misunderstanding of the role of scarcity rent
(defined in Section 1-6.6).
Scarcity Rent
If a competitive generator sells 10,000 MWh for $500,000 and its variable operating
costs are $100,000, then its scarcity rent is $400,000 which is not an unlikely
outcome for a low-variable-cost coal plant when the market price is set by the
marginal cost of a high cost gas turbine. Scarcity rents are necessary for suppliers
to cover their fixed costs, which, as explained in Chapter 1-3, can be thought of
as a constant flow of cost equivalent to the cost of renting the power plant. During
the period in which the plant produced its 10,000 MWh, its fixed costs might have
been $200,000. In this case the scarcity rent was twice what was needed to cover
fixed costs during that period of operation. Such discrepancies are common, and
they convince many that competitive prices are often too high.
Fallacy 1-9.1 Scarcity Rents Are Unfair

If the price paid to generators always equals system marginal cost, generators with
lower variable costs will be paid too much.
This conclusion ignores the fact that in power markets, scarcity rents fluctuate
dramatically. If they are to equal fixed costs on average, they must be higher
sometimes and lower other times. Typically they are lower most of the time but
are occasionally much higher. Ignoring this leads to a fallacy which states that
paying a low-variable-cost plant the marginal cost of an expensive plant is unfair.
This was succinctly expressed by former FERC Chairman Curt Hébert.3
2. The Blue Ribbon Panel that examined this scheme for California’s Power Exchange concluded: “In
sum, our response is that the expectation behind the proposal to shift from uniform to as-bid pricing—that
it would provide purchasers of electric power substantial relief from the soaring prices of the electric
power, such as they have recently experienced—is simply mistaken. . . . In our view it would do
consumers more harm than good." (Kahn et al., 2001, 17)
3. From p. 4 of his concurrence with FERC (2000b).

If the market clearing price for the final increment of needed

capacity is, say, $100 MWh, why should a supplier who bid a
lower figure receive the same value as that afforded to the sup-
plier of [the] higher-priced increment?
There are two answers to this question. First, as explained below, if low bidders
were paid less, they would raise their bids. The Blue Ribbon Panel (Kahn, 2001),
which analyzed FERC’s pay-as-bid proposal for the California Power Exchange,
focused entirely on this first answer. It explained what would have gone wrong
had the proposal been implemented and it used auction theory for its analysis.
The second answer and the root of the problem lies in the desire to hold prices
down to their short-run limit. Notice that the concern is with a “supplier who bid
a lower figure.” Suppliers that bid low are lower-variable-cost, baseload suppliers,
and they bid low in order to guarantee they will run. The chairman’s question was
why, with their low variable cost, they should be paid as much as the high-priced
supplier. The answer is they have higher fixed costs and need more scarcity rent
to cover them.
Many schemes have been proposed to hold prices down to variable cost, and
given sufficient regulatory authority they can be effective. Consequently it is
important to understand that reducing price in the short run will increase it in the
long run. This is not true if prices remain at or above the competitive level, but
the competitive price pays higher and lower cost producers exactly the same when
they both are needed at the same time. FERC’s scheme was intended to hold
baseload prices well below the competitive level. The Efficient-Competition Result
states that competitive price will, over the long run, induce the right investment
in both baseload plants and peakers. If regulators reduce the prices paid to baseload
plants, in the short run they can get away with it, but in the long run, investors will
see that baseload plants cannot cover their fixed costs and will build no more.
How the Market Fights Back

Markets have ways of subverting the best-laid plans of regulators. Sometimes this
causes problems, but in this case, it would prevent most of the damage that FERC’s
scheme would cause if it worked as intended. The plan was to capture the rent of
baseload plants and thereby reduce prices. The method was to pay low bidders their
bid, and as Kahn et al. (2001, 5) explain:
The critical assumption is, of course, that after the market rules
are changed, generators will bid just as they had before. The one
absolute certainty, however, is that they will not.
Markets are composed of clever, highly motivated players who spend a great deal
of time and effort discovering the most profitable way to respond to changes in
rules. The most fundamental mistake a market designer can make is to treat a market
as if it were a machine that does not change behavior when the rules change.
Knowing that bids will change is easy; finding the new outcome of changed
rules and changed behavior is more difficult. One approach is to take a close look

at why the New York Stock Exchange does not follow Chairman Hébert’s prefer-
ence “that sellers ... be paid what they bid ... rather than the market-clearing price.”4
Why the NYSE Pays the Market-Clearing Price

