Notes on J. Weiss, "Market Power Issues..."
Prepared by L. Tesfatsion

Last Updated: 26 June 2002


Prepared by: Leigh Tesfatsion
Date:        6 September 1999   
For:         Econ/Electricity Project Meeting, September 7

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

                        NOTES ON

    "Market Power Issues in the Restructuring of the Electric
       Power Industry:  An Experimental Investigation"

         by Jurgen Weiss, Harvard Business School

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

In this paper, results from a number of experiments with industry
subjects conducted over the Internet illustrate the impact of

  seller concentration -- effects of increasing the number of
                          sellers in different geographical areas,
                          holding everything else the same (permits
                          analysis of "local market power" pockets
                          where consumer prices remain high despite
                          decreases in seller concentration)

  demand side bidding -- active bidding or passive behavior

  transmission constraints -- limited transmission capacity, using a
                              simple but realistic network of
                              transmission lines connecting the
                              various sources of demand and supply
                              (permits effects of buyer and seller
                              locational differences to be examined)

  two alternative pricing rules -- nodal pricing and uniform pricing

on the nature of competition in a simulated market for electricity.
The context of the experiments is a simulated electricity market in
which the interaction between network constraints and market
competitiveness is explicitly modelled.

Previous work closest in spirit to this work: Cardell, Hitt, and
Hogan (1997), Nasser (1997), and Borenstein et al. (1997).

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

TWO BASIC ISSUES:

EFFICIENCY:

Any gains in efficiency achieved by introducing the proper
incentives for profit maximization through deregulation and
privatization have to be weighed against the potential losses of
efficiency resulting from the exercise of market power by the
deregulated or privatized electric power companies.

EQUITY:

Even if electricity markets become more efficient overall, gains in
efficiency may not benefit all market participants equally.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

HYPOTHESES TESTED:

Is an increase in the number of sellers likely to effectively curb
any potential for market power exercise in deregulated electricity
markets, and, if not, what other regulatory strategies are most
likely going to be effective in mitigating any such market power?

   NOTE:  Key aspects of market structure

         NB = number of buyers
         NS = number of sellers
         Size distribution of buyers
         Size distribution of sellers

   Weiss confounds concentration and capacity effects in his
   experiments, e.g., looks at cases with many small buyers
   and a few large buyers, so size and concentration are
   simultaneously changed.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

KEY FINDINGS:

As summarized on page 3, Weiss finds that simply increasing the
number of sellers in a given geographic area may not effectively
reduce market power.  Rather, certain sellers maintain "local market
power" and resulting consumer prices remain high.

His experiments suggest an alternative strategy for reducing
sellers' ability to exercise market power: he finds that
experimental markets with countervailing market power by a few large
buyers who are actively bidding for power were characterized by
essentially competitive prices, even at high seller concentration.

Weiss claims that this finding confirms previous findings by
Bakerman, Denton, Rassenti and Smith (1997) and Borenstein and
Bushnell (1997), who explored this issue in a simpler context.  The
current study's experiments involve a more realistic representation
of a network for transmission of power which allows a number of
additional issues to be addressed, and a more careful study of the
relation between seller concentration and market power. Although the
same findings are supported, Weiss claims (p. 5) they are supported
for very different reasons as far as the impact of seller
concentration on market power is concerned.

The existence of congestion in the transmission system does not
alter this result.

Bottom Line:  Active demand-side bidding by large buyers is a
potentially important policy instrument in the process of rule
making for liberalized electricity markets.

Policy Implications: Investing in technology that leads to greater
demand elasticity, and hence gives active wholesale buyers more
bargaining power, may -- through the creation of countervailing
market power on the demand side -- be at least as effective as
investing in new generating capacity (i.e., reducing seller
concentration) in terms of limiting sellers' ability to exercise
market power.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

MODELLING ASPECTS: (pp. 5 -->)

Least Cost Dispatch, Nodal Spot Prices, and Uniform Prices

ONE APPROACH TO LEAST COST DISPATCH: NODAL PRICING

Centralized dispatch is assumed to solve the following welfare
maximization problem:

     MAX  Net Social Benefits NSB (Consumer + Producer Surplus)

     Subject to  Network Balance
                 Kirchoff's Law Constraints (e.g., when power is
                 transported between two points of the transmission
                 system, power will flow over all available
                 "parallel paths" between those two points -- p. 4)


Lagrangian Form:  NSB  + lambda*[network balance constraints]
                       - mu*[Kirchoff's Law Constraints]



AS LONG AS ALL MARKET PARTICIPANTS BEHAVE COMPETITIVELY -- i.e.,
take prices as given in determining their quantity demands and
supplies, with no attempt to engage in strategic behavior with
respect to any variable (e.g., congestion, capacity) -- least cost
dispatch can be achieved with a system of prices called NODAL SPOT
PRICES.

