University of Southern California Dissertations and Theses

Emissions markets, power markets and market power: a study of the interactions between contemporary emissions markets and deregulated electricity markets

(USC Thesis Other)

Emissions markets, power markets and market power: a study of the interactions between contemporary emissions markets and deregulated electricity markets

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content EMISSIONS MARKETS, POWER MARKETS AND MARKET POWER:
A STUDY OF THE INTERACTIONS BETWEEN
CONTEMPORARY EMISSIONS MARKETS AND
DEREGULATED ELECTRICITY MARKETS
by
Noah Christopher Dormady
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PUBLIC POLICY, PLANNING AND DEVELOPMENT) August 2012
Copyright 2012 Noah Christopher Dormady
ii
Dedication
This dissertation is dedicated to my beautiful wife, Kelly Michèle Dormady. She is the
reason I pursued a graduate degree to begin with, and she has been a source of continual support,
encouragement, enrichment, faith and love. I enjoy going through life with her.
iii
Acknowledgements
This dissertation would not have been possible without the help of a tremendously
supportive and encouraging faculty committee. The faculty, staff and students of the USC Sol
Price School of Public Policy (formerly the School of Policy, Planning and Development) were
continuously helpful and engaging during my time at the University of Southern California. I am
eternally grateful for their time and assistance.
I would like to thank Daniel Mazmanian, my Advisor and Committee Chair, for his
tireless efforts to ensure that my research and work were supported and productive. Without his
generous assistance and willingness to speak up for me, I would not have completed my PhD. It
is because of his support that I was able to fund my experimental research, travel to Berlin,
Germany, and take a tenure-track job.
I would like to thank Adam Rose for his selfless help and encouragement of my work.
Every week for the past four years, he met with me or called me on the phone to make sure that I
was on the right track, did not get caught behind any hurdles, or falter in my progress. He made
sure that my work was not overlooked by faculty or administrators, spoke with colleagues about
my work at conferences, and continually directed me to experts who could provide me with
information or assistance. He was kind enough to involve me in a significant number of research
projects, funded me throughout the summers, and took the time to publish peer-reviewed papers
and professional reports with me. I have learned more from him than anyone.
Similarly, I would like to thank John Jurewitz of Pomona College, and Simon Wilkie of
the USC Department of Economics for their support of my work. John Jurewitz was willing to sit
with me for long hours to make sure that I thoroughly understood the fundamentals of electricity
deregulation and California's electricity market design. Simon Wilkie met with me on numerous
occasions to offer invaluable feedback on my work, particularly my experimental design and
iv
application. Lastly, I would like to thank Anthony Bertelli and Yan Tang for their time and
assistance in helping me formulate my ideas into a coherent and logical work.
I would like to express my heartfelt gratitude to my brother, Micah Dormady, for his
continual willingness to sit down and help me with advanced concepts in mathematics, statistics,
and with computer problems. Whenever I had a problem, he was the first one I called. Similarly,
Javier Livio has been a tremendous friend who was willing to provide hours of his time, away
from his family, to help me with the development of my software applications and research
design. I would also like to thank John Lin, for his many hours of time in helping me flesh out
design problems in my experimental research. I benefitted greatly from his wisdom. And, I
would like to thank Elena Maggioni for her willingness to give me constructive feedback and
advice on my work, and for her many hours of time helping me prepare for exams.
I would like to acknowledge the support of some very helpful administrative staff at USC.
These include June Muranaka, Suzanne Alexander, Aubrey Hicks, and Young Miller. Without
their administrative help, I have no idea where I would be.
Finally, I would like to thank my personal mentors: Rob Hannah, Paul Wollenberg,
Thaddeus Wakefield, Jon Valerio, and Arnold Rodriguez.
v
Table of Contents
Dedication .......................................................................................................................................ii
Acknowledgements ........................................................................................................................iii
List of Tables ................................................................................................................................. vi
List of Figures ...............................................................................................................................vii
Abstracts ........................................................................................................................................ ix
Chapter 1: A Monte Carlo Approach ............................................................................................. 1
1. Introduction ................................................................................................................. 1
2. A Brief Background on the Literature ......................................................................... 2
3. Why Market Power?: Background and Structure......................................................... 3
4. The Two Stage Uniform-Price Sealed-Bid Auction .................................................... 8
5. The Methodology of the Monte Carlo Approach ...................................................... 12
6. Monte Carlo Simulation Results ............................................................................... 15
7. Implications and Conclusions ................................................................................... 30
Chapter 2: An Experimental Approach ........................................................................................ 32
1. Introduction ............................................................................................................... 32
2. Experimental Method ................................................................................................ 34
3. Experiment Design .................................................................................................... 37
4. Strategy in the Simultaneous Two-Stage Market ...................................................... 45
5. Results ........................................................................................................................ 57
6. Conclusions ............................................................................................................... 78
Chapter 3: Regulatory Mechanisms and Policy Approaches ....................................................... 80
1. Introduction ............................................................................................................... 80
2. A Brief Primer on Deregulated Power Markets ........................................................ 81
3. A Brief Primer on Emissions Markets ...................................................................... 88
4. The Exercise of Market Power .................................................................................. 97
5. Virtual "Convergence" Bidding .............................................................................. 103
6. Consignment Auctions ............................................................................................ 108
7. Multilevel Accounts ................................................................................................ 113
8. Conclusion ............................................................................................................... 119
References ................................................................................................................................... 122
Appendix A: Market Conduct and the Herfindahl-Hirschman Index ........................................ 129
Appendix B: Design and Operation of the Monte Carlo Simulation
Software Oligopsony 1.0 .................................................................................... 132
Appendix C: Experiment Environment ...................................................................................... 140
vi
List of Tables
1.1 Herfindahl-Hirschman Index (HHI) of Past RGGI Auctions ..................................... 7
1.2 Auction Price Responsiveness to Demand Reduction ............................................. 24
2.1 Summary of Treatment Effects ................................................................................. 43
2.2 Symmetric Equilibria Predictions ............................................................................. 56
2.3 Auction Clearing Summary Statistics ....................................................................... 60
2.4 Two Sample Wilcoxon (Mann-Whitney) Non-parametric Hypothesis Tests ........... 61
2.5 Regression Results for Aggregate Energy Analysis ................................................. 68
2.6 Regression Results for Aggregate Emissions Analysis ............................................ 69
2.7 Regression Results for Individual Analysis .............................................................. 77
3.1 Key Market Design Criteria of Major Cap-and-Trade Programs ............................. 93
B.1 Sampling Methodology for User-Supplied Market Parameters ............................. 134
vii
List of Figures
1.1 Oligopsonist Profit Under Demand Reduction ......................................................... 16
1.2 Market Clearing Price Under Demand Reduction .................................................... 18
1.3 Mean Total Fringe Profit Under Demand Reduction ................................................ 19
1.4 Oligopsonist Profit Under Demand Reduction (75 Bidders) .................................... 21
1.5 Oligopsonist Profit Under Demand Reduction (25 Bidders) .................................... 22
1.6 Market Clearing Price Under Demand Reduction (75 Bidders) ............................... 23
1.7 Market Clearing Price Under Demand Reduction (25 Bidders) ............................... 24
1.8 Mean Total Fringe Profit Under Demand Reduction (75 Bidders) ........................... 25
1.9 Mean Total Fringe Profit Under Demand Reduction (25 Bidders) ........................... 26
1.10 Kernel Density Plot of Individual Oligopsonist Profit ............................................ 27
1.11 Kernel Density Plot of Auction Clearing Price ....................................................... 28
1.12 Kernel Density Plot of Total Fringe Profit .............................................................. 29
2.1 Energy-Emissions Experiment Flow Chart ............................................................... 41
2.2 Type I Market Power (Supply Reduction) ................................................................ 52
3.1 Hourly Average System-wide Load (California ISO, July, 2012) ............................ 84
3.2 RGGI Auction Clearing Prices ................................................................................. 95
3.3 Competitiveness Measure of Past RGGI Emissions Auctions (HHI) ....................... 99
3.4 RGGI and AB 32 Auction Mechanisms ................................................................. 109
3.5 Flow Chart of CARB Multi-level Accounts ........................................................... 115
3.6 Single Firm Holding Limit under AB 32 ................................................................ 117
B.1 Data Input Screen ................................................................................................... 132
B.2 Individual Bid Matrix ............................................................................................. 137
viii
B.3 Data Output Screen ................................................................................................ 138
C.1 Round #1 Bidding Screen ...................................................................................... 160
C.2 Round #1 Summary Screen .................................................................................... 161
C.3 Round #2 Bidding Screen ...................................................................................... 162
C.4 Round #2 Summary Screen .................................................................................... 163
C.5 Round #1 Bidding Screen for Type C (Oligopolist) .............................................. 164
C.6 Round #1 Bidding Screen for Type B (Fringe Player) ........................................... 165
C.7 Round #1 Summary Screen for Type C ( Oligopolist) ........................................... 166
C.8 Round #1 Summary Screen for Type B (Fringe Player) ........................................ 167
C.9 Round #2 Bidding Screen for Type A (Fringe Player) .......................................... 168
C.10 Round #2 Summary Screen (All Types) .............................................................. 169
ix
Abstracts
Chapter 1: A Monte Carlo Approach
The use of auctions to distribute tradeable property rights to firms in already heavily
concentrated markets may further exacerbate the problems of market power that exist within
those markets. This chapter provides a model of a two-stage emissions market modeled after a
contemporary regional permit trading market in the United States, the Regional Greenhouse Gas
Initiative, Inc. (RGGI). It then introduces Oligopsony 1.0, a C# software package constructed in
the .NET environment that simulates uniform-price auctions using stochastic Monte Carlo
simulation for modeling market power in tradeable property rights auctions. Monte Carlo
methods add a probabilistic element to standard auction theoretic equilibria. The results of these
simulations indicate that there can be significant non-linearities between profit and market power
as exercised through strategic demand reduction. This analysis finds the optimum point of
strategic demand reduction that enables the firm to exploit these non-linearities, and it determines
the probability distributions of these optima using kernel density analysis.
Chapter 2: An Experimental Approach
How will emerging auction-based emissions markets function within the context of
today’s deregulated auction-based electricity markets? This chapter provides an experimental
analysis of a joint energy-emissions market. The impact of market power and collusion among
dominant firms is evaluated to determine the extent to which an auction-based tradeable permit
market influences performance in an adjacent electricity market. The experimental treatment
design controls for a variety of real-world institutional features, including variable demand,
permit banking, inter-temporal (multi-round) dynamics, a tightening cap, and resale. Results
x
suggest that the exercise of market power significantly increases electricity auction clearing
prices, without significantly increasing emissions auction clearing prices, and in some cases, even
significantly suppresses them. The institution of auction-based carbon markets in the already-
concentrated energy sector can further strengthen the market position of dominant firms who can
leverage energy-emissions market linkages to their operational advantage.
Chapter 3: Regulatory Mechanisms and Policy Approaches
Contemporary deregulated electricity markets are defined by a complex array of multi-
settlement markets, with additional market-based mechanisms designed, to a large extent, to limit
the exercise of market power by dominant firms. On top of the already complex nature of these
markets, policymakers are also adding market-based mechanisms to curtail greenhouse gases.
Key linkages exist between electricity and emissions markets that may be utilized by dominant
firms. This chapter provides an analysis of three specific policy mechanisms that are utilized in
contemporary markets to effectively reduce the incentive of dominant firms to exercise market
power. These include convergence bidding, consignment auctions and multilevel holding
accounts.
1
Chapter 1: A Monte Carlo Approach
1. Introduction
As the international community looks to market-based mechanisms to address negative
externalities such as climate change, the success and efficiency of extant markets can play heavily
into design and operation considerations for future market design. Traditionally, transferable
property rights (cap-and-trade) markets have utilized centralized allocation of property rights
(emissions allowances or permits). That approach, although generally effective, has been shown
to lead to inefficiencies such as regulatory capture and political misallocation (Arimura, 2002;
Dewees, 2008; Ellerman & Montero, 1998). Emerging cap-and-trade programs have improved
upon this by utilizing market-based allocation through auctioning initial property rights. Because
the initial allocation can influence both the efficiency and competitiveness of the emissions
market, the performance of these auctions is of central importance. And, because the firms that
operate within these auctions are the same concentrated firms that operate within deregulated
electricity markets, the issue of concentration and the exercise of market power in emissions
auctions is of central importance.
The purpose of this chapter, therefore, is to evaluate the degree to which the strategic
exercise of market power can influence the performance of emissions auctions. Following a brief
review of extant literature, this chapter introduces a model of a contemporary two-stage auction-
based emissions market. A Monte Carlo emissions auction simulation software, Oligopsony 1.0,
is then introduced. A set of simulation results is then presented, based upon parameters roughly
consistent with a contemporary U.S. market, the Regional Greenhouse Gas Initiative, Inc. (RGGI).
Sensitivity analyses and probability density analysis follows.
2
2. A Brief Background on the Literature
The theory of market power in emissions markets is developed by Hahn (1984) whose
analysis of the Los Angeles region emissions market considers the case of a single dominant firm
among smaller competing fringe firms. Hahn's analysis suggests that the nature of market power
is a product of the initial degree of misallocation, which can transform the dominant firm into
either a dominant buyer or seller, who can then reap excessive profits by exploiting the inelastic
portions of competitors' demand curves.
This is furthered by the work of Misiolek and Elder (1989) who suggests that those
dominant firms have altogether higher valuations in the emissions market because they are
willing to pay for increased market share, barriers to entry, and the exclusion of rivals in common
product markets (Rogerson, 1984; Salop et al. 1987, 1984, 1983; Williamson, 1968). Moreover,
limits on the exercise of market power have been extended as far as the sanction cost for
noncompliance (Chavez & Stranlund, 2003; Malik, 2002; Van Egteren & Weber 1996).
However, others have suggested that market power, despite its presence in emissions
markets, is rather weak (Tietenberg, 2006), and has only minuscule impacts on market prices
(Hagem & Westkog, 1998; Liski & Montero, 2005). On the other hand, laboratory experiments
have provided robust evidence on market behavior in recent years (Milgrom, 2004). Emissions
market experiments have provided evidence that the exercise of market power can be rather
extreme (Godby, 2000; Holt, 1989; Muller et al. 2002; Wrake et al. 2008). And Godby (2000) provides even more striking results than Hahn (1984) in terms of the potential for market power
to be exercised.
Auctions have been analyzed as an alternative allocation method for addressing the
problem of misallocation. The literature lauds auctions for their overall system efficiency
improvements (Joskow, Schmalensee & Bailey, 1998; Parry et al. 1999; Ruth et al. 2008;
3
Tietenberg, 2006; Van Dyke, 1991; Wrake et al. 2008), for their strengths in reducing tax
distortions, creating market flexibility, creating innovation incentives, and disincentivizing rent
seeking (Cramton & Kerr, 2002). And they are lauded for their redistributive strengths; their
ability to allow government to offset social costs (Bovenberg & de Mooij, 1994; Bovenberg &
Goulder, 1996; Goulder et al. 1999; Parry et al. 1999; Smith et al. 2002; Wrake et al. 2008;).
Just as the Coase Theorem suggests that overall system efficiency is independent of the
initial distribution of the property right, Vickrey (1961) argues that the efficiency of auctions, and
the revenue they generate, is independent of the format of the auction. However, just as the
Coase Theorem is built upon a series of assumptions that are sometimes tenuous in practice,
Vickrey (1961) makes two major assumptions. He assumes that all bidders are risk neutral, and
that bidder valuations are identically and independently distributed (I.I.D.). These assumptions
have been handsomely challenged (Maskin & Riley, 2000; McAfee & McMillan, 1987).
Furthermore, bidder valuations are fundamentally impacted by market power, and by the
expectation of arbitrage (Garratt & Troger, 2006; Zheng, 2002)-- the 'trade' in cap-and-trade.
The degree of exercisable market power therefore becomes a key issue in the design of
property rights auctions, because it directly affects the valuation of market participants. Market
power has been consistently revealed in related electricity markets (Kahn et al. 2001; Wolfram,
1998). If the same firms that participate in those markets also participate in emissions markets,
the same disproportionate market composition may influence market performance in emissions
markets.
4
3. Why Market Power?: Background and Structure
The Regional Greenhouse Gas Initiative (RGGI) is the first-ever mandatory carbon cap-
and-trade program in the United States, and it heavily influences national and international
discourse on the development of carbon markets. The success or failure of RGGI, particularly in
terms of economic efficiency, is a vital pivot point on the pendulum of future Coasian policy
mechanisms. Although there have been previous tradeable property rights markets such as
Southern California's Regional Clean Air Incentives Market (RECLAIM), the US Acid Rain
Program, and the Virginia NOx Program, RGGI is the first market to target greenhouse gases,
which are far more difficult to mitigate or abate.
RGGI also serves as a model for larger programs because of its key institutional feature--
RGGI is the first cap-and-trade program to use a nearly 100 percent auction allocation method.
The initial distribution of property rights (allowances/permits) plays heavily into both the
efficiency and the equity of the emissions market (Hahn, 1984; Tietenberg, 2006). More
importantly, unmitigated market power in the distribution (auction), if exercised, can heavily bias
the efficiency of the secondary trading market, making price discovery difficult.
Unlike some other emissions markets, RGGI only covers the electricity sector;
transportation, agriculture and other GHG-emitting sectors are not covered entities. As a result,
market power is a larger concern because this sector is already heavily concentrated, and its
participants are the same natural monopoly firms that operate wholesale power generation.
3.1 Background
RGGI began as the pet project of former New York Republican Governor George Pataki,
who invited neighboring state governments to compact with New York in curtailing negative
effects of climate change in 2003. Today, ten east coast states are signatories: Connecticut,
5
Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey,
1
New York, Rhode
Island, and Vermont. Each RGGI state legislature has agreed to the terms of the compact, which
are outlined in the RGGI Model Rule, and each state has independently determined a reasonable
emissions cap (RGGI, 2008).
RGGI auctions include more than electricity sector firms, however, as banks and hedge
funds also participate. Liquidity was a concern from the inception of RGGI, stemming from the
significant concentration of market participants. Market developers therefore decided to permit
non-covered entities (banks and hedge funds) to participate in RGGI auctions, to serve two main
purposes. First, economic theory dictates that greater competition leads to more efficiency, and
thus market developers aimed to increase the quantity of market participants beyond the roughly
30 covered entities that would have otherwise participated by fiat. Second, despite the fact that
many RGGI participants are protected by rate-of-return regulation, electricity firms in particular
have a proclivity for seeking sufficient hedging instruments against economic risk. Participation
by banks and hedge funds can facilitate a much more robust derivatives market.
Auctions are held quarterly for two separate vintages of allowances. Current-term
vintages constitute the large majority of sales, and forward-term vintages, which are outside the
scope of this analysis because they constitute a very small part of the market. Allowances are
fully bankable (can be held for use in future compliance periods), but are not borrowable (which
would otherwise enable emissions in the present for allowance purchases in the future). See
Chapter 3 for a more detailed description of other RGGI program design features beyond market
composition, discussed here.
1
New Jersey has recently discontinued its participation in RGGI.
6
3.2 Structure
Substantive data on the efficiency and performance of RGGI market auctions are not
publicly available. RGGI is structured as a non-profit organization, which has successfully
shielded it from public disclosure legislation(Dormady, 2012). Member state governments, who
have access to the data, have denied public information requests from numerous sources. In
response to significant public inquiry, member governments have deemed all RGGI information
'trade secrets'. Because of this, a qualitative assessment of the efficiency and performance of
RGGI auctions cannot be given.
From publicly-available data, however, one key measure of market concentration can be
determined. The Herfindahl-Hirschman Index (HHI) is a measure of market concentration. It is
calculated as the sum of each firm's squared market share, or
{ }
2
1
N
i i
HHI
s
=
=
å
equivalently. It
ranges from a value of 10,000 in the case of pure monopoly, and approaches zero in the case of
an atomistic market. Appendix A provides a proof of the relationship between the HHI and the
Lerner Index, another popular market power measure, via the Structure-Conduct-Performance
relationship commonly used in Industrial Organization theory.