Each stock on the NYSE accumulates bids and offers over night, and as the market
opens, these are traded at a single market-clearing price. Say there are three bids
to sell 100 shares each at $50, $60, and $70 per share and three bids to buy at $50,
$60, and $70 per share. In this case, the Exchange will set a market-clearing price
of $60, and the two offers to sell for this much or less will be traded with the two
bids to buy at this much or more. All will receive or pay the market-clearing price
of $60.
If the NYSE followed FERC’s scheme for running the California market, the
opening of the stock market would work like this: The NYSE would accept the
two bids from those selling stock (the supply or generator bids) at $50 and $60 and
pay these two their bid price. They would then compute the average cost to be
$55/share and sell the shares at that price to the buyers who bid $60 and $70/share.
Instead of a single market-clearing price, the market would have one price for each
supplier and a different but single price for all buyers.5
The suppliers would then argue that the NYSE should have done just the
opposite. They should have averaged the two buyer’s bids of $60 and $70 and paid
both suppliers $65 for their stock. They have a point. Why should sellers who bid
low receive only their bid price while buyers who bid high are not required to pay
their bid price. Are stock buyers more worthy than stock sellers?
There is a more fundamental problem; traders will not sit idly by. The $50
supplier would realize that a $60 bid would have been more profitable and the next
day bid $60. Then the FERC’s scheme would produce the same result as the
NYSE’s market-clearing auction. If bidding low is rewarded with a low payment,
while bidding the market clearing price is rewarded with the higher market-clearing
price, previously low bidders will become high bidders.
Result 1-9.1 Changing the Market’s Rules Changes Behavior
Auction Rules and Competitive Prices

Predicting the effect of rules and rule changes is a difficult matter which is why
Section 1-9.4 prescribes testing for all proposed rules. Many of the important rules
in power markets, including those just discussed, are auction rules and are the
subject of auction theory, an important and rapidly developing branch of economics.
A principle goal of auction theory is the design of auctions that produce competitive
prices, and while much progress has been made, much remains unknown.
4. “My preference is that sellers in California be paid what they bid, regardless of what that bid is, rather
than the market-clearing price.” From p. 4 of his concurrence with FERC (2000b).
5. Another possibility is to match buyers and sellers by price so the $50 supplier sells to the $50 buyer,
and so on. Applying the concept of total surplus from Section 1-5.2 demonstrates that this yields no total
surplus, while using a market-clearing price gives a total surplus of $20.

When considering auction rules, two categories of problems should be distin-

guished: designing for a market with a competitive structure or designing for a
market with a monopolistic structure. A competitive structure implies the bidders
are small relative to the size of the market, where “small” must be defined relative
to the market’s demand elasticity. If bidders are large or demand inelastic, the
structure is monopolistic. Within a competitive structure, it is reasonable to ask
that the auction rules produce competitive prices. In a monopolistic structure, about
the best that can be hoped for are auction rules that reduce the exercise of market
power relative to alternative rules.
When considering FERC’s proposed pay-as-bid rule change, three questions
should be asked:
1. If the rule change worked as intended, would market prices be nearer the
competitive level?
2. In a competitive market structure, would the rule change bring market prices
nearer to the competitive level?
3. In a monopolistic market structure, would the rule change bring market
prices nearer to the competitive level?
The first question is about intentions, the second and third are about how the rule
would actually work. The answers to these three questions are: (1) No, baseload
prices would be reduced far below the competitive level. (2) No, prices would suffer
a mild increase in randomness and baseload prices would be reduces slightly below
the short-run competitive level. (3) Quite possibly, but this might come with a
reduction in welfare. These results are discussed further in Section 1-9.3.
1-9.2 DESIGN TO PREVENT GAMING