A nodal price (i.e., price of electricity at a node) is the shadow
value of power injection at the node.  That is, it is given by the
value of the Lagrange multiplier for the network balance constraint
for that node in the least cost dispatch problem.

Nodal spot prices can change instantaneously as a function of
stochastic generation or transmission outages and stochastic
weather-related demand changes, and that can differ by location.
(Work by Schweppe et al.)

SECOND APPROACH TO LEAST COST DISPATCH: UNIFORM PRICING

The uniform price is determined by constructing a single pair of
supply and demand curves from the bids and offers submitted, by
ordering all bids in decreasing order and all offers in increasing
order.  The uniform price is determined by the intersection of the
supply and demand curves and represents the bid made by the marginal
generator.

By construction, then, the uniform price does not vary by location,
as do nodal prices.  Also, it ignores any constraints imposed by the
transmission system.

Using the language of the English spot market for electricity, the
uniform price is called the SMP (System Marginal Price).

The pattern of load and generation, upon which the SMP is based, may
not be feasible once transmission constraints are taken into
account.

Some agents who would buy or sell power in the unconstrained case
must then be denied power -- they are called CONSTRAINED-OFF buyers
or sellers.  Other agents who would not be buying or selling power
in the unconstrained case but who do so in the constrained case are
called CONSTRAINED-ON buyers or sellers.

In either case, a transfer payment is made FROM buyers TO
constrained-on and constrained-off buyers and sellers so that they
are indifferent between their output and purchases under either
constrained or unconstrained dispatch.  This transfer payment is a
surcharge per unit of electricity bought and is called an UPLIFT
PAYMENT.

As a result of this mechanism, all costs associated with
transmission constraints are averaged and born by the demand side,
who pay the Pool Purchasing Price (PPP).  It also means that the
bids entered by the demand side represent bids for the SMP, and
implicitly represent an agreement to pay whatever Uplift charges
result from transmission constraints at the time of dispatch.
It can be expected that demand side bidders will take into account
likely uplift payments when making their bids.

    QUESTIONS: Are the uplift charges in the form of side
    payments (i.e., lump sum transfers rather than price
    adjustments)?  How to choose just who gets constrained-on
    and constrained-off?  Can the uplift payments feasibly be
    supported from within the dispatch system per se, are are
    fund injections from outside the system needed?  What is the
    exact form of compensation to constrained-OFF agents?

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

SECTION 4: THEORIES OF MARKET POWER IN ELECTRICITY MARKETS
(pp. 7 -->)

Market Power = a seller's ability to profitably change her bidding
behavior away from competitive bidding, resulting in higher market
prices, reduced market efficiency, or a higher share of surplus
captured by sellers.

       NOTE: This is at a pretty abstract level and confounds
       definition with implications.  Also, the benchmark of
       comparison ("competitive bidding") may not be easy
       to quantify in any actual situation.

4.1 Seller Concentration and Market Power:

Standard presumption: Inverse relation between seller
                      concentration and market competitiveness


        Higher concentration <--->  Higher prices, lower
                                    market efficiency

Problems with this hypothesis:

A. Seller concentration does not take into account any underlying
demand or supply elasticities, both of which have an important
impact on how much prices should rise as the number of firms in a
market increases.

B. Seller concentration indices does not capture the threat of
potential entrants and the threat of additional regulations.

C. The measurement of seller concentration is further complicated by
the difficulty of determining the size of the relevant market.
     Depending on supply and demand conditions as well as on the
state of the transmission system, the ability to supply power from
one geographic area to another may be severely limited during
certain times and easy during others, so that the size of the market
depends on the degree of congestion of the transmission system which
in turn depends on the actions taken by all firms in the market.
     In the context of uniform prices, there is no local market
power as far as the determination of the SMP is concerned, since
transmission constraints are ignored.  However, local market power
can be exercised by manipulating bids of plants, which are likely to
be constrained-on or constrained-off in ways that increase uplift
payments.