7
Table 1.1 Herfindahl-Hirschman Index (HHI) of Past RGGI Auctions
Auction #
Current Vintage HHI
of Market
Participants
Number of Winning
Bidders
Forward Vintage
HHI of Market
Participants
Number of Winning
Bidders
1 1,061 44 N/A N/A
2 1,203 46 N/A N/A
3 1,122 42 2,020 12
4 1,127 48 2,023 12
5 977 34 1,726 12
6 1,372 40 2,753 8
7 825 40 2,098 9
8 933 42 2,016 10
9 1,038 45 3,474 6
10 1,056 38 4,191 4
11 1,517 34 2,255 6
12 1,122 25 4,884 5
13 884 31 N/A N/A
14 2,332 38 N/A N/A
Available at: http://www.rggi.org/market/market_monitor. "N/A" indicates not applicable, as forward vintages were
not sold.
The HHI is a key measure of market concentration utilized by the US Department of
Justice for oversight of monopolies and trusts for purposes of merger review. Recent guidelines
suggest that any market with HHI exceeding 1,000 is moderately concentrated, and any market
with HHI exceeding 1,800 is highly concentrated.
2
Table 1.1 provides the HHI indices of recent
RGGI auctions for both current and forward vintages. Losing bidders are not considered market
participants in the estimation. Given that some scholarship suggests that auction design is less
relevant than the size of the market (Bernard, et al., 1998), the HHI figures from recent RGGI
auctions imply that market concentration can often decline despite declines in the number of
market participants. This is important because the HHI is a measure of market inequality. In a
market with all firms of equivalent share, the HHI would decrease to H=1/N, where N is the
number of firms in the market. Given the background and structure of these contemporary
emissions market auctions, it becomes clearer why market power should be evaluated.
2
In August, 2010, the US Department of Justice raised the HHI thresholds of merger guidelines in the wake of
financial market reforms, thus making a finding of market concentration more difficult. The current threshold for a
moderately competitive market is an HHI above 1,500.
8
4. The Two-Stage Uniform-Price Sealed Bid Auction
4.1 Auction Model
In order to show how market power is exercised in cap-and-trade auctions like the RGGI,
it is important to have a model, or system of equations, that characterizes the market. The form
of auction used in the RGGI is a uniform-price sealed bid auction. This section provides a
mathematical model of the first and second (arbitrage) stage of the RGGI tradeable property
rights market. The uniform-price sealed-bid auction is one in which bidders are permitted to
submit multiple bids. Unlike alternative auction designs, bidders submit bids of both price and
quantity. Quantities are then ordered by bid price, and winning quantities are those quantities that
are above the market-clearing price. The market-clearing price is determined when quantity
demanded equals the quantity for sale.
Let:
B º an Nx3 matrix containing elements
n
P ,
n
Q , and
n
F
It is rank ordered by price, such that
1 n n
P P
+
³ , and 1 n N £ £ .
C º an Nx1 matrix containing elements
n
C
n = The bid rank identifier.
n
P = The bid price associated with n.
n
F = The bidder identifier
3
associated with
n
P . It is an integer.
n
Q = The bid quantity associated with
n
P .
n
C = The cumulative bid quantity associated with
n
P .
T
Q = The total quantity (available) to be sold at auction.
3
This innovation of mine allows us to account for the possibility of multiple bids from a common bidder.
9
F
Õ = The secondary market profit function associated with bidder ‘F.’
0
P = The market-clearing (stop out) price.
'
F
Q = The quantity of allowances awarded to bidder ‘F.’
F
V = The secondary market resale value of allowances for bidder ‘F.’
F
V is exogenous.
, F a
Q = The quantity of allowances bidder ‘F’ must surrender for compliance
obligations.
, F a
Q is exogenous.
1 1 1
N N N
P Q F
B
P Q F
æ ö
ç ÷
º
ç ÷
ç ÷
è ø
M M M ;
,1 n n
P b = ,
,2 n n
Q b = , and
,3 n n
F b = .
1
N
C
C
C
æ ö
ç ÷
=
ç ÷
ç ÷
è ø
M ;
,1 n n
C c = .
1
n
n k
k
C Q
=
=
å
(1) { }
0 1 2
| &
n n T n T
P P C Q C Q
- -
= ³ <(2) Let n’ = n such that
n
P =
0
P . Note: n’ is the bid rank identifier associated with the
market-clearing price.
10
Let the profit function
4
for a chosen bidder be:
0
() '
F F F
V P Q  = - ×= D P·D Q(3) Profit maximization occurs when either the change in price or the change in quantity is
maximized. However, in these markets it is difficult for a monopsonist or oligopsonist to solely
determine the market-clearing price
0
P .
0
P is influenced by the degree of market power of the
monopsonist or oligopsonist. The monopsonist or oligopsonist can seek to maximize its quantity
of allowances '
F
Q in the auction, or set it strategically.
' 1
1
'
k
n
F k F F
k
Q Q  -
=
=
å
,(4) Where  is a version of the Kronecker Delta, such that;
0,
1,
m n
m n
m n
 ¹
ì ü
=
í ý
=
î þ
.
Equation 4 sums all of the successful bid quantities
n
Q , according to bidder identifier.
4.2 Strategy Under Market Power
The profit-seeking oligopsonist may seek to maximize profit by exercising market power
in one of three ways. First, the oligopsonist may seek to use the emissions market to exclude
rivals or raise their costs in a common product market (Rogerson, 1984; Salop et al., 1983; 1984;
1987; Williamson, 1968). This would be equivalent to a strategy of hoarding by buying as many
allowances as possible (essentially maximizing '
F
Q) and forcing firms in the competitive fringe
to face the choice of either reducing output or paying regulatory-enforced sanction costs for non-
4
A alternative profit function can include an exogenous term to account for those allowances a firm must surrender for
non-compliance. Note that by setting
, F a
Q exogenous, the model can account for regulated compliance entities as well
as banks and hedge funds who may be auction participants. The profit function in this case would be:
0 ,
()( ') F F F F a
V P Q Q  = - -
.
11
compliance. A second and related strategy for the oligopsonist is to buy as many allowances as
possible and profit from resale arbitrage (Garratt & Troger, 2006; McAfee & McMillan, 1987;
Zheng 2002). This second strategy depends crucially upon the rent-indifference point for fringe
firms. Oligopsonists can only resell allowances up to the point at which it is less-unprofitable for
fringe firms to either pay the sanction cost for non-compliance, or reduce output, or both. Both of
these hoarding strategies, although important aspects of market power, are outside of the scope of
this analysis because they rely crucially on exogenous factors specific to both the emissions
market and the common product market (electricity).
A third strategy, analyzed in this chapter, is for the oligopsonist to maximize profit by
strategically setting demand parameters (demand reduction) (Ausubel & Cramton, 2002; List &
Reiley, 2000; Webber, 1997; Wolfram, 1998). The oligopsonist can act as a Cournot firm and
strategically set its bid quantities into the emissions auction to maximize profit. The oligopsonist
would, in this case, exercise its influence over the market-clearing price ( 0
P) to simultaneously
push the market-clearing price downward while attempting to minimize its negative influence on
its own auction earnings ( '
F
Q). The exercise of market power through strategic demand
reduction requires the balancing of two competing market forces. Bidding for a quantity of
allowances too high can have a positive influence on the market-clearing price, thus making it
quite costly for the oligopsonist to gain allowances. Bidding for a quantity of allowances too low,
or at a price too low, can make it difficult for the oligopsonist to receive sufficient allowances
necessary to operate in the adjacent product market and maintain its current market share. The
analysis conducted in this chapter finds this optimization point for demand reduction under a
common set of market parameters, relevant to current cap-and-trade markets. It also includes a
set of sensitivity analyses of these results.
12
5. The Methodology of the Monte Carlo Approach
5.1 Parameters of Analysis
The methodology here expands upon the structure-conduct-performance relationship of
traditional Cournot industrial-organization analysis through the use of Monte Carlo repeated
simulation. As mentioned in Section 4, the Cournot oligopsonist seeks to maximize an a priori
profit function, under the assumption that competing firms hold their output fixed. Following
from Waterson (1984), the structure of the market (degree of market power and price elasticity of
demand for emissions allowances) impacts market performance (the profit-revenue ratio) via
market conduct.
Cournot market conduct in emissions market auctions manifests itself as strategic
demand reduction. Auctions in general are nothing more than a complex sorting methodology for
demand for a fixed quantity of goods. An oligopsonist that commands a significant share of the
market would constitute a larger share of aggregate demand, holding all other things equal.
Exhausting its budget to bid for every emissions allowance it could afford, would be in most
cases, highly unprofitable for the oligopsonistic firm.
5
Such behavior would shift the demand
curve upward and increase the market-clearing price for allowances, thus negatively impacting
the firm's profit.
The Cournot oligopsonist faces a balancing decision, between strategically reducing its
demand for allowances and ensuring sufficient allowance holdings to continue operating in the
adjacent production market (e.g., electricity generation). The latter is outside the scope of this
analysis, as it varies with exogenous market parameters and a complex legal-regulatory structure.
The strategic oligopsonist therefore, seeks a level of conduct (strategic demand reduction in this
5
This excludes asymmetric valuations due to the expectation of secondary market resale arbitrage or inter-temporal
arbitrage.
13
case) that maximizes profit. This Monte Carlo analysis pinpoints that optimum level of conduct
given parameter conditions in the underlying structure of the market.
5.2 The Simulation Environment
The method of analysis provided here offers a third approach to traditional methods of
analysis. Traditional game/auction theoretic analysis has many strengths, but it can be limited by
simplifying assumptions that inhibit its ability to generate probabilistic findings. Econometric
data analysis is often preferable, but is often limited by: a) the availability of data; and b) the
inability to alter exogenous structural parameters (e.g., market structure). The simulations
provided here, were conducted using Oligopsony 1.0, a Monte Carlo emissions auction
simulation environment designed by the author in C# in the .NET environment.
6
I designed Oligopsony 1.0 to carry out the operations of the two-stage model, provided in
Section 4, in repeated iterations. Although complex in design and fundamental logic, Oligopsony
1.0 must simultaneously carry out several operations. First, it must construct a static bid matrix B.
Each static bid matrix consists of three columns that each represent a separate operational class
within Oligopsony 1.0. Generating the bid matrix is the most challenging and computation-heavy
operation of the software.
The first column is an array of bid prices. All bid prices are drawn from a Gaussian
(standard normal) distribution with mean µ and standard deviation σ. The user directly supplies
the standard deviation, which is a double variable consisting of an integer and a fraction/decimal
to the hundredths place (e.g., dollars and cents). The mean is drawn from a uniform distribution
in which the user supplies the minimum and maximum values, also double variables. Oligopsony
1.0 redraws a unique µ for each bid and each bidder. The second column is an array of bid
6
.NET is Microsoft's development environment; its main competitor is Java.
14
quantities, which are drawn from a uniform distribution with user-supplied min and max values.
The third column is an array of bidder numbers
n
F matching each unique bid to its bidder. The
software permits multiple bids for each bidder, with a maximum of four unique bids per bidder.
The business logic of the software ensures that the entirety of each bidder's bids does not surpass
its class' (oligopsonist/fringe) quantity constraint.
Upon completion of the static bid matrix, Oligopsony 1.0 calculates winning and losing
bids, market clearing price
0
P , quantities awarded to each bidder, and profits for each bidder.
Profits in Oligopsony 1.0 are determined from a simple profit function, with a user-supplied
exogenous secondary market price
F
V , as given by equation 3 in Section 4.1.
7
This process
iterates to a user-supplied number of repetitions/auctions, and each value, from each iteration, is
stored in the user's system memory. Upon completion of a user-specified number of runs,
Oligopsony 1.0 supplies descriptive statistics (mean, min and max) for each of four summary
measures; market-clearing price, profit of oligopsonists, profit of fringe firms, and fringe loss (the
quantity of emissions allowances that firms in the competitive fringe bid for but did not receive).
More details on the software environment are provided in Appendix B.
7
This proxy profit calculation carries a modest approximation of actual firm-level profits under standard rate-of-return
regulation in markets with cost pass-thru.
15
6. Monte Carlo Simulation Results
The analysis here functionalizes parameters consistent with modern auction-based
emissions markets (RGGI) in large samples via Monte Carlo simulation to operationalize market
conduct through strategic demand reduction. Whereas a traditional Cournot oligopsonist seeks to
optimize quantity given an assumed static response behavior among firms in the competitive
fringe, this analysis allows both oligopsonists and firms in the competitive fringe to draw both
price and quantity parameters from parametric distributions ex ante, as detailed in Section 4.2.
Firms in the competitive fringe are not static in this analysis, although their behavior can
resemble static behavior asymptotically. Results of the Monte Carlo analysis are first provided
for a main case, and then provided for two sensitivity cases.
6.1 Main Case Monte Carlo Results
The main case attempts to approximate input parameters of contemporary auction-based
emissions markets as closely as possible. In this case, the simulation is set for a total of 50
bidders, where the number of oligopsonists is set from 1 to 3. The total quantity of allowances
for sale is 50,000; consistent with the quantity sold in most quarterly RGGI auctions. The
quantity constraint on competitive fringe firms is drawn from a uniform distribution ~U(1, 2500),
where the quantity sum of each fringe bidder's bids in any static auction cannot exceed 2,500
allowances. Both oligopsonist and fringe firms' bid prices are drawn from a Gaussian distribution
~Ɲ (µ ,
2
), where µ is drawn from ~U ($1.00, $2.00), and where  is set at $0.25. And, for
calculation of profit, the exogenous secondary market price
F
V is set at $2.00.
16
Figure 1.1 Oligopsonist Profit Under Demand Reduction
Figure 1.1 provides the main results of this analysis. Oligopsonist profit is plotted as a
function of market conduct. Market conduct is set incrementally across the x-axis, and represents
a single fixed quantity bid by the oligopsonist(s).
8
Each data point represents the expected value,
or average individual oligopsonist profit, for 2,000 Monte Carlo auction runs.
9
The horizontal
axis represents a fixed parameter of quantity, which is the bid specified by the oligopsonist(s) in
each simulation. Three plots are presented in Figure 1.1; monopsony, duopsony, and triopsony.
Clearly there exists a non-linearity between profit and the exercise of market power through
strategic demand reduction.
8
The x-axis of Figure 1.1 can be read from left to right as a decrease in the oligopsonists' bid, or equivalently, as an
increase in demand reduction.
9
Consider the duopsony case as a clarifying example. There are 46 data points plotted in Figure 1 for the duopsony
case profit curve, at a quantity interval of 500 emissions allowances that ranges from 2,500 to 25,000. Each of the 46
data points represents the average profit of 2,000 auctions. As such, the duopsony curve in Figure 1 was plotted from
92,000 individual auction simulations.
1,000
2,000
3,000
4,000
5,000
6,000
7,000
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 50,000
Mean Individual Oligopsonist Profit ($) Oligopsonist Bid (Demand) Reduction
Profit (1 Oligopsonist) Profit (2 Oligopsonists) Profit (3 Oligopsonists)
17
Each case is fitted with a simple second-degree polynomial trend line, each of which has
extremely significant fit (R
2
>0.95). A first-order condition can be derived from the fitted
function to provide the optimum point of strategic demand reduction for each case (e.g.,
monopsony). For the monopsony case, the first derivative of the fitted function is
() 0.50624 0.00001824
F monopoly
q x  ¢ = - , which, when set equal to zero yields a value ofH
27,750 allowances. This value can also be approximated by visual inspection of Figure 1.1. The
most profitable strategic bid for a monopsonist in a uniform-price sealed-bid auction for
emissions allowances with equivalent parameters, therefore, is not to purchase the entirety of the
market, but rather to reduce demand by approximately 45 percent [1-(27,750÷ 50,000)] of the
total quantity of allowances sold. The equivalent values for duopsony and triopsony areH 17,350
and 12,650 allowances, respectively.
An oligopsonist's ability to influence the market-clearing price is a key determinant of its
ability to successfully exercise market power. As the structure of the market changes (e.g.,
monopsony) the strength of an oligopsonistic firm's influence on price can change. This analysis
also provides insight into this aspect of market power. Figure 1.2 provides the relationship
between auction-clearing price and strategic demand reduction. The data points in Figure 1.2
provide the expected value, or average auction-clearing price for the same 2,000 Monte Carlo
auction runs. In Figure 1.2, a linear trend line function is fitted to the data points for exposition,
and each fitted function has similarly significant fit (R
2
>0.95).
By evaluating the slopes (in a manner similar to elasticities) of these linear functions, a
very close approximation of the relationship between demand reduction and auction-clearing
price can be determined. These slopes represent the impact on auction-clearing price of a
foregone unit of demand. A 5,000 allowance demand reduction (10 percent) by a single
oligopsonist, under equivalent parameters, yields a 2.4 cent reduction in mean auction-clearing
18
price. The equivalent values for duopsony and triopsony are 4.36 cents and 6.35 cents,
respectively.
This indicates that firms that exercise a significant portion of demand for emissions
allowances can parlay that market influence into significant reductions in auction-clearing price,
and thereby successfully suppress the value of tradeable allowances in a cap-and-trade market.
Furthermore, matching the optimum profit calculations from above with the price influences
determined here, allows us to determine a price-suppression indifference point. Another
interpretation of the optimal profit point is the point at which it is no longer profitable for an
oligopsonist to suppress the price and forego an additional emissions allowance.
Figure 1.2 Market Clearing Price Under Demand Reduction
1.35
1.40
1.45
1.50
1.55
1.60
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 50,000
Auction Clearing Price ($) Oligopsonist Bid (Demand) Reduction
P0 (1 Oligopsonist) P0 (2 Oligopsonists) P0 (3 Oligopsonists)
19
An interesting side effect of the influence of market power in emissions auctions is the
effect on the profit of firms in the competitive fringe. This Monte Carlo analysis also provides
insights into these effects of market power as well. The results provided in Figure 1.3 show the
relationship between demand reduction as exercised by the oligopsonist(s) and the expected profit
of firms in the competitive fringe. It shows that a kind of price-leadership effect occurs.
Although oligopsonistic firms exercise a significant portion of the market, smaller firms
in the competitive fringe piggyback on the price suppression of the oligopsonist(s). As the
market conduct of the oligopsonist(s) moves toward the optimum point of demand reduction, the
expected value of profit among fringe firms grows almost exponentially. In markets like the
RGGI, fringe firms that are small or independent electricity generators or participating
municipalities can actually profit significantly by the exercise of demand reduction by dominant
firms.
Figure 1.3 Mean Total Fringe Profit Under Demand Reduction
9,500
14,500
19,500
24,500
29,500
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 50,000
Mean Total Fringe Profit ($) Oligopsonist Bid (Demand) Reduction
Fringe Profit (1 Oligopsonist) Fringe Profit (2 Oligopsonists) Fringe Profit (3 Oligopsonists)
20
6.2 Sensitivity Analyses
The robustness of these findings was also subjected to sensitivity analyses of changes in
market structure. In electricity-only cap-and-trade markets like the RGGI, the quantity of firms
participating in the market is relatively fixed, as market participants are generally covered power
generating firms. Fluctuations in the number of bidders are affected by participation rates among
covered firms in quarterly auctions as well as participation by non-covered entities such as banks
and other speculative participants.
This analysis is not limited to electricity-only cap-and-trade markets, however, as the
simulations could apply to a number of auction-based markets or other cap-and-trade markets
moving toward auction-based allocation. The question that remains, therefore, is to what degree
are these findings robust to changes in the number of bidders? To evaluate this, a robustness
check was conducted by varying the number of fringe bidders by ±50 percent (25 and 75 bidders).
All other market parameters were retained.
The results in Figures 1.4 and 1.5 provide the relationship between oligopsonist profit
and strategic demand reduction. The 75 bidder case in many ways carries the same properties as
the 50 bidder case. The same non-linearities exist and the optimal points do not change much, as
the oligopsonist demand structure does not change much; however, the level of profit potential
decreases by nearly 30 percent across the board. As the market expands, the exercise of market
power becomes less profitable as demand-side competition leaves the oligopsonist with a smaller
share of the overall supply of allowances. Recall that each data point represents the mean profit
from 2,000 auction simulations, and as the size of the market increases, the same demand
reduction optimization strategy remains. However, the mean profit from that strategy shifts
downward, as shown by comparison between Figures 1.1 and 1.4.