Markets often require participants to state a price at which they would either buy
or sell. The market then selects the buyers who name high prices and the sellers
who name low prices. This occurs in private bilateral markets and in public ex-
changes. If the buyers who bid high have the highest values for the goods being
purchased, and if the sellers who bid low have the lowest costs of production, then
their trades will be efficient. All those who can produce for less than the market
price will trade with all those with values above the market price.
One way to arrange efficient trading is for all the traders to bid prices that are
equal to their true costs and values. If all bidders tell the truth, the outcome is
efficient. This is not the only way to get the efficient outcome, but it is the most
obvious. An economic theorem states that for a very broad class of markets, if any
market mechanism can be designed to give the efficient outcome, then there must
be one that works by inducing traders to tell the truth.
Economists call a market mechanism that induces truth-telling an incentive-
compatible mechanism. It not only makes the market efficient but also tends to
be easier to discover and simpler to implement. For these reasons economists look
for incentive-compatible designs, but sometimes they are too complex or politically

unpopular. In this case a less efficient design will usually need to be adopted, but
it is generally useful to look for the efficient design first.
Result 1-9.2 Design Market Mechanisms to Induce Truth Telling

If there is a market design that leads to an efficient outcome, then there is a design
that induces traders to tell the truth and produces that outcome. Such a design is
called “incentive compatible.”
The pay-as-bid auction discussed in the previous section is not incentive

compatible for it induces misrepresentation of variable cost. This is also called
“gaming.” As a consequence there will necessarily be some inefficiency unless
the market price is always perfectly predictable. The single-price auctions of most
electricity markets are nearly incentive compatible. Suppliers almost tell the truth
when they are competitive (most of the time) but distort their bids when they have
market power.
1-9.3 AUCTIONS
In the last forty years, economics has developed an extensive theory of auctions.
Perhaps because auctioning electricity is a rather new idea, some have suspected
that auctions are an invention of theoretical economics. In fact, the Babylonians
used them in 500 BC, and Buddhists employed them in the seventh century.
Sotheby’s was established in 1744, some time before Adam Smith gave economics
its start.6
William Vickrey, who won the 1996 “Nobel Prize” in economics largely for
his work on auctions, classified one-sided auctions into four types.
1. English: Buyers start bidding at a low price. The highest bidder wins and
pays the last price bid.
2. Vickrey (second-price): Buyers submit sealed bids, and the winner pays
the price of the highest losing bid. This is also confusingly called a Dutch
auction.
3. Dutch: The auctioneer starts very high and calls out progressively lower
prices. The first buyer to accept the price wins and pays that price.
4. Sealed-Bid (first-price): Buyers submit sealed bids, and the winner pays
the price that is bid.
These four types of auction can all be used in reverse to buy a product instead
of sell it. For example, in a reversed English auction, sellers can call out progres-
sively lower prices until there are no more bids. The lowest bid wins, and the seller
sells at that price. Consequently everything said about auctions to sell, holds for
auctions to buy, but in reverse.
There are two motives for buying: for a buyer’s own use or for resale. In the
first case the value of the purchased good is the private value placed on it by the
6. See Klemperer (1999; 2000a; 2000b) for surveys of auction theory and literature. The first is the most
accessible.

buyer. In the second case, the value is the price for which it can be resold. If the
purchase is for resale, then all potential purchasers will find the value to be the
same, i.e., the market price upon resale. This is called the common value.
The Revenue Equivalence Theorem

The revenue equivalence theorem, a key result of auction theory, states that
the four types of auction will all produce the same revenue if used in private-value
auctions. To help explain how this is possible, consider a comparison of a Vickrey
(second price) auction and a sealed-bid (first price) auction.
Result 1-9.3 Four Types of Auctions Produce the Same Revenue

The four types of auctions, English, Vickrey, Dutch, and sealed-bid, all produce
the same revenue if the bidders have private values. If they have common values,
then their revenues are in the listed order with English producing the most.
Assume the bids in each were $100, $200, and $300 by A, B, and C respectively.
In each auction, C wins, but in the Vickrey auction, C would pay only $200 while
in the first-price auction, C would pay $300. This makes it seem that the revenue
of the first-price auction is clearly greater by $100.
The explanation of revenue equivalence involves the fundamental principle
of rule design, Result 1-9.1. Changing the rules changes the behavior of the bidders.
To analyze these behaviors assume that the bidders have private values correspond-
ing to the above bids. First, consider bidder behavior under a Vickrey auction. C
will bid $300 because he knows that if A and B have bid less he will win and get
the best possible price, the price of the next-highest bidder. If C loses, say to a bid
of $305, he will be glad he lost, for winning would mean paying $305 for something
he values at only $300. If a bidder bids less than her valuation, she increases the
chance of losing when she should win and saves no money. If a bidder bids above
her private value her only extra wins from biding higher will be in auctions she
would rather lose. So all bidders bid their true values, and this design is incentive
compatible. Bidder C will win and pay the second price of $200.
Result 1-9.4 A Vickrey Auction Is Incentive Compatible