     KEY POINT: The likelihood of specific plants being
constrained-on or constrained-off depends only on demand and the
seller concentration AT THE NODE AT WHICH THE PLANT IS LOCATED,
and is not reduced by an increase in the number of sellers located
elsewhere in the network.  Hence the ability to influence uplift
payments may be independent of the overall number of players
owning generating assets.

4.2 Demand Side Activity and Market Power

There is little theoretical work in the industrial organization
literature on the impact of demand side bidding on market prices.

If the concentration of buyers is comparable to that of sellers,
one would expect buyers to act strategically and thereby to extract
a larger amount of social surplus than they would under passive
demand-side bidding.

Demand-side bidding by concentrated buyers is different from entry
by new sellers (leading to lower seller concentration) in that,
under certain circumstances, the incentives of both incumbent and
entrant sellers may be aligned while those of buyers and sellers are
always opposed.

It can therefore be expected that active buyers will exert more
competitive pressure on prices than entering sellers, at least under
some circumstances.

      NOTE: This suggests the need to look at the numbers of buyers
      and sellers in absolute terms and not just the ratio of these
      two numbers.

Two ways in which a small number of active buyers may force markets
to be more competitive (p. 10):

(1) presence of active rather than passive buyers introduces some
uncertainty about actual levels of demand for sellers trying to
decide on optimal bids, which may induce sellers to lower their bids
relative to environments with no uncertainty in which a very large
number of buyers make demand-revealing bids.

(2) active buyers may force markets to be more competitive by
strategically withholding demand, i.e., by underrevealing the true
value of the demand they serve (thinking of large buyers as
wholesale buyers representing a large number of individual
consumers).  This is suggested by experimental findings on
ultimatum and dictator games that suggest buyers are willing to
incur losses in one-period games if the proposed split of social
surplus is perceived to be unfair.


4.3 Transmission Capacity and the Exercise of Market Power

The existence of binding transmission constraints may increase the
degree of market power for any given overall level of market
concentration.  It may be necessary to expand transmission capacity
beyond the capacity suggested by engineering considerations in order
to alleviate any existing local market power.


5. DESCRIPTION OF THE EXPERIMENTS (PP. 11 -->)

Weiss claims his experiments differ from previous electricity market
experiments in two important ways:

(1) They include a more accurate representation of the electric
transmission network -- the concept of nodal pricing as well as the
concept of uplift can only be studied at the level of complexity
studied here.

(2) They make use of experienced industry subjects only, not
students

Buyers and sellers were located at different nodes of a simple
network of transmission lines.

Attention is limited to the case of no transmission losses and to
active power only -- reactive power considerations are ignored.

MY FIGURES 2-3 ARE NOT CLEAR -- DIFFICULT TO DETERMINE ACTUAL
NETWORK SETUP

Fixed generation capacity of 1300 MW distributed equally (??) among
all sellers and located at 4 transmission nodes (a,b,c,d) connected
through five (??) transmission lines.

A competitive fringe owns 400 MW of capacity through four plants
with 100 MW capacity each.  The remaining 900 MW of capacity was
equally distributed among all players (remaining sellers?).

Transmission capacity of the lines connecting the four nodes was
unlimited, except in the case of the line connecting nodes b and d.
The transmission limit on that line was 100 MW during the first 23
rounds of the game and unlimited thereafter.

The market was organized as a clearinghouse or call auction.  In
this institution, all bids and offers are posted simultaneously,
and market prices are generally determined by the intersection of
bid and offer arrays.  No bilateral transactions between individual
players are allowed.

Call auction easier to implement via email than a continuous double
auction, in which the timing of bids and offers is crucial.  Playing
over the Internet, bids and offers are subject to congestion delays.
[This is one reason why earlier plans to include a continuous double
auction mechanism in the California market have recently been
abandoned.]

In both pricing environments tested (nodal and uniform), an
independent system operator (ISO) plays the role of the market
clearinghouse and uses the bids provided to decide on the optimal
pattern of load and generation.  Unfortunately, this system may
not be truth-revealing -- bids entered by market participants
may not correspond to true underlying costs and demand values.