21
Figure 1.4 Oligopsonist Profit Under Demand Reduction (75 Bidders) The 25 bidder case provides an interesting peculiarity. Although the level of oligopsonist
profit increases (shifts upward) in the expected direction, the strategic demand reduction
optimization point is influenced by the supply point. The expected value of oligopsonist profit
begins to shift upward as significant demand reductions occur toward the right hand side of
Figure 1.5. This is due to the fact that with few bidders and large demand reductions by the
oligopsonist, demand from both the oligopsonist and the fringe is insufficient to match the overall
supply of allowances for sale, and residual supply exists. In uniform-price auctions in which
550
1,050
1,550
2,050
2,550
3,050
3,550
4,050
4,550
5,050
0 10000 20000 30000 40000 50000
Mean Individual Oligopsonist Profit ($) Oligopsonist Bid (Demand) Reduction
75 Profit (1 Oligopsonist) 75 Profit (2 Oligopsonists) 75 Profit (3 Oligopsonists)
22
residual supply exists, the auction clearing price converges to zero.
10
As the auction clearing
price converges to zero, the profit peaks.
11
And, as a result the second-degree polynomial trend
line traces the outcome with less precision.
Figure 1.5 Oligopsonist Profit Under Demand Reduction (25 Bidders) 10
Often in uniform-price auctions, a reserve price can act as a constraint on auction clearing prices that would
otherwise converge to zero. In recent RGGI auctions, for example, the auction clearing price has converged to the
reserve price.
11
Because the profit function utilized here (equation 3 in section 4.1) is a basic one, there is no direct mapping
between auction clearing prices and secondary market prices, as the secondary market price is exogenous by
simplification. The residual supply effect shown in Figure 1.5 would be smaller in magnitude in markets in which
changes in the secondary market reflect changes in the primary allocation auction.
3,000
5,000
7,000
9,000
11,000
13,000
15,000
0 10000 20000 30000 40000 50000
Mean Individual Oligopsonist Profit ($) Oligopsonist Bid (Demand) Reduction
25 Profit (1 Oligopsonist) 25 Profit (2 Oligopsonists) 25 Profit (3 Oligopsonists)
23
Figure 1.6 Market Clearing Price Under Demand Reduction (75 Bidders) The profitability of this residual supply effect is driven quite strongly by the price
responsiveness of the auction to changes in strategic demand reduction by the oligopsonist.
Figures 1.6 and 1.7 present the results of the sensitivity simulations for auction clearing price.
Figure 1.7 clearly shows how quickly the mean clearing price can dip downward as strategic
demand reduction by the oligopsonist begins to increase the likelihood of residual supply. Put
another way, the mean auction clearing price becomes significantly more responsive to strategic
demand reduction as that reduction encroaches upon the point at which residual supply is more
likely.
1.48
1.5
1.52
1.54
1.56
1.58
1.6
1.62
1.64
0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
Auction Clearing Price ($) Oligopsonist Bid (Demand) Reduction
75 P0 (1 Oligopsonist) 75 P0 (2 Oligopsonists) 75 P0 (3 Oligopsonists)
24
Figure 1.7 Market Clearing Price Under Demand Reduction (25 Bidders) A comparison of these effects across each of the three cases reveals the degree to which
the impact of market power on auction clearing price is affected by changes in the overall
structure of the market. Table 1.2 provides this. The responsiveness of auction clearing price to
strategic demand reduction is measured as the slope, or elasticity, of the fitted linear relationship
between them.
12
Table 1.2 Auction Price Responsiveness to Demand Reduction
25 Bidders 50 Bidders 75 Bidders
1 Oligopsonist
14.06µ 2.44µ 1.67µ
2 Oligopsonists
29.68µ 4.36µ 3.33µ
3 Oligopsonists
45.51µ 6.35µ 4.53µ
Slopes reflect a 10 percent (5,000 allowance) change in oligopsonist strategic demand reduction.
12
The significance of the linear fit was good (R
2
>0.95) for the 50 and 75 bidder cases. For the 25 bidder case the fit
was less robust(R
2
>0.74) due to the price impact of residual supply.
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 50,000
Auction Clearing Price ($) Oligopsonist Bid (Demand) Reduction
25 P0 (1 Oligopsonist) 25 P0 (2 Oligopsonists) 25 P0 (3 Oligopsonists)
25
As mentioned above, in the simulations with 50 total bidders, a 10 percent (5,000
allowance) demand reduction by a monopsonist, under equivalent parameters, yields a 2.44 cent
reduction in mean auction-clearing price. The equivalent values for duopsony and triopsony are
4.36 cents and 6.35 cents, respectively. In the 75 bidder simulations, a 10 percent demand
reduction by a monopsonist yields a 1.67 cent reduction in mean auction clearing price. And, in
the 25 bidder simulations, a 10 percent demand reduction by a monopsonist yields a 14.06 cent
reduction in mean auction clearing price. Remaining cases are presented in Table 1.2.
Figure 1.8 Mean Total Fringe Profit Under Demand Reduction (75 Bidders) 10,000
12,000
14,000
16,000
18,000
20,000
22,000
24,000
0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
Mean Total Fringe Profit ($) Oligopsonist Bid (Demand) Reduction
75 Fringe Profit (1 Oligopsonist) 75 Fringe Profit (2 Oligopsonists) 75 Fringe Profit (3 Oligopsonists)
26
The mean total of fringe profit also reveals the same profit impacts for firms in the
competitive fringe in both sensitivity cases. The profit of fringe firms is directly impacted by
clearing price reductions as a result of strategic demand reduction by the oligopsonist(s). The
same sort of price leadership effect occurs. The rate of change of mean total fringe profit
increases as the size of the market decreases. This rate of change becomes dramatic as the
probability of residual supply becomes more likely. This is shown in Figure 1.9 by the upswing
in the mean total fringe profit curve.
Figure 1.9 Mean Total Fringe Profit Under Demand Reduction (25 Bidders) 6.3 Kernel Density Analysis
An illustrative method for evaluating the probabilistic nature of events in a Monte Carlo
study is the use of density analysis. Because the data points in the previous section each represent
the expected value, or mean, of 2,000 Monte Carlo runs, they provide a strong summary of the
11,000
21,000
31,000
41,000
51,000
61,000
71,000
0 10000 20000 30000 40000 50000 60000
Mean Total Fringe Profit ($) Oligopsonist Bid (Demand) Reduction
25 Fringe Profit (1 Oligopsonist) 25 Fringe Profit (2 Oligopsonists) 25 Fringe Profit (3 Oligopsonists)
27
events that occurred throughout those thousands of auctions. Summary measures, however, do
not provide further detail into the range of events that populated those summary measures.
Kernel density plots,
13
as provided below, give probabilistic detail to the results of the
main case Monte Carlo analysis at the respective optimum oligopsonist profit points. These
figures show the density, or likelihood, in a manner similar to a histogram, of the events that
populated the summary measures. The probability of auction outcomes in these simulations is
not uniformly distributed, and the distribution of those outcomes changes as the structure of the
market and the conduct of market participants changes.
Figure 1.10 Kernel Density Plot of Individual Oligopsonist Profit
13
Kernel density plots in Section 6.3 each use a Gaussian kernel function.
0 .00005 .0001 .00015 .0002
K-Density of Profit
-10000 0 10000 20000 30000
Individual Oligopsonist Profit ($) kdensity Monopsony kdensity Duopsony
kdensity Triopsony
28
Figure 1.10 provides kernel density plots of the main case results at the respective
optimum profit points for each of the three oligopsony scenarios. The spike in the distributions
about the zero profit point reflects the likelihood of oligopsonist(s)' bids encroaching upon or
falling below the auction-clearing price. Bids falling below the clearing price return zero
commodities (e.g., emissions allowances) to the bidder, and thus zero profit. The right-skewed
tails of the distributions indicate the strong likelihood of actual oligopsonist profit far exceeding
the mean profit points reported in section 6.1. The spike in the tail of the monopsonist profit
density function, for example, indicates that out of 2,000 auctions, the monopsonist actually
yielded profits exceeding $20,000 frequently.
Figure 1.11 Kernel Density Plot of Auction Clearing Price
0 .5 1 1.5
K-Density of Auction Clearing Price
.5 1 1.5 2 2.5
Auction Clearing Price ($) kdensity Monopsony kdensity Duopsony
kdensity Triopsony
29
The density plots provided in Figure 1.11 show the probabilistic range of auction clearing
prices at the respective optimum points. Recall that throughout these Monte Carlo simulations,
constraints on the distribution of bid prices for both bidder types are equivalent. And, recall that
Oligopsony 1.0 draws a unique mean bid price for each bidder's unique bid, and allows for the
possibility of multiple bids from the same bidder in each auction. As a result, the density plots
provided in Figure 1.11 are similar in shape to a platykurtic normal distribution or a tall uniform
distribution. Thus, they indicate that auction clearing prices slightly above and below $1.50 are
nearly equally likely.
Figure 1.12 Kernel Density Plot of Total Fringe Profit
0 .00002 .00004 .00006
K-Density of Fringe Profit
0 20000 40000 60000
Total Fringe Profit ($) kdensity Monopsony kdensity Duopsony
kdensity Triopsony
30
The figure also shows the weak likelihood of auction clearing prices rising above the
secondary market price. The small likelihood of those events produces the few cases of negative
profit shown in Figure 1.10. The probabilistic distribution of total profit for firms in the
competitive fringe also indicates events of high fringe profit at a higher rate for the monopsony
scenario than other scenarios. The right-skewed tails of the distributions indicate that the
likelihood of increasing total fringe profit occurs beyond the mean point, but at a declining rate.
7. Implications and Conclusions
This chapter has evaluated the extent to which one aspect of market failure, market power,
can impact the range of possible outcomes in a Coasian transferable property rights market with
auction-based allocation. It has applied a parameter set roughly equivalent to a current operating
cap-and-trade program in the United States, and has evaluated the range of possible outcomes
under three market power scenarios (monopsony, duopsony, and triopsony), with a set of
sensitivity analyses evaluating robustness to changes in market size.
The analysis has shown that structural tradeoffs exist between positive demand-side
influences on price and negative influences on profit from bids that yield firms insufficient
quantities of goods (e.g., emissions allowances). Because non-linearities exist between these
tradeoffs, optimal bidding strategies of strategic demand reduction were determined. As well,
because the analysis was conducted using Monte Carlo simulation, the probabilistic distribution
of these results was also provided.
Consistent with the structure-conduct-performance paradigm in traditional industrial
organization theory, this analysis has shown that market conduct, as exercised through strategic
demand reduction, can lead to performance changes in the auction market. Moreover, it has
31
determined the specific marginal rate of influence, or slope, of changes in market conduct on
expected auction clearing price for each individual scenario.
The influence of market power was also considered from the standpoint of firms in the
competitive fringe, who participate in the market with weaker price influence. The results
suggest that the exercise of strategic demand reduction by dominant firms is not entirely injurious
to fringe firms, as a sort of price-leadership effect can occur. That is, as strategic demand
reductions by oligopsonistic firms suppress market prices, fringe firms also benefit from that
price suppression as they acquire goods (e.g., emissions allowances) at a lower cost.
As national and international discourse on climate change mitigation looks to market and
auction-based solutions for further policy implementation, it is important for consideration to be
given to the nature of market structure and the impact it will have on efficiency and performance.
If participants in emissions markets are to be the same natural monopoly firms that operate within
electricity markets, then the design and operation of emissions markets should reflect the same
degree of deference to considerations of strategic behavior and market concentration.
32
Chapter 2: An Experimental Approach
1. Introduction
How will emerging emissions markets function within the context of today’s deregulated
electricity markets? Most deregulated electricity markets can be characterized by a handful of
generators bidding in regional electricity dispatch markets to supply power into a grid or regional
market (Brennan, 2007; Hunt, 2002; Joskow & Schmalensee, 1983; Joskow, 1998; Newberry,
1999). These markets can be characterized by a variety of inherent market failures, most notably
imperfect competition and market power, which require a variety of regulatory mechanisms and
policy corrections for efficiency. Concurrently, these same firms participate in regional emissions
markets that do not have equivalent regulatory mechanisms or controls.
Historically, scholars have debated performance in emissions markets given a variety of
market designs and allocation regimes (Arimura, 2002; Cramton & Kerr, 2002; Ellerman,
Convery & De Perthuis, 2010; Ellerman et al., 2000; Joskow, Schmalensee & Bailey, 1998;
Tietenberg, 2006). Until recently, much of the debate has been centered on the issue of allocative
efficiency under imperfect competition (Chavez & Stranlund, 2003; Hahn, 1984; Misiolek &
Elder, 1989; Sartzetakis, 1997; Van Egteren & Weber, 1996). Stemming from Hahn’s (1984) seminal paper, this literature has made significant headway into understanding the
interrelationships that exist between emissions markets and imperfect competition in adjacent
product markets.
More recent scholarship has appropriately looked to auction theory that stemmed from
the pioneers of spectrum auction design (Milgrom, 2004). Today, contemporary Coasian
33
emissions markets
14
make use of auctions for allocation of tradable emissions permits, which
overcome a variety of shortcomings inherent to emissions market design under alternative
allocation methodologies such as grandfathering. However, the efficiency of contemporary
auction-based emissions markets under the Northeast’s Regional Greenhouse Gas Initiative
(RGGI) has been met with mixed results at best (RGGI Annual Report, 2010). Auction clearing
prices have been consistently below pricing expectations, with prices clearing at the price floor
(reserve price) for nearly the entirety of the program, and the secondary trading market has been
almost completely inactive. However, much of this can be attributed to factors beyond imperfect
competition, such as economic downturn and weak demand, over-allocation and the availability
of low-cost emission reductions techniques.
Yet key linkages exist between extant electricity markets and emissions markets
(Downward, 2010; Murray, 2009). Past scholarship recognizes the implicit role these energy-
emissions linkages played during the California crisis (Joskow & Kahn, 2002; Kolstad & Wolak,
2003). But it is unclear how these linkages will affect prices in both markets under auction-based
carbon regimes and imperfect competition in both markets. This chapter hypothesizes that when
firms who are dominant producers in electricity markets are also allowed to be dominant buyers
in adjacent auction-based carbon markets, these firms will collusively act to inflate electricity
prices and suppress emissions prices. Therefore, this chapter seeks to test the degree to which
these energy-emissions linkages can be operationalized by dominant firms.
This chapter provides a laboratory experiment of a joint energy-emissions market under a
stylized market power case with a variety of real-world market features, including stochastic
electricity demand, collusion among dominant firms, inter-temporal dynamics, a tightening
14
The Northeast’s Regional Greenhouse Gas Initiative (RGGI) makes use of a nearly 100% auction allocation, and
California’s AB32 Cap-and-trade program makes use of an auction-based allocation method in which the proportion of
allowances allocated by auction increases across time. The interested reader is encouraged to see the RGGI Model
Rule and the California AB32 Final Regulation Order.
34
emissions cap and resale arbitrage. Past experiments have been used in a similar vein to look at
market power in emissions markets (Godby, 2000), market power in auction markets in general
(Holt, 1989), emissions market allocation methodologies (Wrake et al., 2008), collusion in
emissions auctions (Burtraw et al., 2008) real-world features of stochastic emissions and
allowance banking (Cason & Gangadharan, 2006), and enforcement and compliance under a
dynamic market (Stranlund et al., 2011). This chapter adds to this important line of scholarship
by internalizing greater real-world market dynamics than any prior experimental analysis.
Furthermore, this chapter adds to this literature by being the first of its kind to analyze a joint
energy-emissions market.
This chapter provides details of the experimental methodology and design in Sections 2
and 3 respectively. Section 4 provides details on strategy and profit under market power in the
stylized dynamic energy-emissions market. Results of the experiment are provided in Section 5.
Conclusions and implications for public policy are provided in Section 6.
2. Experimental Method
2.1 Recruitment and Pilot Sessions
The experiments were conducted at the University of Southern California Law School.
Pilot experiments were conducted at the USC School of Policy, Planning and Development
(SPPD), USC Department of Economics, and USC Law School. Subject recruitment occurred via
both electronic and physical contact with students. Subjects were contacted through student
listservs both through the Department of Economics and the School of Policy, Planning and
Development, as well as through organizational listservs including the Indian Students
Association and fraternal honor societies. Subjects were also contacted directly, as the author
attended a variety of large introductory economics courses and spoke on the topic of experimental
35
economics. And finally, a small number of subjects were recruited via advertisement signs
posted in departmental front offices. The majority of subject recruitment came from majors and
fields such as business, economics and policy.
Potential subjects were queried on prior experimental participation to ensure that no
biasing effects of past experimental participation would occur, such as deception. Only a very
small number of participants had prior experimental participation, and of those, two subjects were
excluded from participation due to participation in psychology experiments involving deception.
Of those few students who participated in past experiments, the queries showed that those
experiments were mainly market research, namely a large housing survey conducted in the USC
Marshall School of Business Administration.
Subjects were queried on their scheduling availabilities to participate in a variety of
session times. Within their scheduling constraints, subjects were randomly matched to
experimental sessions. To ensure further randomization, treatments were randomly assigned to
scheduled sessions after subjects had been scheduled. Of the six experimental sessions conducted,
the order of treatments was 1) control, 2) treatment, 3) treatment, 4) control, 5) treatment, and 6) control. And to ensure further randomization, within the treatment group, subject class/type was
also randomized.
2.2 Experiment Instructions and Remuneration
Experiment sessions lasted 3 hours. During the first hour of the experiment, subjects
received instructions on participating in the experiment. This consisted of a 60 minute MP4
video that covered the experiment, client software, and a very brief set of example scenarios to
illustrate profit calculation and resale in the two-stage game. Experiment videos were recorded
using Jing Pro (TechSmith Inc., Okemo, MI), and the text transcript of the instructional portion of
36
the videos is provided in Appendix C. Subjects also received three handouts. These included the
subject payment form, market diagram, and a single page summary of the experiment setup.
Subjects each received a $10 show up payment. Session payments had a 0.1 conversion
rate. This means that in addition to the show up payment, subjects were incentivized to perform
by keeping ten cents for every dollar they made in the session. Because the operational sessions
lasted two hours, and because the experiment was designed to test market power under collusion,
the potential for significant earnings existed under sustained cooperation. A safety valve was put
in place to guard against overcompensation, which restricted individual session payouts to $50,
for a total earnings cap of $60. This cap turned out to provide a constraint on only one subject;
one individual had payout total of $64. Subjects were informed of this constraint during
recruitment and during the subject instructions phase.
Another constraint was put in place to guard against subject losses. Each period of the
experiment consisted of two rounds. In each round, the subject could lose money; however, if the
total two-round, or period profit was negative, the subject would receive a payout of zero dollars
rather than receive a negative payout. This was particularly important for three reasons. First,
negative losses could occur during early periods of learning. And, negative losses could occur if
the subject was randomly assigned to a class without market power, and with high production
costs under a competitive market. Second, because there were a large number of periods in any
single session (approx. 20-30), we believe that early losses could have potentially dissuaded
performance throughout the session, if the subjects were to become discouraged by negative
earnings early on. And third, Federal experimentation rules regarding negative financial losses
by human subjects are prohibitive.
37
2.3 Software
The software environment utilized for this analysis was the Zurich Toolbox for Ready-
made Economic Experiments (ZTREE)(Fischbacher, 2007) and its companion client application
Z-Leaf. Coding in ZTREE requires mastery of the University of Zurich’s proprietary software
language, which in many ways is similar to older forms of Java. Although ZTREE is not an
Object-oriented Language (OOL), its structure utilizes tables, in place of objects, that maintain
much of the same functionalization as OOL.
Although ZTREE does not possess many higher level commands like more advanced
programming languages (e.g., C#, Java, Python), it does allow for basic core functions requisite
in most experimental economics applications (e.g., matrix manipulation, strings, arrays, loops).
ZTREE was selected against other programming environments because of its ease of management
in experimental economics applications, particularly for the server-side of the software. Because
of the complexity of this experimental application, however, the coding included more than 50
table programs (similar to objects) and spanned more than 200 pages of text in total.
3. Experiment Design
3.1 Market Representation
The experiment is designed to simulate a simultaneous electricity and emissions
allocation auction with inter-temporal dynamics. There are a variety of market designs for both
sets of markets, and no single experiment could simulate the structural dynamics that exist in each
of them. Given this, the design of this experiment is intended to provide an illustrative case that
is broadly representative of the dynamics given by these market designs in general. More directly,
there are specific design structures, regulatory constraints, holding requirements, etc., that exist in
some markets but not in others, that are excluded from this experimental design.