A second-price auction causes bidders to bid their true value, or if it is an auction
to purchase, they bid their true cost.
Now consider a first-price auction. Bidding strategies depend on what is known

about others’ bids. Assume that C knows that A and B almost surely have values
of $100 and $200. Then C knows B will almost surely not bid higher than $200
and perhaps lower. If C is sure enough he will bid $201 or less. So the revenue
of a first price auction will be much less than $300 because C will not bid his true
value. This auction is not incentive compatible. The calculation of bidding strategies
in such an auction is quite difficult, and it is remarkable that the revenue equiva-
lence theorem can cut through this complexity and produce such a clean result.
The winner in the first-price auction will pay $200 on average.

One disadvantage of the first-price auction is the complexity of the optimal

strategy. Bidders are forced to guess the probability distribution of other bids and
then make a complex computation based on this guess. This introduces errors into
the bids. For example, in a power market, bidders in a Vickrey auction will bid
their marginal cost and that will reproduce the standard merit order from which
the system operator will dispatch generation efficiently. With a first-price auction,
all generators with marginal costs below the expected market price attempt to bid
just under the market-clearing price. This produces a cluster of bids that have little
to do with the merit order and which differ more because of estimation errors than
because of costs. Dispatching the “cheapest” generators from this assortment of
bids will produce an inefficient dispatch.
Revenue Equivalence and the Pay-as-Bid Design

If the auction sells many units of the product, many megawatt-hours of power, and
there are many winners, then a second-price auction pays all winners the same while
a first-price, sealed-bid auction pays each winner his bid. This gives rise to the terms
single-price auction and pay-as-bid auction. Economists call a “pay-as-bid” auction
a discriminatory auction. As discussed in Section 1-9.1, FERC recently argued
for replacing California’s single-price auction with a discriminatory auction on
the grounds that it would result in lower prices. If the auction under consideration
were buying a single indivisible unit of power of a known quantity, then the revenue
equivalence theorem would prove that switching to pay-as-bid would have no affect
on revenue, which means the average price paid would be the same.
Unfortunately, electricity auctions are multi-unit procurement auctions in which
demand is elastic and uncertain, and the revenue equivalence theorem does not
apply. A paper by Frederico and Rahman (2001, 2) demonstrates that for such
auctions, in a competitive market structure, pay-as-bid reduces the price paid but
also reduces efficiency. They show that under monopoly conditions, “the exercise
of market power is more difficult under pay-as-bid, and that firms with market
power may react in inefficient ways to a switch to pay-as-bid” thus reducing
welfare. They are also optimistic that under oligopolistic conditions pay-as-bid will
significantly reduce market power but have not modeled this situation.
These results are complex and ambiguous, and those who believe they see easy
answers have not understood the question. Theory may never yield definitive
answers, and testing of auction designs may prove the only reliable method of
evaluation. Testing such a design for the first time in a 10 billion dollar per year
market may not be the most cost-effective approach.
1-9.4 TESTING A MARKET DESIGN

In a turbulent environment such as the restructuring of the U.S. electric industry,
new market designs are nearly always conceived and implemented without rigorous
testing. In fact a complete absence of testing is common; however, few designs
are implemented without claims that they produce “the correct incentives” and

“efficient outcomes.” Unfortunately, when such designs are tested against even
the most well-behaved hypothetical situations, they do not always live up to their
claims.
A moderately rigorous test would involve a laboratory simulation of the market
design. While this is recommended, especially when gambling with sums that can
run into the billions of dollars, a more modest course of testing is presented here.
No design should be implemented without this minimal level of testing, which is
called the “bottom-line test” because it tests the effect of a market design on the
total cost of supply. It does not provide a cookbook procedure, but it provides a
structure that is often missing, and it sets a minimum standard. The test procedure
can be broken into three steps.
Test The Bottom-Line Test of Market Rules