All power plants in our experiments had constant variable costs over
the entire range of output and had the same capacity.  Plants A and
E differed from other plants by a penalty incurred if they were not
at least partially dispatched.  This penalty was created to reflect
the must-run character of many base-load plants in real electricity
markets.

       QUESTION: What is a "base-load plant"?

A substantial part of demand from the largest customer group was
"must-serve" load.  Not buying enough power to serve all must-serve
load resulted in a severe penalty for buyers.

Transmission system including four nodes permits loop flow effects
to be studied, since there are several "parallel" paths for
transporting power between any two points.  Weiss claims his
model is the first to permit loop flow effects to be systematically
examined in lab setting.

5.3 SUBJECT POOL (p. 14 -->)

The 180 participants in the experiments were recruited
internationally via email advertisements to individuals associated
with the electricity industry.  [Participants told they could gain
insights into the workings of competitive electricity markets and
that they could win a prize of $2000.]

Benefits:  Electricity rules of the game difficult for students
           to comprehend, easier for actual market participants

Danger: Market participants have vested interest in outcomes and
        could game the experiments to make outcomes support their
        own particular point of view

5.4 EXPERIMENT LOGISTICS (p. 16-->)

Conducted over a twelve week period in 1997.  A typical week
consisted of 3 rounds.  Overall, 33 rounds were completed, of which
the first five were practice rounds.

Each participant was randomly assigned to an experimental group and
to a role within that group.  Participants then received a detailed
description of the game. All trading took place via email over the
Internet.

Each subject received feedback about his own performance after each
round of play.  Each player earned tickets to a lottery over a grand
prize of $2000.  Game profits --> lottery tickets

5.5 TREATMENT VARIABLES (pp. 17 -->)

Four treatment variables

Seller Concentration:

      Three levels  -- one seller (monopoly), three sellers
      (triopoly), and six sellers.

Demand Side Activity Level: Active or Passive

       Active --> wholesale buyers were actual players making bids
                  to buy power

       Passive --> Demand side was simulated as fully demand
                   revealing

Pricing Algorithm:

       Nodal pricing and uniform pricing

Transmission Constraints:

       Only imposed on one line, and only for the first 23 rounds
       of the game.  Final ten rounds were played without the
       transmission constraint.

The experimental Design is depicted in Table 2 (page 18).

The design was not fully crossed.  Only eight cells were played,
with six observations in each cell.  The choice of active cells
was guided by a desire to test as many realistic mitigation
strategies as possible.  Therefore, an active demand side was only
introduced for the three-seller environment.


6.  MARKET POWER RESULTS (PP. 18-->)

The market power analysis measures both efficiency and equity and
considers both measures important indicators of market power.

Market efficiency measure = deadweight loss (dwl)

Market equity = share of social surplus captured by the sellers
                relative to the share captured in a competitive
                environment (relshare)

              = measure of market power

Because the concept of price is different in nodal and uniform price
environments, price cannot be used as a measure of market power
whenever the two types of environments are directly compared.

6.3 GENERAL RESULTS (PP. 19-->)

Tables 4--6 give regression results

    COMMENT BY MATT SMITH: Why not a panel data study in
    place of a regression study?  (Panel data is data
    that separately tracks outcomes for specific individuals
    over time rather than averaging outcomes across agents.)

FINDINGS FOR SELLER CONCENTRATION: (pp. 26--->)

In all environments not including a monopoly seller, results during
medium and low demand periods were insignificantly different from
competitive outcomes. Consequently, Weiss limits his attention to the
high demand period in all following analyses.

In general, nodal prices tend to be lower the larger the number of
sellers in the market.  Monopoly prices are higher than tripoly
prices which in turn tend to be higher than six-seller environment
prices.  There are, however, some interesting differences across
nodes with respect to how prices change as the number of sellers is
increased that tend to support the hypothesis regarding the relation
between prices and local market power strengths.

Results are similar for uniform price environments -- the SMP drops
consistently as the number of sellers is increased.  Also, the
magnitude of the uplift does not change dramatically as the number
of sellers is increased.

It is not easy from the graphs to see any consistent pattern in
deadweight loss across experiments.  Therefore, relshare may be a
better indicator of market power than the efficiency measures in the
context of this experimental design.

The portion of social surplus captured by producers is highest for
monopoly environments, followed by three-seller environments.
Six-seller environments and environments with active demand side
bidding tend to exhibit the lowest producer shares, hence tend to be
closest to the competitive outcome.