38
Emissions allocation auctions are the preferred design of contemporary emissions
allocation policy. Historically, tradeable property rights (i.e., allowances, permits, credits) have
been distributed directly (grandfathered) by the regulatory or planning oversight agency to
covered entities, and this has led to inefficient market signals, political misallocation and barriers
to entry (see Ellerman et al., 2010). More contemporary market designs have called for direct
allocation via auctioning. The preferred format of auction chosen by policymakers in extant and
future-planned markets is the uniform-price sealed-bid auction.
15
The electricity auction component of this experiment is intended to translate to a more
generalized electricity production auction. Electricity generators, such as independent power
producers or cooperatives, compete in auctions to sell electricity to firms with service obligations
such as regulated utilities or municipalities. The sale of electricity into a centralized regional grid
or network occurs throughout a variety of regions in the United States and elsewhere, and the
electricity auction component of this experiment is intended to simulate these production auctions
in a general, yet representative sense. The electricity component of this experiment could also be
broadly representative of capacity/resource adequacy auctions held by many regional grid
corporations and state governments. In those auctions, generators offer supply-ready and
dispatchable power to grid operators on a continual and contractual basis in exchange for a fixed
capacity payment.
15
The uniform-price sealed-bid auction is a non-discriminatory auction design. This means that auction winners are
not required to pay their bid price, but rather pay a uniform clearing price. A mathematical model of a multi stage
uniform-price sealed-bid auction is provided in Chapter 1. The interested reader is also encouraged to see Krishna
(2004). This auction format is utilized in RGGI, AB32 auctions, and was the preferred format of the Waxman-Markey
and Kerry-Boxer national cap-and-trade proposals.
39
3.2 Simultaneous Auctions
The experimental design requires subjects to participate in both the electricity and
emissions markets simultaneously. In the electricity auction, subjects are given a pre-specified
portfolio of power that they can sell into the market. Their portfolio consists of generation assets
(e.g., generators) that they bid to sell into the grid separately. Bids specify the minimum payment
that they would like to receive to sell power from that asset into the power market. The market
clearing price is set when price-ranked supply meets demand. Demand for power is set
exogenously, and all subjects are made aware of the current period’s demand for power at the
beginning of each auctioning period. This equates to a demand forecast in a forward electricity
auction or capacity market.
The emissions market is slightly more complex. Each subject is given a pre-specified
budget, depending upon treatment group, with which to purchase emissions allowances. Subjects
are required to hold emissions allowances on a 1:1 ratio for simplicity. That is, subjects are
required to hold 1 emissions allowance for each unit of power they supply into the power market.
The supply of allowances is fixed and decreasing, to simulate a cap-and-trade market with a
declining, or tightening cap. All subjects are made aware of the supply of allowances at the
beginning of each auctioning period. Unlike bids into the electricity market component that
specify only a price for a fixed unitary supply of power, bids into the emissions market specify
both a price and quantity. That is, they specify how many allowances they would like to purchase,
and the maximum marginal price that they are willing to pay.
3.3 Inter-temporal (Multi-Round) Dynamics-- Banking
The experimental simulations consist of multiple periods. Within each period, there are
two rounds of bidding. In each round, both the electricity and emissions market auctions are held.
40
In the first round, approximately 30 percent more emissions allowances are sold than in the
second. This simulates an emissions market with a declining, or tightening cap across time. A
flow chart of the intra-period experiment is provided in Figure 2.1.
Compliance periods are simulated on a period-by-period basis in this experimental design.
Subjects are not required to surrender, or turn in, their allowances until the end of the second
period. That is, just as in operating emissions markets, participants can run a deficit or a surplus
until the end of the compliance period, at which point they must turn in a minimum quantity of
allowances consistent with their emissions. The experimental server application maintains
separate accounting for each subject, across rounds, and banking behavior is simulated by
allowances held by a subject from the first round into the second round. Throughout this
experiment, banking behavior was significant. Borrowing, or the act of counting future purchases
of allowances ahead of time for present emissions, is not explicitly simulated in this design.
16
3.4 Demand for Energy
The demand for electricity is fixed and exogenous within any period (2 rounds), and
stochastic on a period-by-period basis. That is, akin to true electricity markets, this design
simulates the stochastic nature of electricity demand patterns to a degree. Demand in this
experiment is set at a value that is high, intermediate, or low. High demand simulates behavior
when electricity markets are resource-constrained, such as in the case of peak electricity demand.
Low demand simulates behavior when the demand for electricity resources on a regional market
is low, and only the most efficient generators win production rights.
16
The economic rationale for banking (and borrowing) is well established in the literature. Stevens and Rose (2002) provide an analysis of inter-temporal dynamics in a permit trading model under a fixed annual emissions cap. They
find that banking is an explicit strategy utilized by firms when the rate of expected future abatement cost increases
(due, for example, to having to move up the marginal abatement cost curve in the face of more stringent emission
requirements under economic growth and a fixed emissions cap) exceeds the discount rate.
41
Figure 2.1 Energy-Emissions Experiment Flow Chart
42
3.5 Generation Assets -- Portfolios
This experiment internalizes real-world differences that exist in electricity markets of
different portfolios of generation assets. That is, in operating electricity markets, some firms
have generation assets with low production and operating costs, while others have high
production costs. In this experiment, there are two forms of generation assets; low cost
generators and high cost generators. Throughout each treatment, low cost generators incur a
production cost of $1, and high cost generators incur a production cost of $2. And, in all
treatments, there were a total of 10 of each generation asset, for a total production capacity of 20
generation assets.
3.6 Allowance Market with Resale Arbitrage
Whereas in operating electricity markets under emissions market regimes, the acquisition
cost of emissions allowances raises the production cost of electricity. The same dynamic is
analyzed in this experiment. All generation assets are required to acquire one emissions
allowance for every unit of generation they produce.
In total, there are 35 emissions allowances sold in any period. Within each period, there
are two rounds, and within each round, a portion of the total quantity of emissions licenses is
auctioned. In all treatments, there are a total of 20 emissions allowances auctioned in the first
round. And, there are a total of 15 allowances auctioned in the second round. However, that
quantity can exceed 15, if there are either unsold allowances from the first round that are carried
over to the second round, or if there are allowances sold by other subjects. These quantities
effectively permit the simulation of a peak electricity market with an emissions allowance
shortage under market power.
43
Resale, which exists in most tradeable allowance markets, constitutes the “trade” in cap-
and-trade. This experiment institutes a simplified version of this trading phase. Subjects, who
purchase allowances in the first round, can sell them into the auction in the second round to other
subjects who need those allowances. These subjects do this with the expectation of earning a
higher return on the allowances in the second round, when the quantity of allowances sold
tightens. One caveat is that this experiment institutes the business rule of public priority. That is,
no subject can resell any allowances in the emissions auction until the government’s allowances
are sold first.
3.7 Treatments -- Market Power
Within this experiment there is one control group and one treatment group. The control
group simulates a competitive market where all firms have equal market share, equal budget, and
no collusion takes place. The experimental, or treatment group simulates a market with market
power and allowed collusion among oligopolists/oligopsonists. Table 2.1 provides a summary of
the distinctions between these treatments.
Table 2.1 Summary of Treatment Effects
Parameter Control Group
Number of Subjects 10
Number of Generators 20
Supply of Emissions Allowances Sold (Round 1) 20
Supply of Emissions Allowances Sold (Round 2) 15 + Resale + Unsold (Round 1) Total 2-round Quantity of Allowances 35
Stochastic Demand (Low, Medium & Peak) Yes
Banking Yes
Market Power No
Collusion Among Oligopolists Not Permitted
Types All Equal Fringe (Type A) Fringe (Type B) Oligopolist (Type C) Number of Subjects within Type 10 4 4 2
Generation Asset (Low MC) Quantity 1 1 0 3
Generation Asset (High MC) Quantity 1 0 1 3
Opening Budget $20 $10 $10 $60
Treatment Group
Yes
Yes
Yes
Permitted
10
20
20
15 + Resale + Unsold (Round 1) 35
44
In the control group, there are a total of 10 equal subjects, and each subject has a
portfolio of two generation assets; 1 high cost generator and 1 low cost generator. Each subject
begins with a budget of $20. In the treatment group there are also a total of 10 subjects.
However, two of those subjects are oligopolists/oligopsonists, and the remaining 8 have
diminished price influence. Every period, type assignment is set randomly. The oligopolists each
have a portfolio of six generation assets; 3 high cost generators and 3 low cost generators. The
remaining subjects each have one generator, either a high or a low cost generator, also assigned
randomly. Oligopolists each begin with an opening budget of $60, and the remaining subjects
each begin with an opening budget of $10. Budget assignment in both treatment groups equates
to $10 per generation asset. The two subjects who are oligopolists also have access to a chat
room to cooperate or collude with each other.
3.8 Context Sanitization
To avoid biasing the experimental results from confounding factors related to the
electricity or emissions market context, market-specific terminology was removed from the
experiment altogether. Instead of using phrases such as “high cost generator,” “high emissions
generator,” “low cost generator,” or “low emissions generator,” the experiment used the term
“Product X” and “Product Y.” And instead of using the phrase “Emissions Allowance” or
“Emissions Permit”, the subjects saw the phrase “License.” The sanitization was sufficiently
effective that several interested subjects inquired of the experimenter after the experimental
sessions as to the context of the experiment.
Similarly, in the treatment group, no reference to “market power” was provided to
subjects. Rather, subjects were provided with classification of types. Subjects were assigned to
either Type A, Type B or Type C. Subjects of Type A were those subjects who received a single
45
high cost generator, and subjects of Type B were those subjects who received a single low cost
generator. Furthermore, subjects of Type C were those subjects who were
oligopolists/oligopsonists, who received a portfolio of 6 generators, a correspondingly larger
budget, and had access to coordinate with the other Type C subject via a chat room.
The term “resale” was used in the experiment; however, the terms “arbitrage” or any
derivative of the term “speculate” were not used. Similarly, for the treatment group, no reference
to “collusion” or “cooperation” was made.
4. Strategy in the Simultaneous Two-Stage Market
4.1 Profit in the Simultaneous Two-Round Game
Within a joint electricity and emissions market, there are a variety of strategies that each
type of market participant can pursue. The design phase of this experiment considered symmetric
equilibria in determining the most appropriate structural design; those equilibrium strategies
under which profit is maximized when players play an equivalent strategy. Strategy is critically
shaped by constraints inherent to the conditions imposed by a given subject’s type and to
parameters specific to the market itself.
For the control group (without market power or collusion), there is only one type. All
subjects are competitive bidders of equal market share. For the treatment group, differences in
subject endowment determine constraints which shape strategy. In the treatment group, a subject
who is randomly assigned a type without market power is of type Fringe. And, two subjects are
randomly selected each period to be of type Oligopolist; which consists of a larger opening
46
budget, a larger endowment of electricity generation assets, and the ability to communicate with
the other oligopolist (collusion).
17
Optimal symmetric strategies are considered those that yield the largest overall period
(two-round) profit. Under this stylized multi-round, simultaneous energy-emissions market,
profit is given by a linear additive function, as given by Equation 1. Profit in equation 1 is given
by five separate components. First, production revenue is generated when a subject sells
electricity into the electricity market in round 1 or 2. It is given by the sum over two rounds of
the price quantity pair
2
1
et eit
t
P q
=
å
, where the price of electricity
et
P is endogenous, and represents
the market-clearing price of the uniform-price sealed bid electricity market auction. Chapter 1
provides a simple two-stage model of an equivalent uniform-price sealed bid auction market.
et
P
is a uniform, or common, price and therefore does not carry the subscript i. The quantity of
electricity that is sold by subject i in either round 1 or 2 ( eit
q), is similarly endogenous, and is
determined by business rules inherent to the uniform-price auction. As provided in Chapter 1,
eit
q is the entirety of subject i’s bid quantity if that bid is below the clearing price, and takes an
inframarginal value if that bid sets the clearing price. Simply put, winning bids sell all of the
electricity they bid to produce, and only those bids that are on the margin and set the clearing
price may be split and result in production of a fraction of the electricity offered into the market.
18
17
The Oligopolist-Fringe model of emissions markets was originally developed by Hahn (1984), and later extended by
others to include a fringe type that has more moderate price influence as provided here.
18
The business rules of this experiment simplified this inframarginal quantity issue by restricting bids to unity. That is,
bids were placed separately for each single unit of production capacity, and all ties were decided randomly. This
precluded the possibility that subjects would sell a fraction of energy into the energy market.
47
Let:
et
P = price of electricity(auction clearing price) at auction t
mt
P =price of emissions allowances (auction clearing price) at auction t
eit
q =quantity of electricity sold by firm i into auction t
mit
q = quantity of emissions allowances sold by firm i into auction t
pi
p = cost to firm i of producing a single unit of electricity(exclusive of emissions cost) wit
q = quantity of emissions allowances won/purchased by firm i at auction t
N
P = the non-compliance penalty per unit
ni
q = quantity of shortage of emissions allowances (shortage quantity) 2 2 2
2 2
1 1 , 1
high
i et eit m mi pi eit mt wit N ni
t t c low t
P q p q p q p q P q
= = =
Õ = + - - -
å åå å
(1) Where
2 2
2
1 1
2 2
2
1 1
0,
eit wit mi
t t
ni
eit wit mi
t t
q q q
q
q q q
= =
= =
ì ü
£ -
ï ï
ï ï
=
í ý
ï ï
- +
ï ï
î þ
å å
å å
Second, production costs are those costs that are incurred by the electricity producer for
producing electricity, exclusive of the internalization of emissions costs. For electricity producers
in operating markets, these include such costs as capital outlays, fuel and labor costs for plant
operation. Within this experiment these costs are fixed, and set equivalently to either a high or
low cost ($1 or $2), and the quantity of each type of generator is equal (10 generators of each).
The distribution of these generators among subjects is set randomly each period. This is fairly
48
consistent with operating markets in which operating costs are relatively fixed in the short run. In
this experiment, these costs are given by
2
1 ,
high
pi eit
t c low
p q
=
åå
, which is the simple sum across two
rounds, of each subject’s endowment of high and low cost generators. Those generators that do
not win production rights in the electricity market do not produce electricity, and thus, do not
carry any operating costs ( 0
eit
q =). For example, a firm that has both a low and a high marginal
cost generator, and wins production of both generators in both rounds, would incur a period
production cost of $6.00
2
1
$1 $2
t=
æ ö
+
ç ÷
è ø
å
.
Third, firms incur costs of acquiring emissions allowances. In this experiment, emissions
costs are determined endogenously, and are given by
2
1
mt wit
t
P q
=
å
, which is the simple sum of
emissions allowances purchased at each round’s emissions auction clearing price, summed across
both rounds. A complicating factor not included in this experiment is the endogenization of
emissions factors. Emissions factors are a fixed amount of emissions effluent (e.g., carbon
dioxide, sulfur dioxide, oxides of nitrogen) per unit of electricity production. Including this
would complicate an already heavily-complicated experiment, and will be pursued in future
research. For simplicity, all generators that win production require a single emissions allowance,
which is equivalent to requiring one emissions allowance for each unit of electricity generated.
Fourth, firms operating in emissions markets may resell unused emissions allowances to
other firms that require them for purposes of compliance. Firms may also operate as arbitrageurs
and acquire emissions allowances early in the game when the emissions cap is less restrictive, and
sell them to firms at later periods when the cap is tighter and when the demand for those
allowances has increased. Resale in this experiment is similarly endogenous. Firms that acquire
49
emissions allowances in round 1 can post them for resale in the emissions auction in round 2. In
Equation 1, resale revenue is given by
2 2 m mi
P q . Note, because this experiment’s inter-temporal
dynamic lasts only two rounds per period, resale revenue can only be acquired in the second
round, and thus this term is not summed across periods as the other terms are. This experiment is
designed to emulate operating emissions markets with a declining cap across time and allowance
scarcity in later rounds. Thus, 25 percent fewer emissions allowances are sold by the government
in the second round of this experiment.
Also noteworthy is the fact that this term allows for the possibility of loss of revenue in
resale, which can occur in operating emissions markets as well, and which in fact did occur at
multiple occasions throughout the operation of this experiment. Loss of resale revenue occurs
when firms purchase emissions allowances at a high price in early rounds, and sell those
allowances in later rounds at a lower price than they paid for them, or do not sell them at all and
end up with more allowances than requisite.
In emissions markets where early compliance periods are marked by uncertainty and fear
of not acquiring sufficient quantities of allowances, firms are often willing to pay a premium to
ensure they are not, figuratively speaking, left without a chair when the music stops. Then, at
later periods when fears subside and demand calms, the price declines to more competitive levels.
Firms that attempt to sell emissions allowances that they purchased during the initial high-priced
frenzy into the calmer market in later rounds can lose money in resale. This is a kind of inter-
temporal winner’s curse, identified by McAfee and Vincent (1993) as the "Declining Price
Anomaly."
Finally, firms that operate in the electricity market and produce electricity but do not
carry sufficient emissions allowances to cover their emissions, are penalized by the regulatory
agency or governmental organization with a non-compliance penalty. Non-compliance penalties
50
are similarly endogenous within this experiment, and are given by
N ni
P q . In this experiment, a
non-compliance penalty of $5.00 is levied on a subject for each missing, or short, emissions
allowance, therefore
N
P = $5.00. These levies only come at the end of a period (i.e., at the end of
round 2), which emulates a compliance period. This enables subjects to carry a surplus or deficit
of emissions allowances in any single round, and thus gives them sufficient flexibility to acquire
allowances in the round that they believe will be most profitable.
4.2 Strategies with Market Power
There are three main strategies profiles that can be pursued by an oligopolist in an
electricity market with an emissions market overlay. For purposes of exposition, I identify these
strategies as Type I, II, and III Market Power. However, it should be clear that there is one
particular set of market parameters that is considered here, those relevant to the design and setup
of this experiment. Policymakers are still designing and implementing new market parameters
and designs for future markets. These can be generalized to a case with more or fewer
oligopolistic firms, with different budget constraints, different rules regarding collusion, or
different auction formats. As such, these strategies may be equivalent, irrelevant or inapplicable
under different market parameters.
4.2.1 Type I Market Power
A Type I market power strategy is consistent with supply/demand reduction and anti-
competitive bidding in both the electricity and emissions markets. Under this strategy profile,
firms within the competitive fringe are assumed to bid competitively within both markets, which
under symmetric behavior, is the most profitable behavior for fringe firms. Oligopolistic firms
51
know ex ante that they carry a small portion of the residual supply into the electricity market,
even if all firms in the competitive fringe bid competitively (at marginal cost).
Because the uniform price electricity auction will clear at the highest cost supplying
generator, any remaining generation that is needed to clear the electricity market after all firms in
the competitive fringe have supplied their generation at marginal cost, can be offered by
oligopolistic firms at the highest possible price that expropriates the full rents from the
government. That is, without the oligopolist, the electricity market cannot clear. And, because
the auction is a uniform-price auction, firms in the competitive fringe receive this rent as well,
even though they bid competitively. This is equivalent to supply reductions in operating
electricity auctions, where firms that are at the margin can swap their competitive generation
assets for more costly generation assets, and raise the common clearing price to all generation
assets within the control area, or simply bid marginal generation assets at the full rent level and
set the common region-wide electricity price for all generation. This equivalently raises the
payout to firms in the competitive fringe as well. In terms of Equation 1, this is a strategy by
which the oligopolist seeks to maximize
2
1
et eit
t
P q
=
å
, and minimize
2
1 ,
high
pi eit
t c low
p q
=
åå
and
2
1
mt wit
t
P q
=
å
.
The success of Type I Market Power is dependent upon two crucial factors. First, it
depends heavily on the demand for electricity. In this experiment, demand for electricity is
stochastic and endogenous, just as it is in operating electricity markets. Second, it depends upon
cooperation between the oligopolists. These two factors interact with each other simultaneously.
52
Figure 2.2 Type I Market Power(Supply Reduction) In the treatment group, when the demand is low, ten units of electricity are demanded by
the government. The most electricity that can be supplied by firms in the competitive fringe is 8
units. The residual 2 units requisite for closing the market must come from the oligopolists. If
the oligopolists cooperate, they can supply those two units jointly at the price cap of $10.00 rather
than marginal cost plus allowance cost. It is more profitable for oligopolists to supply one unit
each at $10.00 than all six of their units at marginal cost. Supplying all six units raises both their
production cost and their cost of acquiring emissions allowances. Doing so would require each
oligopolist to purchase 12 emissions allowances and pay $18.00 in production costs, across two
rounds.