1. Model the cost functions of the players in the market.
2. Compute the minimum possible cost of serving the target load level.
3. Compute the cost increase of serving the load under proposed rules.
If consumer response to price (demand elasticity) is important for the rules being
tested, then the bottom-line test must be modified by examining the decrease in
total surplus caused by the design instead of the increase in cost. Fortunately, most
market rules can be tested adequately under the assumption of completely inelastic
demand. Then consumer benefit is unaffected by market rules and drops out of
the test procedure. Occasionally, adding demand elasticity can simplify a model
by removing “knife-edge” discontinuities in behavior. In this case the net-benefit
test should be used.
Lack of demand elasticity confers market power on suppliers, and most designs
are not efficient in the presence of market power. The bottom-line test is only
intended to test designs for competitive markets, so to counteract the effects of
inelastic demand, suppliers should simply be assumed to behave competitively.
The Least-Cost Focus of Economic Testing

As discussed in Section 1-5.2, least-cost production is the criterion for both short
and long-run supply-side efficiency. Statements about sending the right price signals
to generators, providing optimal incentives, and being efficient mean nothing if
the design does not keep the delivered cost of production low. Conversely, if the
design does minimize generation costs, it necessarily sends the right price signals
to generators and provides the right incentives.
The supply-side implications of a competitive market design can be tested just
by comparing the total cost of supply under the proposed design with the minimum
cost of supply in an ideal world. The point is not that the ideal must be achieved
before a design is approved, but the ideal provides a benchmark, and cost is the
sole criterion. As simple as this test is, it can still include such complexities as
environmental concerns by including their costs. (This is the only method of
inclusion that makes otherwise implicit trade-offs explicit.)

The Advantage of Simple Models

Economic modeling is as much art as science, and mistaken conclusions result as
often from poor model design as from faulty calculation. Simple models minimize
such errors. Those who prefer their ideas not be tested usually argue that simple
models cannot test a design adequately and should not be used. Even when a
complex model is available, simple models should be used to check the reasonable-
ness of the more complex models, and a simple model is always better than no
model.
Simple models can give clear-cut results when they are negative and can save
the considerable effort of a more elaborate test. One disadvantage, however, is that
while they can lend credibility to a design, they cannot serve as proof that it will
succeed in practice. That such testing can prove a design wrong but cannot prove
it right may be discouraging, but it is always better to learn of failure during the
design phase.
1-9.5 TECHNICAL SUPPLEMENT: EXAMPLE OF A “BOTTOM-LINE TEST”

This example demonstrates how to test a design with the bottom-line test and how
to avoid spending a lot of money on complex tests when a design has an elementary
flaw. Detailed cost calculations are postponed to the end in order to focus on
methodology. This example was an actual proposed design, and the inappropriate
test was carried out.
A Test of Two Transmission Charging Rules

The design to be tested consists of two rules that are intended to charge generators
for wires in such a way that new generators are induced to build in the optimal
location. The first rule specifies that the power flows caused by a generator are
calculated by assuming its power flows proportionally to every load on the system.
(This is not a law of physics but only an accounting rule.) If a generator injects
100 MW and there are two loads on the system, of 2000 and 3000 MW, then 20
MW flows to the first load and 30 MW to the second, no matter where the two are
located relative to the generator.
Rule 1. Power is assumed to flow proportionally to every load.
Rule 2. Flows are charged their proportional use of each line’s capacity times
the line’s cost.
Having used the first rule to compute the generator’s nominal power flow on
every wire in the system, the second rule is used to compute the charge for using
each wire. The charge is based on the absolute value of the fraction of a line’s
capacity used each hour.7 In other words, if a generator causes a 20-MW flow on
a 100-MW line with a fixed cost of $50/h, then the generator is charged $10/h no
7. This might be termed an absolute impacted megawatt-mile approach and is a design that has been
proposed and inappropriately tested.