OVERALL CONCLUSIONS:

Some support for the traditional view of the relationship between
seller concentration and the exercise of market power.  Market power
does tend to decrease as the number of sellers is increased, but
there may be important local market power leading to persistently
high prices at some locations or to persistently high uplifts in
uniform price environments.

The regression results confirm the graphical analysis.

While both nodal and uniform price environments show evidence of the
exercise of local market power, the nature of local market power
differs between the two environments.

In nodal price environments, only players with plants that are
likely to be constrained-on have local market power.  In uniform
price environments, both players with constrained-on and with
constrained-off plants can exercise market power by manipulating
their offers for plants which are likely to be either constrained-on
or constrained-off.  As a result, profits in nodal price
environments are concentrated in the hands of those players who own
constrained-on plants.  They are distributed more evenly in uniform
price environments.

FINDINGS FOR DEMAND-SIDE ACTIVITY: (pp. 30--->)

Active demand-side bidding significantly lowers market power for a
given number of sellers and therefore promises to be an excellent
market power mitigation strategy.  It is successful in removing any
local market power found in nodal price environments...  It also
significantly lowers the SMP in uniform price environments, but has
no impact on the magnitude of the uplift.

It appears that a large portion of the significant reduction in
market power with active demand-side idding can be attributed to
the fact that in those environments sellers face increased
uncertainty about actual levels of demand...

However, we also observe that more than half of the actively bidding
buyers in our experiments act strategically, and aggressively try to
lower market prices.  This in turn further reduces the extent of
market power exercise by sellers.

This second benefit of active demand-side bidding by a small number
of buyers is unlikely going to be achieved without the creation of
bilateral markets in which buyers are allowed to concentrate to
levels comparable to sellers.

6.6 FINDINGS REGARDING TRANSMISSION CAPACITY (PP. 33-->)

Results strongly support hypothesis 4: when the transmission
constraint is binding, which is most often the case during high
demand periods, certain producers are effectively excluded from
competing with their entire capacity, thus reducing the amount of
overall capacity available to meet demand and hence increasing the
effective market share of those sellers not excluded through an
existing transmission constraint.  Sellers in the experiments were
able to exploit this situation successfully.

7 OVERALL SUMMARY (PP. 34-->)

Standard solution for decreasing market power -- an increase in the
number of sellers competing in the total market -- may not be
sufficient to lower prices at all market locations if transmission
capacity is limited and under certain demand conditions.

While an expansion of the transmission system will remove such local
market power, we find that demand side bidding may be an equally
powerful market power mitigation strategy, in particular in a system
of nodal prices.

The pricing mechanism has important implications for the
distribution of rents among buyers and sellers as well as for the
sources and remedies of market power.

Under nodal pricing, the profits accruing to buyers and sellers in
the market will differ substantially across the newtork, even when
the underlying cost structure of plants and resale values of demand
are identical.  On the other hand, these differences are an
indicator of where the value of electricity is high and where it is
low, and can thus guide decisions for either entry of new or
decommisssioning of existing generation facilities.

With respect to high nodal prices at some locations, an effective
mechanism for lowering them is active bidding power by a small
number of buyers on the demand side of the market.  Therefore,
policy measures aimed both at increasing demand elasticity through
technology and aiming at increasing the bargaining power of the
demand side by allowing active participation of concentrated
intermediate buyers may provide promising means of making the
resulting markets more competitive.

Uniform prices as used in the British spot market for electricity
have their own set of problems, most of which are confirmed by our
experiments.  Strategic use of uplift.  The total amount of rents
extracted from consumers via uplift payments in uniform price
environments is not substantially reduced as market concentration is
lowered due to the influence of transmission constraints.

In conclusion, it appears that both nodal and uniform pricing rules
create similarly strong incentives to exercise local market power
and that this local market power is not easily mitigated by
increasing the number of sellers in the market.

Split of profits more uneven under nodal pricing.

But nodal pricing provides correct signals for entry and exit in the
long run and correct price signals for equating short run marginal
costs and benefits.  Also, active bidding by a small number of
buyers as a means of reducing seller market power is more effective
in nodal spot pricing environments than in uniform pricing
environments.

Therefore, nodal pricing may be the preferable alternative over a
system of uniform pricing.