However, cooperation on two remaining units of demand is, in reality, quite difficult to
maintain, because there is tremendous incentive for one oligopolist to slightly undercut the other
oligopolist, even by a penny. This sort of defection occurred frequently in this experiment.
Doing so would enable the defecting subject to earn the rents on one or two additional generating
units.
53
Peak demand for electricity, on the other hand, requires no collusion and allows full rent
expropriation by any firm. When the demand is at peak levels, the quantity of electricity
demanded by the government in this experiment is 20 units. The total quantity of available
supply is 20, and therefore, at peak demand, all subjects, fringe and oligopolist, have the same
level of price influence. In essence, at peak demand, every last generator is needed to close the
market and stave off a blackout, and thus, any generating firm can ask for the price cap.
The most interesting case is the case of intermediate demand for electricity. The
government demands 15 units of electricity in the intermediate demand case, and thus, after all
firms in the competitive fringe supply their generation units at marginal cost, the oligopolists face
a residual demand of 7 units. A residual demand of 7 units has two very interesting properties.
First, under competitive equilibria, it makes either of the two oligopolists a monopolist on the
final residual demand, if the other oligopolist supplies his six units of electricity competitively.
Under collusive equilibria, the two oligopolists must decide not only to cooperate, but which of
the two firms will supply the coveted 4
th
unit of generation. That is, two firms must work
together to supply an odd number quantity of electricity. For equal expropriation of rents under
two-round collusive equilibria, one firm must be willing to supply only 3 units in the first round
and trust the other firm to allow him to supply 4 units in the second round, and not undercut him.
This sort of cooperation was quite difficult for subjects to maintain in this experiment.
4.2.2 Type II Market Power
A Type II market power strategy is consistent with hoarding behavior designed to raise
the operating costs of rival firms (Misiolek & Elder, 1989; Salop & Scheffman, 1983; 1987).
This analysis also considered this strategy for potential dominance. A Type II strategy is an
explicit effort to increase the quantity of electricity sold
eit
q , and raise the cost of emissions
54
allowances
mt
p to other firms acquiring them. This is a particularly risky strategy that requires
oligopolistic firms to forego cooperative behavior and bid their generation assets below the price
cap (at or near marginal cost), and bid up the price of emissions allowances in the first round,
creating scarcity and sufficient demand in the second round to either increase resale revenue, or
make other firms face the non-compliance penalty.
Overall, a Type II strategy is less profitable for several reasons. However, under
alternative market scenarios, it may be a profitable strategy. First, the non-discriminatory
properties of the uniform price auction make raising costs to other firms difficult, because a firm
would also be raising its own costs. Second, there are no cases in which increased resale revenue
compensates for lost production revenue. Third, the threat of not acquiring allowances is limited
by the power of the sanction for non-compliance, which in this case is $5.00 per missing
allowance. Finally, and most notably, a Type II strategy ends up sufficiently raising
2
1
mt wit
t
p q
=
å
to
the firm exercising this strategy, because under all cases except peak electricity demand, the firm
has to purchase a sufficient quantity of allowances that it will never use. Thus, it becomes quite
costly to a firm exercising Type II market power to create scarcity in the emissions allowance
market.
4.2.3 Type III Market Power
A Type III market power strategy is consistent with competitive bidding by firms with
market power. Firms with market power bid into the electricity market at marginal cost plus
allowance cost, and into the emissions allowance market to competitively acquire the quantity of
emissions allowances consistent with their expected level of allowance need
2
1
eit
t
q
=
å
. In this way,
55
excess revenue is not spent on purchasing allowances that will never be used or resold, and firms
only pay a competitive level of production costs.
Overall, a Type III competitive bidding strategy is less profitable than a Type I strategy.
A firm with market power that exercises even a slight amount of market power to either raise
production revenue or suppress allowance prices is, ceteris paribus, a more profitable firm than
one that neglects the exercise of its market power. Nonetheless, this strategy was utilized by
several bidders with market power throughout this experiment. This is largely due to the fact that
cooperative strategies are difficult to maintain under intermediate and low electricity demand
levels.
4.3 Expected Market Prices under Dominant Strategy
Table 2.2 provides the expected auction clearing prices in both the electricity and
emissions market for each of the three levels of stochastic demand, for both treatments and under
the most profitable symmetric market power strategy for each case. Under low demand for
electricity, only the lowest ten (of 20) generation bids are accepted. In the control group without
market power, $1.99 is the threshold price point at which low marginal cost generators can under-
bid the high marginal cost generators, because the remaining ten generators have a marginal cost
of $2.00. With market power, oligopolistic firms face a residual demand for electricity of two
units, and the most profitable market power strategy is a Type I strategy where they collude to
each produce a single unit at the $10 price cap. In both treatments, because only ten units of
electricity are demanded, the total demand for emissions allowances is 20, and the total supply of
allowances is 35, therefore the price is expected to fall to zero in both the control and
experimental treatment under both Type I and Type III market power.
56
Table 2.2 Symmetric Equilibria Predictions
Note: These symmetric equilibria are predicted under the assumption of no backwardation. That is, in repeated periods,
inter-temporal differences across rounds will converge to zero.
Under intermediate demand for electricity, the same principle applies in the emissions
allowance market, under Type I and Type III market power. In the control treatment, we expect a
$2.00 clearing price, because low marginal cost generators will competitively win production and
there will be competition among high cost generators to supply the remaining 5 units of
electricity, which will be competitively supplied at the $2.00 marginal cost. In the treatment with
market power, the most profitable strategy is Type I market power, in which colluding
oligopolists face a residual demand for 7 units of electricity, and cooperate to share production of
these units at the $10 price cap. Again, cooperation here is difficult because two subjects must
split an odd quantity, and coordinate across two rounds which subject will produce the coveted
fourth unit of electricity.
Under peak demand for electricity, we expect all strategies to result in market clearing
prices at the $10 price cap in the electricity market, in both the control and market power
treatments. This is because when demand is at its peak levels, all firms are monopolists over the
residual demand for electricity, and any single firm can set the market clearing price of the
uniform auction at the price cap. In the peak demand case, strategy in the emissions market
Control Treatment
Market Power
Treatment Control Treatment
Market Power
Treatment
Low Energy Demand
$1.99 ≤ $10.00 $0.00 ≤ $0.00
≤ ≤ ≤ ≤
Intermediate Energy
Demand
$2.00 ≤ $10.00 $0.00 ≤ $0.00
≤ ≤ ≤ ≤
Peak Energy Demand
$10.00 ≤ $10.00 $5.00 ≤ $5.00
Energy Market Clearing Price Emissions Market Clearing Price
57
becomes important. To meet peak electricity demand, all 20 generation units are required to
operate in both rounds. In this case, demand for emissions allowances is 40, and the total two-
round supply is 35. Thus, we expect that the binding emissions allowance price will reflect the
point of indifference between operating a generation unit without an allowance and facing a $5.00
sanction and operating that generation unit with a purchased allowance. Thus, the sanction cost
becomes the point of indifference between compliance and non-compliance. This is consistent
with van Egteren and Weber (1996).
5. Results
The six experimental sessions provide for a robust dataset from both the control and
treatment groups. In total, four separate auctions were conducted a total of 144 times each. That
provides for a sample size of 144 auctions from each round, excluding practice/trial rounds, in
both the electricity and emissions markets. In each auction, there were a total of ten subjects
bidding, and the individual analysis provides for a total of 1,440 observations. In many cases, the
mean clearing prices converge closely with expected equilibrium predictions, and in others,
alternative behavioral events occurred. The results are first presented in Section 5.1 and 5.2 for
aggregate market outcomes, where the variable of explanatory interest is the market clearing price
in each round of both the electricity and emissions market. Then, in Section 5.3 the individual
market results are provided, where the dependent variable of interest is the subject’s profit.
5.1 Aggregate Market Results
Descriptive statistics for aggregate market results are provided in Table 2.3. In total, 77
periods were conducted in the control group, and 67 were conducted in the experimental (market
power) group. The quantity of auctions conducted in the control group, and within each demand
58
category is not equal because demand was set stochastically, and because in some experimental
sessions, bidding decisions were made by subjects, in the aggregate, more quickly than in other
sessions. It is appropriate that there are fewer sessions conducted in equivalent time in the
experimental treatment group, because greater time is taken for coordination and chatting
communication among oligopolists.
Table 2.4 provides Mann-Whitney (non-parametric) two sample hypothesis tests of the
equality between the control and experimental treatment in each demand category, for each of the
four markets. In the low demand case, the electricity market cleared fairly consistently with price
predictions in the control group. The experimental group, on the other hand, did not clear at the
predicted price in the low demand case, on all occasions. As mentioned previously, larger
supply/demand
19
reductions on the part of the oligopolist require greater cooperation between
oligopolists, and it becomes increasingly more difficult for one oligopolist not to have the
incentive to undercut the other oligopolist and produce an additional unit of electricity. And on
several occasions, Oligopolists played a Type III strategy. As can be seen by Table 2.3, however,
there were several occasions in which cooperative equilibria were maintained and the $10 price
was achieved even in the low demand case. As can also be seen in both Tables 2.3 and 2.4, the
low demand electricity market price in the experimental treatment was statistically significantly
higher than the control group at the p<.01 level, with 75 percent probability that the true mean
price in the market power treatment is higher than in the control treatment.
In the intermediate demand case, mean electricity auction clearing prices in the control
group do not conform to predicted equilibrium conditions. This is easily explained by learning
behavior associated with repeated-measures aspects of the data. The statistics in both Tables 2.3
and 4 include all rounds of bidding, even early rounds of bidding, but excluding two initial trial
19
Supply reductions in the energy auction, and demand reduction in the emissions auction.
59
periods of bidding. A simple plot of the data for this intermediate demand case (excluded for
brevity) shows that after initial periods of learning, market clearing prices quickly converge
toward the expected $2.00 price. The mean electricity auction clearing price for the control group
for rounds 1 and 2, excluding the first ten periods of bidding for learning, is $2.33 and $2.74,
respectively. Thus, a complete set of summary statistics does not comprehensively explain the
full range of behavior and learning.
By comparison, the market power treatment group mean price is significantly higher, and
is statistically greater than the control group for this case in both rounds, as shown in Table 2.4.
A Type I market power strategy was attempted in a variety of periods, but it was not fully
achieved often. As mentioned previously, collusion among the two oligopolists to supply reduce
in the electricity auction was difficult because of the incentive to slightly undercut the other
oligopolist. In contrast, the mean electricity auction clearing price for the treatment group in the
intermediate demand electricity case where collusion was pursued is $8.78 and $8.82 for round 1
and 2, respectively. Collusion was coded as an explicit signal from one oligopolist to the other,
via chat room transcripts, to set the market price in some way.
Interestingly, explicit deceit defined some of the communication among oligopolists. In
some notable cases, two oligopolists signaled to each other that they would simultaneously bid at
the price cap and explicitly set the clearing price at the cap. In these cases, the initiating
oligopolist would tell the other oligopolist that he/she would bid at $10, and actually placed bids
at $9.95 to slightly undercut the other subject. In one notable case, the other subject expected this
behavior by the initiating oligopolist, and placed an across the board bid for $9.90. In the next
round of chat, when the initiating oligopolist, who expected to undercut the other, learned that
he/she had been undercut, the initiating oligopolist expressed his/her frustration with the other
oligopolist. In essence, the defector was defected upon.
60
Table 2.3 Auction Clearing Summary Statistics
Mean S.D. Min. Max. N
Low Energy Demand $1.82 $0.70 $1.02 $3.00 19
Intermediate Energy Demand $3.04 $1.06 $1.99 $5.99 28
Peak Energy Demand $10.00 $0.00 $10.00 $10.00 30
Low Energy Demand $2.75 $1.73 $1.40 $10.00 25
Intermediate Energy Demand $5.68 $3.42 $1.65 $10.00 22
Peak Energy Demand $9.99 $0.01 $9.98 $10.00 20
Mean S.D. Min. Max. N
Low Energy Demand 1.71 0.47 1.01 2.59 19
Intermediate Energy Demand 3.03 1.05 1.99 5.99 28
Peak Energy Demand 10.00 0.00 10.00 10.00 30
Low Energy Demand 3.01 2.75 1.00 10.00 25
Intermediate Energy Demand 5.86 3.42 2.01 10.00 22
Peak Energy Demand 9.99 0.00 9.99 10.00 20
Mean S.D. Min. Max. N
Low Energy Demand $0.48 $0.76 $0.00 $2.99 19
Intermediate Energy Demand $0.95 $0.93 $0.00 $3.99 28
Peak Energy Demand $3.63 $1.04 $2.00 $5.00 30
Low Energy Demand $0.92 $0.96 $0.00 $3.01 25
Intermediate Energy Demand $1.18 $1.09 $0.00 $3.51 22
Peak Energy Demand $2.77 $0.88 $1.00 $4.00 20
Mean S.D. Min. Max. N
Low Energy Demand $0.00 $0.00 $0.00 $0.01 19
Intermediate Energy Demand $0.41 $0.76 $0.00 $2.50 28
Peak Energy Demand $4.07 $1.05 $1.80 $5.75 30
Low Energy Demand $0.06 $0.22 $0.00 $1.00 25
Intermediate Energy Demand $0.43 $0.93 $0.00 $3.60 22
Peak Energy Demand $2.55 $1.78 $0.00 $4.99 20
Mean S.D. Min. Max. N
Low Energy Demand 3.52 3.65 0 14 19
Intermediate Energy Demand 3.89 4.03 0 16 28
Peak Energy Demand 2.03 1.43 0 4 30
Low Energy Demand 2.72 2.76 0 10 25
Intermediate Energy Demand 2.23 1.84 0 7 22
Peak Energy Demand 2.10 2.19 0 8 20
Energy Market Clearing Price (Round #1) Energy Market Clearing Price (Round #2) Emissions Market Clearing Price (Round #1) Emissions Market Clearing Price (Round #2) Resale Bids (Count of Emissions Permits Posted to Auction) Control Group (No Market Power) Treatment Group (Market Power) Control Group (No Market Power) Treatment Group (Market Power) Control Group (No Market Power) Treatment Group (Market Power) Control Group (No Market Power) Treatment Group (Market Power) Control Group (No Market Power) Treatment Group (Market Power)
61
Table 2.4 Two Sample Wilcoxon (Mann-Whitney) Non-parametric Hypothesis Tests
In the peak demand electricity market, both rounds cleared at expected price levels in
nearly all cases, as shown in Table 2.3. That is, because in peak demand cases, all subjects face a
residual demand for electricity, over which they are a monopolist. The peak demand electricity
market however, fails to meet expectations in the hypothesis test. It is expected to be
insignificant, but because many subjects placed bids at $9.99 instead of $10 in the market power
treatment, the sign of the test statistic switches, and the two-sample test indicates that $9.99 is
statistically different from $10. It is unclear why a number of subjects chose $9.99 instead of $10
in the market power treatment.
20
If the $9.99 prices were recoded to $10, the two-sample test
would be insignificant, as per expectations.
20
It is our belief that the subjects’ interpretation of the experimental instructions contributed to this difference. That is,
it seems that some subjects understood the $10 price cap to include $10 as an ineligible bid, for which the highest
possible bid was $9.99. For exactness, the data are analyzed as they are, rather than recoded.
Energy Market Clearing Price (Round #1) P-value Pr(MP>Control) Low Energy Demand 2.830 *** 0.004 75%
Intermediate Energy Demand 2.360 * 0.018 69%
Peak Energy Demand -2.850 *** 0.004 38%
Energy Market Clearing Price (Round #2) P-value Pr(MP>Control) Low Energy Demand 2.160 ** 0.030 69%
Intermediate Energy Demand 2.600 *** 0.009 72%
Peak Energy Demand -2.850 *** 0.004 38%
Emissions Market Clearing Price (Round #1) P-value Pr(MP>Control) Low Energy Demand 1.160 0.240 60%
Intermediate Energy Demand 0.640 0.510 55%
Peak Energy Demand -2.660 *** 0.007 27%
Emissions Market Clearing Price (Round #2) P-value Pr(MP>Control) Low Energy Demand 1.140 0.254 56%
Intermediate Energy Demand -0.661 0.508 45%
Peak Energy Demand -2.970 *** 0.003 25%
*p<0.1, **p<0.05, ***p<0.01
Z-value
Z-value
Z-value
Z-value
62
However, most subjects in the peak demand case did not bid at the price cap, and instead
continued to bid as they had bid in prior demand cases. This is partly due to the fact that the
instructions provided to the subjects were unbiased and agnostic on strategy and some subjects
simply did not figure out that they possessed monopoly power in peak demand cases.
Interestingly, other subjects understood that they possessed monopoly power, but continued to bid
at marginal cost or lower, because they knew that it would only take one subject to set the market
clearing price at a peak price.
In one notable case, one subject repeatedly placed bids at $0.01, far below marginal cost,
and in all demand categories, because he understood that that bid would ensure him a win in the
electricity market, and because he knew that other subjects would set the price on the margin for
him. Upon interviewing this individual following the experiment, he expressed clearly that this
was his rationale for the excessively low bids, and that he relied on other subjects to worry about
setting the market price at or above marginal cost. Clearly this subject was not playing a
symmetric strategy. This is a strategy that is, to date, not mentioned in the literature of uniform-
price auctions, to the best of this researcher’s knowledge. Therefore, I am calling bidders who
play this strategy “Collective Action Bidders”, because they bid excessively low bids to ensure a
win, and rely on the collective action of others to set the market price at an efficient level.
The emissions market was expected to converge to zero dollars in the low and
intermediate demand case, in both treatment groups. For nearly all periods this expectation was
approximately achieved. Excluding the first ten periods in each session for learning, the mean
control group emissions market clearing price for the low electricity demand case was $0.14 and
$0.00, for rounds 1 and 2, respectively. The equivalent values for the market power treatment are
$1.00 and $0.07, respectively. For the intermediate electricity demand case, the respective values
for the round 1 and 2 control treatment are $0.68 and $0.17, respectively. For the intermediate
63
electricity demand market power treatment, those values are $0.69 and $0.10. There is no
statistically significant difference between the two treatments in either the low or intermediate
demand emissions markets, as expected.
There are three interesting features of the emissions market in the low and intermediate
electricity demand cases. First, due to the fact that the electricity and emissions auctions occur
simultaneously, some subjects expect that they will beat the odds and produce electricity even
though the demand for electricity is low. When this belief is maintained, the demand for
emissions permits is inflated, because subjects purchase allowances for electricity that they
expect to produce, and when they lose production bids in the electricity market, they end up
sitting on unused emissions allowances in round 2. This is one of the main reasons why
emissions prices in round 2 are categorically lower than in round 1.
Second, a sort of "Declining Price Anomaly" occurred in the emissions market. That is,
in round 1, subjects sought to meet their two round compliance obligations by buying the quantity
of emissions allowances they believed they would need if they won production rights in both
round 1 and 2. In other words, subjects sought to buy all of their emissions allowances in round 1.
Subjects seemed to be willing to pay a premium in round 1 to ensure that they would not be,
figuratively speaking, left without a chair in the second round emissions market when the music
stopped. This was consistent in both treatment groups, excluding cases of peak electricity
demand, and only declined slightly in later periods. In a few periods of the market power
treatment, prices in the emissions market peaked when this behavior was pursued by oligopolistic
subjects, who had a collective two-round allowance demand of 68 percent of the total supply of
allowances, if they produced their full quantity of electricity. This accounts for the maximum
values in the low and intermediate electricity demand cases. By extension, this strategy would be
64
unprofitable for large electricity producers in operating emissions markets and directly
contravene a Type I strategy of demand reduction.
Third, Type II market power strategy contributed slightly to these price increases. On a
few occasions, oligopolistic subjects communicated with each other the intent to purchase a large
quantity of emissions allowances. However, purchasing many allowances without attempting to
resell them into the second round is unprofitable. The attempt to profit from resale arbitrage, on
the other hand, is a strategy that is not limited to oligopolists. Few subjects actually profited
much from resale arbitrage in the low and intermediate electricity demand cases, because for
reasons just mentioned, the second round emissions market cleared lower than the first round
market, rather consistently. Secondly, because emissions allowances offered into the second
round raised the second round supply of emissions allowances, resale was frequently unprofitable,
particularly as a symmetric strategy. Thus, the reality of this declining price effect invalidates
inter-temporal arbitrage in auction-based markets.