matter which way it sends power over the line. This completes the description of
the design in question; these two rules are sufficient to compute the transmission
access charge for every generator. In all other respects the market is assumed to
be perfectly competitive.
The stated purpose of the design focuses only on the linkage between this market
and the market for generation plants, and it ignores linkages to the energy market.
Although these linkages deserve testing, for simplicity, the example will test only
the effect of the rules on the location of new power plants.
These rules were “tested” by constructing a model of all the power flows in
the state of California and computing them for every hour on every wire for a given
year. Graphs were made showing how much power flowed on wires operating at
various voltages. It was claimed that since most of the power flowed on high-voltage
lines, and those were the ones being charged for, the market design would “send
the right signals.” As a consequence of sending the right signals, investors would
build plants in the right locations, and the market would operate efficiently. But
such “testing” misses the point.
Notice that although some costs may have been calculated, no total minimum
cost was calculated (step 2), and no total cost under the proposed design (step 3)
was contemplated. The test did not rule out the possibility that generation costs
would be double the economic minimum under the proposed market design. The
central economic question of cost minimization was simply not considered. These
oversights are now remedied by applying the bottom-line test.
Step 1: Modeling Costs and Benefits

Basic economics always assumes market participants will act to maximize their
benefit net of costs, so cost and benefit functions must be specified to calculate
the predicted behavior. To perform a “bottom-line test,” first determine what costs
and benefits are relevant. In this case, the relevant costs are (1) the cost of genera-
tion including its dependence on location, and (2) the cost of wires. A useful
economic model must include both.
Modeling elastic demand would complicate the computation of benefits and
is not relevant to the benefits being claimed. The basic bottom-line test which
focuses only on supply costs will be sufficient. The first test of a design should
be as simple as possible, and the model needs at least one wire and two generation
locations to evaluate the design. These are specified in the final sub-section which
uses the details of the model to compute costs.
Step 2: Computing Minimum Cost

The bottom-line test requires computing the minimum cost of producing and
delivering the power required to meet a fixed level of demand. In principle this
can be done by trying all ways of producing the required output. Generally a little
mathematics will simplify the process. In the present example, the minimum cost
of delivered power is found to be $362,000/h. For a particular specification of the
test model, step 2 always gives a simple answer that is just one number. No excuse

should be accepted for a more complex or less precise answer although it may be
useful to test several variations of the model (each will have its own one-number
result for step 2).
Step 3: Computing Cost Under the Proposed Rules

The behavior of suppliers must be considered when computing the effect of the
rules on the total cost of delivered energy. Suppliers will react to market rules, and
this reaction must be predicted. Economics enters the test process at this point by
dictating the assumption that suppliers will maximize their profits. This may not
be true when a rule is first put in place, but with practice, suppliers should learn
how to adjust their behavior to the rule and maximize their profit. The profit
maximizing assumption is key to economics, and though not precisely true, is
generally the best approximation available.
In the present model, taking into account the profit maximizing behavior of
generators, the total cost of delivered power will be $376,000/h. Step 3 should
always produce as simple an answer as does step 2.
Interpretation
The result of this example can be summarized as a $14,000 (4%) increase in cost
caused by the proposed rules. This might be tolerable, but a qualitative inspection
of the model raises more concerns. The rules cause all of the generation in the
system to locate at the load center and consequently cause an eight-fold increase
in the cost of transmission. This is unreasonable, and the rules should clearly be
rejected because they have failed such a simple and reasonable test.
The benefit of the bottom-line test is dramatic. By applying it first, the expense
of a complex study can be saved. Additionally, it provides an intuitive understanding
of the design’s flaws that would be difficult to extract from a more complex
simulation. The following subsection builds the model and uses it to compute the
costs for steps 2 and 3, as well as to develop the intuition of the interpretation.
Step 1: Building the Model and Computing Costs. Assume there are two
cities in the market, and they are connected by a transmission line as shown in
Figure 1-9.1. Assume for simplicity that generation can be located near either city,
and the cost is the same: $10/MWh for capacity and $20/MWh for energy. Assume
peak load on the system is 16,000 MW with 12,000 MW located at city A and 4000
MW located at city B. Assume the average hourly load is 10,000 MW, assume that
line capacity costs $4/MWh, and for reliability purposes assume the minimal line
capacity is 500 MW. This is a sufficient model of the relevant costs.
Step 2: Finding the Minimum Cost. Because the above model assumes that
generators cost the same at either location, the cheapest arrangement is to locate
them where they are needed, that is 12,000 MW of generation at A and 4000 MW
of generation at B. In this case, no power line is needed to facilitate trade. The total
long-run cost of power in this model is found by summing three components as
follows:

Figure 1-9.1
Minimum cost locations
with flows calculated
using proposed Rule 2.
16,000 MW × $10/MWh = $160,000/h for generation capacity

+ 10,000 MW × $20/MWh = $200,000/h for energy
+ 500 MW × $4/MWh = $2,000/h for line capacity
for a total of $362,000/h. This is the minimum total cost of serving the load.
Step 3: Finding Total Cost under the Proposed Rules Under the proposed
transmission pricing rules, generators will pay different transmission charges
depending on their location. In particular, if they were located in the minimal cost
arrangement just described, the generators at A would have 1'4 of their flow assigned
to the line from A to B because by Rule 1, 1'4 of it goes to the 25% of load at B.
The generators located at B would have 3'4 of their flow assigned to that line. (This
second power flow is in the opposite direction and cancels the first, but this does
not affect the charges.) Because charges are proportional to flow, the generators
at B are charged three times more than the generators at A. Whenever a generator
is retired at B, it will be replaced with a generator at A because, counting transmis-
sion charges, A is the cheaper place to locate.
The long-run consequence of this dynamic is that all new generators will locate
at A and all expansion of existing generators will take place at A. In the long run,
all generating plants will be located at A. Consequently a 4 GW transmission line
will be built to supply B’s peak load. That is eight times bigger than the 500-MW
line required for reliability in the least-cost configuration. This increases the total
cost of energy by the cost of the extra transmission line. An extra 3500 MW of line
costs an extra $14,000/h. This brings total cost under the proposed design to
$376,000/h.

Part 1

Uploaded by

Copyright:

Available Formats

Part 1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Part 1

Uploaded by

Copyright:

Available Formats

What are the main parts of the book discussed in the prologue?

What are the main parts of the book discussed in the prologue?

How does the author propose to analyze power markets in the different parts?

How does the author propose to analyze power markets in the different parts?

Part 1

Power Market Fundamentals

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

P OWER SYSTEM ECONOMICS PROVIDES A PRACTICAL INTRODUC-

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

READING TO DIFFERENT DEPTHS

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

READING OUT OF ORDER

Conventions Specific to this Book

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Conventional Terms That Are Sometimes Misinterpreted

Profits. “Profit” is used by economists to mean long-run economic profits, and

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

I N THE BEGINNING THERE WAS COMPETITION—BRUTAL AND INEFFI-

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

On the scale of an isolated city, provision of electricity is a natural monopoly

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

1-1.1 CONDITIONS FOR DEREGULATION

Did Cheap Gas Turbines End the Natural Monopoly?

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

1-1.2 PROBLEMS WITH REGULATION

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Regulation has two fundamental problems: (1) it cannot provide a strong

The Regulator’s Dilemma

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

1-1.3 THE BENEFITS OF COMPETITIVE WHOLESALE MARKETS

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

1-1.4 THE BENEFITS OF REAL-TIME RATES

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Result 1-1.1 Savings from Real-Time Rates Would Be Small

This 3% reduction in supply cost is offset by a cost increase on the demand

1-1.5 PROBLEMS WITH DEREGULATING ELECTRICITY

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Two Demand-Side Flaws

Demand-Side Flaw 1: Lack of Metering and Real-Time Billing

The first demand-side flaw causes a lack of demand responsiveness to price

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Complexity and Local Market Power

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Nothing is more terrible than activity without insight.

D ELIVERED POWER IS A BUNDLE OF MANY SERVICES. These include

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

speculation about granting rights to congestion rents to investors in transmission

1-2.1 ANCILLARY SERVICES AND THE SYSTEM OPERATOR

Regulation and Balancing

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

3. See Chao and Wilson (1999a) and Wilson (1999).

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Voltage Support and Black-Start Capability

The System Operator Service

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.

Nonprofit “monopolies” are quite different from for-profit monopolies. For

1-2.2 UNIT COMMITMENT AND CONGESTION MANAGEMENT

© February2002. Steven Stoft, Power System Economics (IEEE/Wiley) ISBN 0-471-15040-1.