There is evidence of these effects in operating emissions markets today. In the RGGI
market operating in the Northeast U.S., early emissions auctions saw large price and quantity bids
that dissipated over time. In early rounds, demand exceeded a supply that was far in excess of
actual emissions anyway, and in later rounds prices converged quickly to the price floor (reserve
price) and as demand lessened and resale markets stagnated (RGGI Annual Report 2010). The
ETS markets in Europe also witnessed peak demand in early rounds that drew down significantly
in later rounds, some of which is also attributable to economic downturn.
In the emissions market under peak demand for electricity, results fit squarely with
oligopolistic theory. That is, an oligopolist will seek to maximize profit by increasing the price of
electricity and suppressing the price of emissions allowances. Proving that both effects occur
simultaneously has, to date, not been achieved. As shown in Tables 2.3 and 2.4, emissions
65
market prices cleared consistently lower in the market power treatment. And, the Wilcoxon tests
show, at a high level of significance and in the expected direction, that emissions prices are
statistically lower under market power.
5.2 Aggregate Regression Analysis
Analysis of the aggregate experimental data using regression analysis to determine
appropriate marginal effects of market conditions and behavioral features under further statistical
control is also provided. Tables 2.5 and 2.6 provide Tobit regression analyses for each of the
round 1 and 2 electricity and emissions markets. Tobit regressions are utilized because the
dependent variable in the electricity market is censored at the bid cap of $10, and in the emissions
market at the bid floor of $0.
The reference case excluded from the electricity models of Table 2.5 is the peak
electricity demand category. Because the peak electricity demand market in both treatments
cleared consistently at the expected price cap, including the peak market and excluding one of the
others would provide less detail. The downside however, is that demand categories now carry a
negative sign and must be interpreted counter-intuitively as deviations from the $10 price cap.
The reference case excluded from the models of Table 2.6, for the emissions market, are the
opposite. Detail for those markets is important at the peak and intermediate electricity demand
levels, and thus the low electricity demand category is excluded there.
Recall that the anticipated results of this analysis are to show that under market power,
dominant firms seek to increase profit by simultaneously increasing the clearing prices in the
electricity market and suppressing the clearing prices in the emissions market. Comparing the
coefficients across markets and treatments in Tables 2.5 and 2.6 confirms this hypothesis. Model
1 in both tables provides the most reduced form model whereby the average effects of the
66
treatment condition and demand for electricity are assessed. The dominant driver of electricity
prices is the demand for electricity, however the treatment condition provides a statistically
significant moderating condition. Average prices in our market power treatment were statistically
significantly higher than in our competitive treatment by approximately 22-26 percent. The
intermediate electricity demand market also consistently cleared higher than the low demand
market.
Models 2-4 are interactive models that provide a decomposition of these effects. In
Model 2, the intermediate electricity demand market in the market power treatment (Market
Power * Intermediate Energy Demand) was statistically significantly higher than the intermediate
demand market in the competitive control market (Control * Intermediate Energy Demand), by a
factor of nearly $3.00. The low electricity demand markets, on the other hand, carry coefficients
that are nearly equal, indicating the difficulty of exercising market power in low demand markets.
Model 3 provides the addition of Collusion. When oligopolists collude, the market
clearing price increases by an average of 61-75 percent over the average market with or without
market power, and this effect is highly statistically significant. Notice also that the disparity
between the two treatment's intermediate demand electricity market coefficients diminishes
moderately between Model 2 and Model 3, with the addition of Collusion. This indicates that the
inflated electricity prices of intermediate demand markets is driven in part by collusive behavior.
Model 4 provides a further decomposition of collusion by demand categories. As
mentioned above, collusion is most profitable in the intermediate electricity demand market. The
average round 1 intermediate electricity demand market cleared at $3.04, and the average
intermediate electricity demand market under market power cleared at $5.68. However, the
average intermediate electricity demand market where collusion occurred, cleared at $8.78.
Equivalent round 2 values deviate only very slightly. Clearly collusion is a key ingredient to a
67
successful Type I market power strategy. Collusion * Low Energy Demand also provides
evidence that collusion can be profitable in low electricity demand markets. This coefficient is
significant at the 0.05 level for round 2 only, falling just short of significance for round 1.
For the emissions market analysis provided in Table 2.6, Market Power falls short of
statistical significance in Model 1 for round 1. Although the descriptive statistics and Wilcoxon
tests provide strong evidence that emissions prices are lower in the market power treatment, they
are not across the board lower in all electricity demand categories. As expected, in all markets
except Type II hoarding markets, the emissions price in intermediate and low electricity demand
categories converges toward zero. Because of the inherent energy-emissions linkages, as larger
quantities of electricity are required to clear the aggregate market, greater quantities of emissions
allowances are required. Thus, unless a Type II hoarding strategy is pursued by dominant firms,
intermediate and low electricity demand markets under market power will see statistically-
equivalent zero price convergence in the emissions market.
Looking beyond the reduced form Model 1 provides evidence that emissions prices in
peak electricity demand markets with market power are statistically lower than in peak electricity
demand markets without market power. Comparison across the coefficient for Market Power *
Peak Energy Demand and Control * Peak Energy Demand shows that emissions prices are
consistently higher in competitive markets. This same effect for the intermediate electricity
demand market only holds consistently in the second round, as the first round provides weaker
evidence.
68
1.23 ** 1.48 **
(-2.77)(3.08) -8.66 *** -8.99 ***
(-14.33)(-13.97) -10.72 *** -11.05 ***
(-16.65)(-16.49) -6.94 *** -7.55 *** -8.49 *** -7.19 *** -7.87 *** -8.54 ***
(-7.93)(-9.76)(-14.17)(-7.65)(-9.71)(-12.84) -10.05 *** -10.12 *** -9.75 *** -10.14 *** -10.17 *** -10.33 ***
(-18.29)(-17.24)(-19.27)(-16.57)(-15.92)(-20.12) -9.75 *** -8.94 *** -8.91 *** -10.14 *** -9.12 *** -9.27 ***
(-16.33)(-19.65)(-18.65)(-15.24)(-18.51)(-18.65) -10.91 *** -10.07 *** -10.01 *** -11.48 *** -10.39 *** -10.51 ***
(-18.15)(-22.77)(-21.38)(-17.40)(-22.67)(-22.32) 3.44 ** 4.29 ***
(3.53)(3.87) 5.92 *** 5.67 ***
(5.85)(5.66) 1.97 4.31 *
(1.58)(2.45) -0.02 0.01 0.02
(-0.04)(0.31)(0.50) -0.02 -0.02 -0.05 * -0.06 *** 0.01 0.00 -0.03 -0.04 *
(-0.74)(-1.13)(-2.29)(-3.45)(0.18)(0.02)(-1.22)(-2.01) 12.65 *** 13.14 *** 12.64 *** 12.74 *** 12.61 *** 13.25 *** 12.51 *** 12.79 ***
(20.76)(21.38)(21.04)(21.54)(18.70)(19.69)(18.86) -20.71
F 95.47 *** 80.90 *** 89.24 *** 93.24 *** 80.56 *** 67.73 *** 83.32 *** 79.97 ***
N 144 144 144 144 144 144 144 144
Log Pseudo-likelihood -221.68 -214.21 -195.02 -180.05 -232.14 -225.68 -201.73 -193.41
Pseudo R
2
.34 .36 .42 .46 .31 .33 .40 .43
AIC 455.36 442.42 406.05 378.10 476.28 467.36 421.46 406.82
BIC 473.18 463.21 429.81 404.83 494.10 491.12 448.19 436.51
Market Power * Intermediate Energy Demand
Market Power * Low Energy Demand
Control * Intermediate Energy Demand
Energy Market (Round 1) Model 1 Model 3 Model 4
Market Power
Tobit regression results with White's Robust T-values in parentheses. * p<0.05, ** p<0.01, *** p<0.001. Right-upper censoring at $10 Energy Market price cap. Peak Energy Demand excluded
from models as reference case
Energy Market (Round 2) Model 1 Model 3 Model 4 Model 2 Model 2
Constant
Control * Low Energy Demand
Collusion
Collusion * Intermediate Energy Demand
Collusion * Low Energy Demand
Period
Allowance Resale
Intermediate Energy Demand
Low Energy Demand
Table 2.5 Regression Results for Aggregate Energy Analysis
69
-0.09 -0.71 *
(-0.51)(-2.14) 2.84 *** 5.29 ***
(11.36)(12.44) 0.54 * 1.42 **
(2.08)(2.98) 2.60 *** 2.50 *** 2.59 *** 4.78 *** 4.68 *** 4.69 ***
(7.70)(7.22)(7.40)(6.88)(6.65)(6.43) 0.94 * 0.75 0.60 1.90 * 1.71 ** 1.70 *
(2.55)(1.88)(1.32)(2.54)(2.36)(2.26) 0.49 0.38 0.49 0.86 0.75 0.85
(1.29)(0.93)(1.28)(1.22)(1.06)(1.21) 3.48 *** 3.47 *** 3.48 *** 6.37 *** 6.35 *** 6.36 ***
(10.51)(10.50)(10.51)(9.87)(9.91)(9.93) 0.74 * 0.74 * 0.73 * 2.29 *** 2.27 *** 2.27 ***
(2.22)(2.22)(2.22)(3.35)(3.35)(3.35) 0.43 0.39
(1.48)(0.75) -0.06 0.38
(-0.13)(0.47) 0.80 0.43
(1.76)(0.48) -0.24 *** -0.23 *** -0.23 ***
(-3.68)(-3.60)(-3.58) -0.01 -0.01 -0.01 -0.01 0.00 -0.00 -0.00 -0.00
(-0.42)(-0.51)(-0.67)(-0.67)(0.01)(-0.23)(-0.28)(-0.24) 0.55 0.24 0.28 0.28 -1.59 *** -1.76 ** -1.74 ** -1.76 **
(1.95)(0.70)(0.80)(0.81)(-3.34)(-2.74)(-2.71)(-2.73) F 46.72 *** 35.18 *** 29.95 *** 26.56 *** 66.19 *** 52.21 *** 45.43 *** 40.60 ***
N 144 144 144 144 144 144 144 144
Log Pseudo-likelihood -205.97 -201.09 -200.07 -199.65 -162.59 -151.35 -150.99 -151.04
Pseudo R
2
.21 .23 .24 .24 .31 .36 .36 .36
AIC 423.93 418.19 418.14 419.31 337.18 320.69 321.98 324.09
BIC 441.75 441.94 444.87 449.00 354.99 347.42 351.68 356.76
Market Power * Intermediate Energy Demand
Market Power
Peak Energy Demand
Intermediate Energy Demand
Market Power * Peak Energy Demand
Emissions Market (Round 1) Emissions Market (Round 2) Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4
Market Power * Low Energy Demand
Constant
Tobit regression results with White's Robust T-values in parentheses. * p<0.05, ** p<0.01, *** p<0.001. Left-lower censoring at $0 Emissions Market price. Low Energy Demand excluded from
models as reference case
Control * Intermediate Energy Demand
Collusion
Collusion * Peak Energy Demand
Collusion * Intermediate Energy Demand
Allowance Resale
Period
Control * Peak Energy Demand
Table 2.6 Regression Results for Aggregate Emissions Analysis
70
Prices in the second round are consistently lower than in the first round, as is evidenced
by the declining price anomaly discussed above. However, the larger coefficients for the second
round models in Table 2.6 can be misleading without consideration given to the change across
rounds of the intercept term. Round 1 emissions prices are consistently higher because
participants are willing to pay a premium for the certainty of knowing that they will have
sufficient allowances for the entirety of the compliance period. This effect does not however,
hold for the competitive peak electricity demand market. In the competitive market with peak
electricity demand, some participants realize at the end of round 1 that their bids were not
accepted and that, in order to continue producing electricity in round 2, they will need to raise
their emissions bids in round 2 to ensure a sufficient quantity of allowances to cover their
electricity production from round 2, and also to cover their losing bids from the round 1
electricity market. In a more dynamic emissions market with quarterly auctions and multi-year
compliance periods, this demand adjustment should be more smooth.
Collusion, however, does not have any statistically significant impact on emissions
market prices in any electricity demand category. This provides evidence that participants
collude to set electricity prices but inherently treat a Type II strategy in the emissions market as a
less profitable strategy, requiring the unnecessary purchase of allowances.
Finally, allowance resale into round 2 statistically significantly decreases emissions
prices. Participants who resell allowances into round 2 increase the supply of allowances in that
round and thus, the price drops slightly. The variable Allowance Resale is a non-binary variable,
and thus, can be interpreted approximately as a 23 cent decrease in the round 2 emissions prices
for every additional allowance posted for resale. This is consistent with the RGGI market, which
has a nearly non-existent secondary market characterized by exceptionally low volumes of trades.
Demand for allowances can be met by the auction and adjustment bids can simply be carried over
71
to the next quarterly auction. This coefficient provides evidence that a Type II market power
strategy is not a dominant strategy. The regressions in Table 2.5 likewise confirm that emissions
allowance resale has almost no effect on electricity prices.
5.3 Individual Regression Analysis
This section provides a regression analysis of the individual subject data where the
dependent variable of interest is the individual subject's profit. Profit is measured at the level of
the period (both rounds), consistent with Equation 1 above, and includes revenue from production
of electricity in both rounds, production costs, emissions costs, revenue from allowance resale,
and any non-compliance penalty. Models 1 and 2 are Linear Mixed Models (LMM) (Rabe-
Hesketh & Skrondal, 2008; West et al., 2010) with random effect intercept parameters within
experimental sessions to account for session-specific differences in the data. That is, the models
allow for a subjects-within-sessions assessment. The models also allow independence in the
residual errors but permit error variance to vary by treatment group, as appropriate. Treatment-
specific error variance highly increases the robustness of the models and reliability of estimates,
as there is greater variance in market clearing prices in the market power treatment in three of the
four models. The models also include a variable for “period” because the data are collected in
repeated-measures (West et al., 2010), enabling the model to account for subject learning across
time periods.
The first eight interaction terms provide for a decomposition of profit across electricity
demand levels and subject types. Recall that within the control group, all subjects are of the same
type, but in the market power treatment, subjects are randomly assigned each period to be of
either type; fringe or oligopolist. The coefficients of these models confirm that it is clearly most
72
profitable for all firms under peak electricity demand, and increasingly so for oligopolistic firms.
The competitive subjects under low electricity demand are excluded for comparison.
Of particular note are the coefficients for fringe subjects under intermediate electricity
demand. Recall that fringe subjects can produce at most a single unit of electricity in each round,
whereas the competitive subjects who have more market influence can produce at most two units
each round. Despite this, the coefficient for fringe profit is larger than that for the equivalent
demand level among competitive subjects. Fringe subjects profit from the increased electricity
prices and suppressed emissions prices at the hand of oligopolistic firms. Low electricity demand
cases on the other hand, are statistically insignificant.
The next six interaction terms decompose drivers of profit by electricity production
across subject types. Each subject can produce a variable quantity of electricity from its
endowment of high and low marginal cost generators. Appropriately, we expect the coefficient
for high marginal cost generation to be smaller than low cost because low cost generation is more
efficient. For every unit of low marginal cost generation that is sold into the electricity market by
oligopolistic firms, their profit increases by approximately $2.00, irrespective of the level of
demand for electricity. This coefficient is about fifty cents larger for fringe firms, mainly because
fringe firms can produce that single unit of electricity with the purchase of only a single
emissions allowance. Competitive subjects, on the other hand, profit significantly less for each
unit of electricity that they produce, for both high and low marginal cost generation, due to the
more competitively priced electricity market, as provided in the aggregate analysis.
The purchase of emissions allowances is included in these models also to account for the
impact of emissions market linkages on profit. Three interaction terms are included in the models
of Table 2.6 that interact subject type with the quantity of emissions allowances the subject
received in both rounds. Of note is the sign reversal between a positive and statistically
73
significant coefficient for the competitive market and a negative and significant (in Model 2 only) coefficient for the oligopolists’ purchase of allowances. Whereas in both markets, the purchase
of emissions allowances incurs costs to the subject, purchases in the competitive market are made
under competitive pricing under which allowance acquisitions are made for generators that are
incurring a slight profit. The coefficient for oligopolists’ acquisition of emissions allowances is
however, negative and less significant. Oligopolists who supply a significant portion of system-
wide electricity are responsible for a significant portion of allowance demand, and thus are forced
to purchase the lion’s share of emissions allowances in peak and intermediate demand markets.
Under a competitive market, costs of acquiring emissions allowances are passed through in the
electricity price at marginal cost or slightly higher. However, in an anti-competitive market,
electricity prices are unnaturally inflated and the cost of electricity does not reflect passed through
costs of emissions. To dominant firms who can raise the price of electricity anyway, additional
allowance acquisitions are simply an annoyance that diminishes their rents, even if they can
suppress the allowance prices slightly.
Whereas the aggregate analysis showed that collusion had a significant impact on
clearing prices, the individual analysis includes collusion as an explanatory variable at the subject
level to explain profit. The coefficient for collusion in Model 1 suggests that collusive behavior
is exceptionally profitable, at a high level of statistical significance. Model 2 decomposes
collusive behavior across the three electricity demand categories, and provides evidence that
collusion is most profitable in the intermediate demand scenario. In the peak demand scenario,
collusion is less relevant because every last generator is a monopolist over some residual demand
for electricity, and thus collusion is unnecessary to set the electricity price at the price cap.
Collusion under intermediate and low electricity demand however, can allow for a coordinated
74
Type I market power strategy of supply/demand reduction, thus providing strong evidence of the
profitability of that strategy for dominant firms.
Model 3 provides a Quantile Regression analysis that provides further insight into the
determinants of profit in the joint energy-emissions market. Quantile regression techniques are a
useful non-parametric technique for analyzing complex data beyond the mean of the dependent
variable (Koenker, 2005; Rose & Dormady, 2011; Rose, Wei & Dormady, 2011). The dependent
variable in this analysis falls short of standard normality assumptions, as appropriate, because of
differences across experimental treatments; namely significantly larger profits incurred by
oligopolists, and spreads across the three electricity demand levels. Model 3 provides three
benchmark quantile models, weighted at the 25
th
quantile (t =.25), median regression (t =.50),
and the 75
th
quantile (t= .75). These three models assign greater explanatory weight to
observations near their respective quantile along the distribution of profit. The 75
th
quantile
model therefore, provides insight into the drivers of those observations in the highest tiers of
profit. The 25
th
quantile model provides insight into the drivers of those in the lowest tiers of
profit, and the median regression model provides an alternative to a standard OLS approach, by
providing insight into those observations at the median of profit.
A majority of the variables in Model 3 carry signs and magnitudes in the expected
direction, with some notable exceptions. We see a sign change of statistical significance in the
lower quantile model for fringe firms in low electricity demand markets. At lower quantiles of
profit, fringe subjects experience significantly less profit than subjects in competitive markets
under low demand. Although the average fringe subject profit under low demand markets is
more than twice as large in magnitude as the average competitive subject’s profit ($1.93 versus
$0.76), the variance of that figure is significantly larger for fringe subjects. This stems from the
fact that fringe subject profit swings widely based on the coordination of oligopolists. That same
75
figure for fringe subjects when oligopolists collude is $5.37. The sign of this coefficient at the
lowest quantile reflects fringe subject profit in those auctions in which oligopolists did not pursue
a Type I market power strategy. Therefore, it is clear that fringe firms can expect to piggyback
on the earned rents of dominant firms that play a strategy of supply/demand reduction, but when
they do not play that strategy, fringe firms can expect to earn altogether less money than firms in
a competitive market.
Another interesting sign change revealed by the quantile models are those negative signs
for oligopolistic subjects supplying high marginal cost electricity into the market. Those
coefficients are negative and significant in the lower and median quantile model, yet positive and
significant in the higher quantile model. Oligopolists who supply their inefficient generation into
the low and intermediate demand electricity markets are clearly not playing a Type I market
power strategy. If they were, they would supply the fewest and most efficient generators possible
to maximize their profit. Recall that under a low electricity demand market, the oligopolists face
a combined residual demand of 2 units of electricity each round, and under the intermediate
demand market they face a residual demand of 7 units. Combined, the two oligopolists have an
endowment of 6 units of low marginal cost electricity and 6 units of high marginal cost electricity.
Supplying additional high marginal cost electricity into either the low or intermediate demand
electricity markets means that they are attempting to play a Type III market power strategy, and
are not attempting to supply/demand reduce. Therefore, this variable carries negative and
significant coefficients in the lower and intermediate quantiles of profit, reflecting the
profitability of supply/demand reduction in all but the most profitable peak demand markets,
where every last unit of generation is demanded anyway.
This same effect provides insight into competitive markets as well. The variable for high
marginal cost electricity supplied by competitive firms carries a negative and significant
76
coefficient in the lower quantile and a positive and significant coefficient in the higher quantile.
Competitive subjects supplying high marginal cost generation into low paying markets, which are
typically low electricity demand markets, are supplying electricity into a market that pays them
less money than it costs them to produce that electricity. This provides policy insight for
deregulated markets that attempt to incentivize market entry through resource adequacy or
capacity payment policies. Even firms in competitive markets can profit from reducing the
overall supply of electricity into the market, particularly high marginal cost generation.
77
Table 2.7 Regression Results for Individual Analysis
Fixed Effects
50.77 *** 51.73 *** 56.65 *** 63.89 *** 61.64***
(12.44)(13.84)(218.55)(193.25)(54.15) 12.14 *** 3.53 5.29 *** 10.74 *** 21.58***
(3.37)(1.05)(23.40)(39.19)(23.71) -2.24 -0.27 1.09 *** 1.87 *** 2.04*
(-0.70)(-0.09)(6.97)(8.25)(2.44) 6.38 * 6.37 ** 7.24 *** 9.16 *** 6.93***
(2.50)(2.74)(59.13)(62.49)(11.94) 4.68 4.52 * -0.16 1.79 *** 7.25***
(1.87)(1.98)(-1.37)(12.91)(13.67) 0.14 0.03 -0.78 *** -0.02 -0.77
(0.06)(0.01)(-7.93)(-0.16)(-1.44) 15.24 *** 15.26 *** 14.12 *** 15.68 *** 17.40***
(42.70)(42.77)(173.03)(177.68)(55.26) 0.39 0.42 0.24 *** 1.00 *** 1.20***
(1.37)(1.44)(3.92)(14.06)(4.66) 1.89 *** 2.16 *** 0.72 *** 0.75 *** 1.01***
(4.64)(5.83)(24.90)(20.13)(6.89) 0.86 1.29 ** -0.11 *** -0.59 *** 0.47***
(1.88)(3.10)(-3.54)(-13.81)(3.06) 2.48 *** 2.51 *** 0.79 *** 1.26 *** 2.98***
(4.56)(5.10)(18.21)(25.99)(16.26) 1.83 *** 1.87 *** 0.05 0.41 *** 2.31***
(3.21)(3.62)(1.14)(8.05)(11.82) 0.95 *** 0.93 *** 0.24 *** 0.64 *** 0.56**
(4.47)(4.40)(6.64)(12.78)(3.01) 0.56 ** 0.55 ** -0.12 ** 0.02 1.07***
(3.02)(2.97)(-3.04)(0.44)(6.33) -0.41 -0.58 * -0.37 *** -0.13 *** -0.25**
(-1.55)(-2.45)(22.40)(-0.51)(-2.56) 0.08 0.10 0.52 *** 0.57 *** 0.66***
(0.24)(0.35)(13.46)(17.86)(7.19) 0.27 *** 0.28 *** 0.06 *** 0.28 *** 0.28***
(3.74)(3.93)(3.76)(16.14)(5.59) 11.64 *** 2.98 *** 9.75 *** 17.92***
(8.26)(27.43)(76.56)(38.03) -4.04
(-1.62) 29.75 ***
(14.82) 3.24
(1.54) -0.00 -0.06 -0.09 *** -0.36 *** 0.01
(-0.01)(-0.37)(-3.44)(-10.05)(0.07) -0.14 *** -0.14 ***
(-12.25)(-12.16) 0.52 0.47 -0.36 *** -1.12 *** -0.56
(0.32)(0.32)(-5.83)(-11.85)(-1.44) Random Effects & Variance Components
σ
2
constant
(session) 7.3 6.07
σ
2
constant
(subject: session) 0.19 0.18
σ
2
e
(control treatment) 6.99 6.99
σ
2
e
(market power treatment) 47.68 39.31
Model Information Criteria
Wald Chi-squared 10012.75 *** 10785.46 ***
REML Log Likelihood -4098.91 -4030.89
AIC 8247.81 8115.78
BIC 8379.62 8258.13
Psuedo R
2
.56 .62 .64
Observations 1440 1440 1440 1440 1440
Model 3
Quantile .25 Quantile .50 Quantile .75
Linear Mixed Model (LMM) Restricted Maximum Likelihood Z-values in parentheses. T-values in parentheses for Quantile Regression models
(Model 3). *p<.05, **p<.01, ***p<.001
Oligopolist*Low Energy Demand
LMM
Oligopolist*Peak Energy Demand
Oligopolist*Intermediate Energy Demand
Model 2 Model 1
Oligopolist*Emissions Allowances
Fringe*Peak Energy Demand
Fringe*Intermediate Energy Demand
Fringe*Low Energy Demand
Competitive*Peak Energy Demand
Resale (Emissions Allowances) Period
Constant
Collusion*Intermediate Energy Demand
Collusion*Low Energy Demand
LMM
Fringe*Emissions Allowances
Competitive*Emissions Allowances
Collusion
Collusion*Peak Energy Demand
Oligopolist*High MC Generation
Fringe*Low MC Generation
Fringe*High MC Generation
Competitive*Low MC Generation
Competitive*High MC Generation
Competitive*Intermediate Energy Demand
Oligopolist*Low MC Generation
78
6. Conclusions
This chapter evaluates the effect of emissions markets on electricity markets under real-
world market characteristics, including market power, stochastic electricity demand, collusion
among dominant firms, inter-temporal dynamics, a tightening emissions cap and resale arbitrage.
Using students at the University of Southern California as experimental subjects, we perform a
series of laboratory experiments to investigate a simultaneous multi-round energy-emissions
market under a stylized market power scenario. We investigate the data at a variety of levels to
test the degree to which dominant firms can utilize linkages that exist between these markets to
inflate electricity prices and suppress emissions prices.
The results suggest that dominant firms can utilize emissions-energy market linkages to
their operational advantage to inflate electricity prices without systematically inflating emissions
prices, and in cases of peak electricity demand, suppress emissions prices. Firms that possess
market power in these markets have a variety of strategies at their disposal with which to seek
higher profits. This chapter characterizes three main types of market power, and finds that, under
this stylized case, the most profitable market power strategy is a collusive Type I market power
strategy of supply reduction in the electricity market and demand reduction in the emissions
market.
In operational auctions there can exist tremendous temptation for dominant firms to
deviate from collusive agreements. To avoid Type I error, this experimental design provides
cases in which it is more, rather than less, difficult for dominant firms to collude. This is done by
splitting residual demand across periods and limiting the residual demand that is faced by
dominant firms in cases of low electricity demand. These experimental results suggest that,
despite the operational difficulty of collusion, collusive supply/demand reduction is far more
profitable in both markets than competitive behavior.
79
Policymakers have long acknowledged the operational hurdles provided by market
imperfections such as market power, in deregulated electricity markets. Contemporary electricity
markets include a variety of regulatory controls to limit strategic bidding behavior and other
exercises of market power, depending upon the rules of the regional regulatory regime and the
nature of the market. Whereas it is often the same firms participating in regional auction-based
emissions markets, equivalent controls do not exist. As the Northeast’s Regional Greenhouse
Gas Initiative (RGGI) and California’s AB32 greenhouse gas market move forward, greater
consideration must be given to the linkages that exist between operating electricity markets and
developing emissions markets, particularly with regard to the nature of imperfect markets
characterized by a handful of dominant firms in a regional oligopoly.
80
Chapter 3: Regulatory Mechanisms and Policy Approaches
1. Introduction
Electricity deregulation brought to light many of the troublesome implications of market
characteristics inherent in the electric power sector (e.g., inelastic demand, non-storability of
power, scale economies and market power). The market design in California culminated in
skyrocketing prices, blackouts, bankruptcies, and a successful gubernatorial recall campaign.
Since that time, policymakers have iteratively updated the structure and regulation of electricity
markets with a variety of revised and additional mechanisms intended to alleviate some of the
inherent structural flaws that have emerged in these markets. Today, these markets are defined
by a complex mix of multi-settlement auction-based price mechanisms, overlaid by a variety of
complex auction-based hedging markets for financial rights to locational price differences caused
by transmission limitations (commonly referred to as "transmission congestion").
On top of this, policymakers are adding market-based regulatory mechanisms to
electricity markets to mitigate greenhouse gas emissions. This chapter aims to address what few
scholars have, by evaluating the interrelationships that exist between the already-complex mix of
regulatory mechanisms and policies in contemporary electricity markets, and those additional
regulatory mechanisms and policies that are being added to these markets under auction-based
GHG cap-and-trade programs.
This chapter aims specifically to address the issue of market power and its exercise in
contemporary electricity markets overlaid by cap-and-trade regimes. This chapter provides an
analysis of three regulatory mechanisms, scantly covered in the literature, which effectively
reduce the incentive to exercise market power in both the electricity and emissions markets. This
chapter begins with a brief primer on contemporary electricity and emissions markets in sections
81
2 and 3 respectively. Section 4 provides a brief primer on market power and its exercise in these
markets. Section 5 provides an analysis of virtual (convergence) bidding as a mechanism to
reduce the incentive to exercise market power. Section 6 provides an analysis of the California
Assembly Bill 32 (AB 32) market’s use of consignment auctions as a mechanism to neutralize the
incentive to exercise market power through strategic demand reduction. Section 7 provides an
analysis of the AB 32 market’s use of multilevel accounts with both strategic holding and
purchase limits that reduce the potential for the exercise of market power through hoarding
emissions permits to exercise control over the secondary, or resale, markets. Section 8 concludes.
2. A Brief Primer on Deregulated Power Markets
2.1 Background
The purpose of this section is to provide a brief background on contemporary electricity
markets, within the U.S. context. Although a full exposition of the subject would span volumes,
this section is intended to provide an appropriate survey of the nature and structure of these
markets to illuminate the reasons why market power is a subject of focus, particularly within the
context of contemporary market-based emissions programs. Design changes within these markets
in the past few years have provided for significant changes in the incentives for exercising market
power. This section builds mainly upon the market designs of the California wholesale electricity
market.
The production and delivery of electric power on an integrated network involves five
sequential operations: generation, scheduling of generation, transmission, distribution and
retailing. Generation, scheduling and transmission generally constitute the wholesale market, in
which electricity is produced and sold for the purposes of resale. This chapter concerns itself
primarily with wholesale markets. The distribution function is typically carried out by
82
distribution companies, also known as Load Serving Entities (LSEs), of which the reader may be
more familiar: Investor Owned Utilities (IOUs), Publicly Owned Utilities (POUs), and rural and
agricultural cooperatives. The retailing function is carried out by Load Serving Entities (LSEs) comprised of the aforementioned distribution companies plus competitive retailers.
Across these five operations there exists a divergent and complicated regulatory structure.
The interstate wholesale markets are regulated by the federal government, namely the Federal
Energy Regulatory Commission (FERC). Pursuant to the Federal Power Act of 1935, the FERC
is tasked with ensuring that the prices in these interstate wholesale power markets are “just and
reasonable.” The federal jurisdiction in these markets is based on the fact that these wholesale
power sales and related transmission services are typically transacted in interstate commerce.
The retail side, however, is regulated in most states by public utility commissions or public
service commissions.
21
State regulators are tasked with ensuring that IOUs are allowed to charge
prices that produce a fair rate of return, while at the same time protecting consumers from abuses
of monopoly resulting from significant economies of scale inherent in electricity production and
distribution, and their franchise monopoly status.
Public service commissions ensure this fair rate of return through a process known as
“retail ratemaking.” IOUs sell the power that they generate or purchase to businesses and
households at state-approved regulated rates. Rates to end users are set, such that, consumers are
protected from potential monopolistic abuses while at the same time, IOUs receive sufficient
financial incentives to provide reliable, safe and sustainable electricity. The wholesale markets
from which the IOUs acquire much of their power are not subject to the state commission’s
regulatory oversight. Within the federal system, it is the responsibility of the FERC to oversee
21
Some states have adopted retail competition in which third party suppliers distribute electricity to end users. In this
case, rates are not chosen directly by a public service commission, but rather by market competition among these
suppliers of retail electricity.
83
the prices in the interstate wholesale markets. In the event the acquisition of power in the
wholesale markets becomes more costly for an IOU, the IOU can petition the state commission
for an increased retail rate structure to defray those costs. The commission is compelled under
the “filed rate doctrine” to allow these FERC-regulated wholesale costs to be passed through,
unless the IOU has engaged in some objectionable purchasing practice.
2.2 The Demand Side
Of preeminent importance is the relationship between demand for electricity in retail
markets and the supply of that power in the wholesale markets. Demand for electricity is
different than most other economic commodities, because there is very little short-run price
response because demand is highly inelastic. Unlike other consumption goods in which price
factors immediately into the consumption decision, businesses and households cannot be
expected to deal with the informational complexity of changing electricity prices continuously
throughout their day as prices change on the wholesale electricity market. State regulators set
rates to businesses and households (i.e., retail prices) in part, to shield them from having to adjust
to price changes that occur every five minutes, on the wholesale markets. Because of this, the
regulatory structure is such that retail prices are disconnected from wholesale prices.
The demand for electricity by businesses and households is also subject to diurnal load
cycles. Throughout any 24 hour period, there are periods of the day in which more electricity is
demanded because of natural cycles of consumption and weather. This systematically varying
time-pattern of aggregate consumer demand is typically referred to as a Load Curve. Demand is
low during hours in which a majority of the population is sleeping, and demand increases during
hours in which more electricity consumption activities are taking place. In warm climates like
California, demand peaks during hot summer afternoons when business and household activity
84
levels peak, and air conditioning is utilized. Consumer demand also exhibits a much smaller
stochastic component from moment to moment within the prevailing load cycle. These "ripples,"
though relatively small compared to the overall diurnal cycle, must nonetheless be accommodated
in load balancing.
This demand-side dynamic fundamentally shapes wholesale markets and requires
corresponding supply-side adjustments. Because the demand for electricity is exogenous,
cyclical, stochastic within short time intervals, and unresponsive to short-run price changes,
wholesale markets are highly complex and must adapt to these inherent market characteristics.
Wholesale markets are designed to ensure that electricity is supplied to match these inherent
structural characteristics both reliably and efficiently, while sending appropriate price signals to
both the supply and demand sides of the wholesale markets.
Figure 3.1 Hourly Average System-wide Load (California ISO, July, 2011) Minimum and Maximum Values Provided in Confidence Bars. Data retrieved from: http://oasis.caiso.com
20,000
25,000
30,000
35,000
40,000
1 4 7 10 13 16 19 22
System-wide Load (MW) Hour of the Day
85
2.3 Demand Driven Supply
The supply side of electricity markets consists of the mechanics by which generation of
electricity is supplied to LSEs consistent with those patterns of demand. The sellers who generate
electricity are IOUs, POUs, independent power producers (IPPs), and power marketers. The
federal government also generates a significant share of electricity.
22
The ultimate buyers in
wholesale markets are the LSEs (IOUs, POUs, Co-ops, and competitive retailers) who retail that
power to businesses and households.
23
For purposes of reliability, supply and demand must be balanced by a centralized system
operator within any given region of the grid. These regions are called "control areas." Some of
these control areas span very large geographic regional areas crossing state lines and containing
large population centers. These are known as Independent System Operators (ISOs) or Regional
Transmission Organizations (RTOs).
24
Within the U.S., there are six main ISOs: New England
(ISO-NE), Pennsylvania, New Jersey and Maryland (PJM-ISO), New York (NY-ISO), California
(CAISO), Midwest (MISO) and Texas (Electricity Reliability Council of Texas, or ERCOT). The
system operators are authorized and regulated by the FERC. All of the permissible activities
pertaining to market operations for the sale of power, transmission and scheduling are codified
and circumscribed in a legal document called a Tariff. Each ISO/RTO must file its own tariff and
have it approved by the FERC.
Under FERC authority, system operators utilize an array of market-based tools to ensure
that the supply of electricity is balanced with demand, given physical transmission limitations
22
Federal power projects often include large scale projects such as the Tennessee Valley Authority, Bonneville Power
Administration, and the Hoover Dam under the U.S. Bureau of Reclamation.
23
Wholesale markets can also include another category of market participant called marketers. These are wholesale
purchasers of electricity that do not have a service obligation, often banks and hedge funds, and purchase electricity as
brokers and arbitrageurs.
24
The FERC provides a helpful map of the regional areas in the US and Canada, at
http://www.ferc.gov/industries/electric/indus-act/rto/elec-ovr-rto-map.pdf
86
inherent to the power network. Balancing is almost entirely accomplished by manipulating
generation of electricity to meet load cycles and stochastic events. To accomplish this large task,
these tools consist mainly of auction-based forward and real-time markets that match demand
bids from LSEs with supply bids from power suppliers. Most system operators operate markets
for the time period one day ahead (usually called a Day Ahead Market or DAM), one hour ahead
(Hour Ahead Market or HAM), and a market in real-time (Real Time Market or RTM). RTMs
are also known as spot markets(although sometimes this is used to refer to all of these short-term
markets). They also operate a market to deal with ancillary and stochastic occurrences, known as
an Ancillary Services (AS) market. This market includes a variety of products, mainly consisting
of spinning and non-spinning reserves that includes generators that are paid a premium to standby
with dispatch-ready power in the event of an outage or unpredictable event.
In addition to the complexities of the temporal dimension, system operators must also
deal with physical and geographic dimensions. Electricity must be delivered over a physical
network of transmission lines, and it must be delivered to load centers where LSEs can utilize it
to meet retail demand. The network consists of a large quantity of nodes, at which power is
“injected” or “withdrawn.” In the California grid, for example, there are over 4,900 nodes.
Transmission of electricity on the grid is limited by physical capacity constraints inherent to the
power lines themselves, and these limit the amount of power that can be transmitted over a given
line.
The ISO’s key function is to achieve reliability through markets by balancing demand at
each node point with an adequate supply of electricity generation, at the lowest possible cost. In
doing so, the ISO must also account for physical transmission capacity limitations. As such, the
scheduling of generation and transmission, or dispatch, is not simply done to minimize costs, but
also to minimize costs given transmission constraints. This is referred to as “security-constrained
87
economic dispatch.” As a result, the prices that clear at each of the 4,900 node points for each of
the markets (e.g., DAM, HAM, RTM), accounts for demand at those node points, for that given
time interval, subject to physical capacity constraints inherent to the injection or withdrawal of
that power.
These prices are referred to as Locational Marginal Prices (LMPs). They are the result of
uniform-price generation auctions in which the owners of generation bid their supply into the
market. That is, these auctions are not pay-as-bid, but rather a common or uniform price clears
the market. LMPs include three components: energy, loss and congestion. The energy
component is the main component and is the result, or clearing price of the system-wide
electricity auction. It is uniform across all nodes. The loss component is generally very small,
and accounts for node-specific energy loss inherent to inefficiencies in transmitting electricity. It
is a node-specific price. The congestion component prices the physical transmission capacity
limitations, and a cost is incurred when voltage transfers begin to run up against the physical
capacity constraint of the lines themselves.
Another fundamental demand-side characteristic that shapes the supply side is the
inherent ramp in the load curve. This is seen in Figure 3.1 as a significant upswing in aggregate
load between 5am and 5pm. To meet this significant temporal demand-side factor, ramping rate
becomes a crucial factor in electricity generation. Ramping rate refers to the speed at which a
generator can respond to demand by increasing (or decreasing) its output of electricity, by either
turning on additional turbines or boilers, or increasing the thermal capacity of the units
themselves by burning fuel at a greater rate. For example, a plant currently producing at 300 MW
may not be able to increase this flow by more than 50 MW by the end of the hour. Those units
that have quicker ramps are advantaged in more remunerative short-term markets, because they
88
can respond to this upswing.
25
The aggregate demand ramp that the ISO must meet is entirely
demand-driven. The fact that this aggregate demand ramp must be met by multiple generators
that have individual ramping constraints means that more generators must be used than would be
necessary in the absence of these individual generator ramping constraints.
Because of the demand-side impact on these markets, those generators that are more
accommodative (in terms of ramping and dispatchability in general) are naturally more
remunerative. This feature matters tremendously for the increased integration of renewables and
renewable standards, and carbon markets. Wind and solar generation are inherently intermittent
and often cannot match upswings in demand. As such, they are inherently disadvantaged in more
remunerative short-term markets. This also holds significant implications for increased
renewable utilization through renewable portfolio standards (RPS). As more renewable power is
required by environmental regulatory policies, the remainder of system-wide generation must be
increasingly more accommodative.
3. A Brief Primer on Emissions Markets
3.1 Background
The purpose of this section is to provide a brief background on the design of market-
based emissions programs, including both mature programs and contemporary auction-based
programs. Emissions markets are Coasian transferable property rights markets, designed to
realign incentives in product markets by internalizing pollution externalities. Economic theory
suggests that gains from trade can allow for this incentive realignment at the least possible
welfare loss(Coase, 1960). In other words, whatever level of aggregate emissions is desirable to
25
For a generation unit to be eligible to participate in the California Ancillary Services (AS) market, it must be able to
ramp to a pre-specified level within a very short, ten-minute duration.
89
achieve, doing so through a cap-and-trade system will minimize the cost of achieving the targeted
reduction. Depending upon the context, emissions markets have historically served society well
as a policy tool, with a few notable caveats (Arimura, 2002; Ellerman et al., 2000).
Contemporary emissions markets for global pollutants(e.g., greenhouse gases/GHGs),
vary in some important regards from emissions markets of the past. Emissions markets of the
past were designed at the regional or national scale to mitigate local and regional pollutants for
ecological or health protection, such as the mitigation of smog and ozone, or sulfuric deposition
(acid rain). Contemporary markets on the other hand, seek to mitigate global greenhouse
pollutants.
26
Historically, any severity of economic cost associated with emissions markets for
local pollutants has been ameliorated by significant offsetting behaviors, such as substitution,
technological innovation and mitigation.
27
There is much greater skepticism regarding similar
reductions in global pollutants, for which production relocation is not a direct option because they
affect the natural environment equally from any origin. As well, mitigation is much more
difficult for these pollutants because they are a natural byproduct of all forms of combustion and
there are no available end-of-pipe technologies to mitigate them completely. Nonetheless, there
are numerous sources of GHG emissions with varying costs of GHG emission reductions. It is
this variation in the costs of achieving reductions across emissions sources that suggests that
either a tax on emissions or a cap-and-trade program would be superior to a command-and-
control regime for achieving emission reductions in an overall best-cost manner.
26
These include carbon dioxide (CO
2
), methane (CH
4
), nitrous oxide (N
2
O), hydrofluorocarbons (HFCs),
perfluorocarbons(PFCs) and sulphurhexafluoride (SF
6
). A detailed list is provided in the IPCC’s 4
th
Assessment
Report, available at:
http://www.ipcc.ch/publications_and_data/publications_ipcc_fourth_assessment_report_synthesis_report.htm
27
These include fuel switching behaviors, end-of-pipe technologies and regional production shifts. The interested
reader is encouraged to see: Ellerman et al. 2000; 2010; and Tietenberg, 2006.
90
3.2 The Emissions Cap
A key element of any Coasian transferable property rights regime is the creation and
allocation of the property right itself. Emissions market designers and policymakers decide an
acceptable level of pollution ex ante, which is known as a “Cap.” Subdivisions of that cap are
divided down to an acceptable level, typically in single or thousand metric ton increments.
Pollution “permits” or “allowances” are then created in those denominated increments, and those
constitute the property right, which authorizes the holder to emit that quantity of pollution
without penalty.
Firms that are required to participate in the program, known as “Covered Entities,” must
“surrender” enough allowances to cover their individual quantity of pollution by the end of a pre-
specified compliance period. That compliance period is usually annual, or every 3-4 years,
depending upon the program. Firms that do not hold sufficient allowances must reduce their
output of emissions, purchase additional allowances, or pay a non-compliance penalty.
3.3 Allocation
The initial distribution, or initial “allocation,” of the property rights is a key design
element in any tradeable property rights regime. Historically, emissions markets have utilized a
method of initial allocation known as “grandfathering,” by which covered entities are freely given
a pre-specified quantity of allowances based roughly upon pro-rata reductions from their past
levels of pollution. Presumably, such grandfathering mechanisms have been used in order to
secure sufficient political support to achieve policy adoption. Furthermore, it is common for
these programs to be designed so that the aggregate cap and periodically allocated emissions
rights steadily decline over time, reflecting an increasingly stringent tightening of overall allowed
emissions. There have historically been some problems associated with this system, such as rent-
91
seeking, barriers to entry and mixed incentives (Arimura, 2002; Ellerman et al., 2000). This
method has also been the preferred method of allocation for national proposals in the US for
GHG mitigation markets, with some notable exceptions.
More contemporary market proposals and one key regional emissions market in operation,
have utilized a more market-based approach. This approach does not distribute the allowances
freely, but rather forces firms to compete in purchasing allowances in an auction. The benefits of
utilizing auctions are recognized within the literature and include efficiency and transparency
(Cramton & Kerr, 2002; Milgrom, 2004). However, according to theory, these auctions are
perfectly efficient only if market participants have symmetric valuations for emission permits, no
firms have market power, collusion among firms participating in the auction does not occur, and
all bidders are risk neutral with identically and independently-distributed (i.i.d.) private values
(Vickrey, 1961).
28
3.4 Compliance
Program compliance is also a key structural feature of any emissions market. Covered
entities are typically required to receive third-party or sponsored agency verification of emissions
levels, and at the end of each compliance period, “surrender,” or turn-in sufficient allowances to
meet those levels. Verification, in any emissions program, market-based or otherwise, can often
be costly to the firm and constitutes a significant transaction cost in a Coasian market, as firms
are often required to hire costly technical experts or consulting agencies to assist in the
verification process. The administering agency typically has statutory authority to ensure firm-
level compliance, and can ensure compliance by levying non-compliance penalties to firms that
28
Much of the literature on auction theory, a subfield of microeconomics and game theory has tested both theoretically
and empirically the nature and limitations of these key assumptions as they pertain to generalized auctions outside of
the scope of emissions markets. The interested reader is encouraged to see Klemperer (2004).
92
do not carry a sufficient quantity of allowances by the end of the compliance period. These
penalties are typically equivalent to the market cost of its allowance shortage, plus possible civil
penalties. The administering agency also has a public-relations tool at its disposal to ensure
compliance, as non-compliance can be reported publicly and the firm can be publicly labeled as a
polluting firm.
3.5 Inter-temporal Dynamics, Banking and Borrowing
The temporal dimension is also important. Participating firms know that the emissions
cap declines over time until it eventually reaches the aggregate target. As a result, firms can
expect an increasing degree of scarcity in the future emissions market, which will likely translate
into increasing allowance prices over time. This key design element necessitates a forward-
looking strategy on the part of profit-maximizing firms. There are some permissible strategies
available to firms to deal with the expectation of future declines in aggregate allowance supply.
These include proactive mitigation behavior, long-term contracting, substitution, and “banking”
of permits if this is allowed.
93
Title IV (U.S. Acid Rain
Program) Regional Clean Air Incentives
Market (RECLAIM) Regional Greenhouse Gas
Initiative Inc. (RGGI) E.U. Emissions Trading
Scheme (ETS) California Cap-and-Trade (AB
32) Location United States Southern California
Northeast and East Coast of
U.S.
Europe California
Covered Sectors Electricity Electricity and Industrial Electricity
Electricity, Industrial and
Transportation
Electricity and Industrial
Pollutant
Sulfur Dioxide & Nitrogen
Oxides
Nitrogen Oxides Carbon Dioxide & Equivalents Carbon Dioxide & Equivalents
Carbon Dioxide and
Equivalents
Time Horizon 1995-2005 1994-Present 2008-2018 2005-Present 2012-2020
Size of Market Approx. 175 Firms Approx. 300 Firms Approx. 40 Firms
Information not publicly
released
Information not publicly
released
Approximate No. of
Installations/Facilities
2000 400 650 10000 400
Allocation Method
Direct Allocation &
Auctioning in Small
Quantities
Direct Allocation Auctioning (Quarterly) Direct Allocation (Phase I & II) / Auctioning (Phase III) Direct Allocation &
Auctioning. (Portion
auctioned increases annually) Compliance Period Annual Annual 3 Year 3 Year
3 Year with annual surrender
provisions
Banking Allowed Not Allowed Allowed Allowed Allowed
Borrowing Not Allowed Not Allowed Not Allowed Not Allowed Not Allowed
Vintaging
Current and 7-year advance
vintage
Current, Forward and Infinite-
Year Block Credits (IYBs) Current and next compliance
period vintage
Current compliance period
vintage
Current and next compliance
period vintage
Offsets Not Allowed Not Allowed Allowed Allowed Allowed
Holding Restriction None None None None
Holding limit based on annual
allocation
Purchase Limit None None
Purchase limit 25% of any
auction
TBA. Phase III may include
Purchase limit 15% (excludes
electric utilities) of any
auction for covered entities,
and 4% for non-covered
entities (banks/hedge funds) Non-Compliance
Penalty
$2,000 (adjusted for inflation) per ton of shortage
Shortage credits deducted
from following year allocation
plus possible civil penalties
3 times the number of
shortage allowances must be
purchased, plus state-specific
penalties
1 times the number of
shortage permits must be
purchased, plus ¬100 per ton
penalty
1 times the number of
shortage permits must be
purchased
Table 3.1 Key Market Design Criteria of Major Cap-and-Trade Programs
94
Emissions market designs typically include some built-in structural features to also allow
firms to more adequately deal with the temporal dimension. Designs can include “banking” or
“borrowing” mechanisms.
29
If banking is permitted, the covered entity can retain an allowance in
the present and surrender it for compliance in a future period. This enables firms to purchase
allowances in early years when the cap is less stringent and allowance prices may be lower, hold
them at bay for a number of years, and then utilize them in future years when the cap tightens.
Firms can choose to make early reductions or substitutions and retain current term allowances for
use in future years. Evidence from the U.S. National Acid Rain Program suggests that banking
behavior can be quite significant (Ellerman & Montero, 2005). Borrowing, on the other hand,
occurs when a covered entity is permitted to surrender allowances in future years to cover
emissions in the current year. Although a less prevalent policy mechanism in emissions markets,
the option of borrowing can provide covered entities greater pricing certainty as a response to
market volatility.
29
Stevens and Rose (2002) discuss various incentives to banking and borrowing, including the discount rate and
technological change.
95
Figure 3.2 RGGI Auction Clearing Prices
3.6 Vintages
A less salient structural feature of many emissions markets is the use of “vintaging.”
Many emissions markets assign a vintage to allowances issued or sold that specify the earliest
compliance period, or year, in which the allowance can be utilized. For example, an allowance
issued with a 2015 vintage and with banking permitted, could be used to cover emissions in the
year 2015 or later, but not prior to 2015. Typically, forward vintages are sold in significantly
smaller quantities than current-term vintages. They serve as price discovery mechanisms for the
current period, and they have zero current compliance value. In the early RGGI markets in the
Northeast US, forward vintages began selling for an average of 15 percent lower than current
term vintages. RGGI recently discontinued the use of forward vintages altogether, because prices
consistently cleared at the price floor (reserve price). Forward vintages also offer covered entities
the opportunity to both hedge against future pricing uncertainty, as well as plan for long-term
1.75
1.95
2.15
2.35
2.55
2.75
2.95
3.15
3.35
3.55
3.75
Price ($/ton CO2e) Quarterly Auction Date
Current Vintage Forward Vintage
96
acquisitions. The power industry, in particular, has a long planning horizon and any forward
certainty has merit to its firms.
3.7 Offsets
Another planning and price-relief policy option available to covered entities in many
emissions markets is “offsets.” Offsets permit firms to emit GHGs in one location and make up
for that pollution with GHG reduction actions in another location. Depending upon the market,
these actions are typically required to be both “verifiable” and “additional,” which means that the
offset must be an approved action and witnessed by a third party and it must be an action that
would not have otherwise occurred in the absence of the offset’s issuance. Offsets, although
fiercely debated, can offer covered entities cost certainty and containment. However, if emissions
programs permit excessive offsets, this can have a demand-side effect and suppress emissions
permit demand, and as a result, emissions market prices. Offsets also provide an incentive for
program participation by firms not covered by the emissions program.
3.8 Leakage
Finally, GHG emissions markets typically incorporate statutory language to constrain or
disincentivize pollution behavior within the geospatial barriers of the regulated jurisdiction.
Covered entities may find that as a result of an emissions market, fiduciary incentives exist such
that it becomes less costly to simply circumvent the jurisdiction of the emissions market
altogether by moving production outside. For example, an LSE may find it less costly to
purchase power from a neighboring state than from a local producer who must factor allowance
costs into its production cost determination. Depending upon the regulatory mechanism used,
“leakage” constraining policies can have varying degrees of impact on adjacent power markets.
97
These policies typically ascribe generalized or weighted-average carbon content for imported
power, the responsibility for which is generally born by the first jurisdictional importer of that
power.
4. The Exercise of Market Power
4.1 Market Power and Misallocation
The exercise of market power can occur when influence in a market is sufficiently
concentrated in the hands of one or several participants who can use that influence to influence
prices or quantities within that market (Tirole, 1988). Market power among firms is a classic
concern in the electricity sector because that sector emerged out of traditional regulated natural
monopoly markets. Much of the literature addressing the potential exercise of market power in
emissions markets is dominated by examinations of the potential impacts of using grandfathering
as a method for directly allocating emissions allowances. Hahn (1984) provides a seminal
analysis of the issue. Hahn suggests that when there is an emissions market containing one or
more dominant firms and the government implements a direct allocation that over-allocates
(under-allocates) allowances to the dominant firms, those firms will become the monopolist
(monopsonist) from whom smaller fringe firms must purchase (sell) emissions allowances. His
analysis suggests that market power is shaped by the degree of initial misallocation, which can
transform dominant firms into either a dominant seller or dominant buyer who can reap excessive
rents by exploiting the inelastic portions of competitors’ demand curves.
Other scholars have suggested that dominant firms have even larger incentives to exercise
market power, because by doing so they can exercise “exclusionary manipulation” and increase
their market share, restrict entry into the market by would-be competitors, or raise costs to rival
firms (Misiolek & Elder, 1989; Rogerson, 1984; Salop, et al., 1983; 1984; 1987). Van Egteren
98
and Weber (1996) suggest that there is an upper bound on raising rivals’ costs, however; that their
costs can only go as high as the non-compliance penalty in the emissions market. Later analyses
have also factored into the market power equation the risk of non-compliance due to market
cheating (Chavez & Stranlund, 2003; Malik, 2002), and endogenously determined levels of
market power (Lange, 2012).
Implicit within much of this literature is the notion of hoarding allowances. This means
that a dominant firm exercising monopoly power over a market can exclude rivals outright, or
raise their costs by hoarding allowances to tighten the residual supply available to others. Or
equivalently, that firm can force its rivals to face the difficult calculus of either reducing their
output to match their holding of allowances-- which is difficult in markets like electricity where
demand is relatively price unresponsive-- or purchase those allowances from the dominant firm at
its monopoly rent.
4.2 Market Power and Strategic Bidding
In more contemporary auction-based markets, market power is exercised as strategic
bidding in either or both the wholesale electricity auction and the emissions auction, to inflate the
price of electricity and suppress the price of emissions allowances. In deregulated electricity
markets, strategic bidding is operationalized as supply disruptions that create financial shortages
in wholesale supply schedules (Hunt, 2002; Hortacsu & Puller, 2008; Joskow & Kahn, 2002;
Perekhodtsev & Baselice, 2010; Wolfram, 1998). Firms exercising this behavior can capitalize
on the inherent inelasticity of demand for electricity by adjusting their bids on marginal
generation to “withdraw” capacity from efficient dispatch and create financial shortages that shift
the supply curve upward. Essentially, firms exercise supply reduction by withdrawing capacity
from marginal points on the supply curve, and profit by the increase in aggregate market clearing
99
price on their remaining generation resources in operation.
30
This is colloquially referred to as
“hockey stick bidding.”
In auction-based emissions markets, strategic bidding is operationalized in nearly the
same manner, except it takes place as reductions in demand. Demand reduction is profitable
when dominant firms, that have heightened price influence in emissions auctions due to the
volume of their demand, reduce their demand bids to suppress the clearing price in emissions
auctions (Ausubel & Cramton, 2002; List & Lucking-Reiley, 2000; Porter & Vragov, 2006;
Webber, 1997). Firms with market power can seek to acquire a slightly reduced quantity of
emissions allowances at a significantly suppressed price. Chapter 1 provides a detailed treatment
on this behavior. The misallocation problems of grandfathering approaches, described above,
become equally problematic in auction-based emissions markets if the exercise of market power
results in an over or under-allocation to dominant firms.
Figure 3.3 Competitiveness Measure of Past RGGI Emissions Auctions (HHI) HHI calculated based on auction awards data. Available at: http://www.rggi.org/market/co2_auctions/results
30
This same effect can also be achieved through physical withholding of electricity generation.
0
1,000
2,000
3,000
4,000
5,000
6,000
Herfindahl-Hirschman Index (HHI) Quarterly Auction Date
HHI Current Vintage HHI Future Vintage
100
4.3 Market Power in Operating Markets
Because of the inherent concern for market power and its exercise in the electricity and
emissions markets, a variety of measurement methodologies have been created to estimate the
potential for market power. Structural and conduct measures have helped provide insights into
the potential for market power and its exercise. There are shortcomings associated with each
measure, and no measure is uniformly utilized across all regulatory and oversight institutions.
The most common structural measure, utilized by the US Department of Justice, is the
Herfindahl-Hirschman Index (HHI).
31
It is measured as the sum of each firm’s squared market
share, and ranges from 10,000 in the case of a perfectly unitary monopoly to nearly zero in the
case of an atomistic market. Typically any market with an HHI below 1,000 is considered
unconcentrated, and any HHI between 1,000 and 1,800 is considered moderately concentrated.
HHIs above 1,800 are considered concentrated.
Figure 3.3 provides HHI results from past emissions auctions for both current term and
forward vintages, for the operating RGGI emissions market. Published results of the auctions
suggest that auction earnings are consistently concentrated in the hands of a few large winning
bidders. Forward vintage auctions have been significantly concentrated, with approximately 60
percent fewer firms participating in those auctions. At the same time, auction prices have
consistently cleared at the lowest possible price (reserve price), as shown in Figure 3.2.
The HHI is a particularly robust measure for emissions auctions because it directly
corresponds to quantities and holdings of emissions allowances. However, it is a particularly
poor indicator of market power in electricity auctions because it provides little insight into the
spatial and temporal nature of market power in the supply of electricity. In electricity markets, as
discussed above, during periods of peak demand, a single supplier of electricity, or a small group
31
The Herfindahl-Hirschman Index is given by the formula: HHI =
2
1
N
i
i
s
=
å
where s
i
is firm i’s market share.
101
of suppliers, can be pivotal, when they would otherwise not be during periods of relaxed demand.
For example, in hot summer afternoons in California when demand moves toward system-wide
capacity limits, a single firm owning generators that make up as little as 5 percent of the total
supply may be a pivotal supplier who can set prices system-wide. As the market share of supply
owned
32
by dominant firms increases, it requires less severe demand scenarios for a single firm or
small group of firms to be pivotal.
Similarly, because of inherent transmission and generation limitations at certain spatial
locations of the grid (i.e., because ISOs utilize “security-constrained economic dispatch”), a
single supplier, or small group of suppliers, can be pivotal in specific physical locations along the
grid, at specific periods of time. This is often referred to as “local market power,” because a
firm’s ability to exercise market power is not system-wide, and results from “load pockets” that
exist on the grid. These pockets are created in locations where competing generators cannot
physically deliver generation due to inherent limitations in transmission capacity.
Given these real-world limitations, regulators often utilize an alternative measure of
market power known as the Residual Supply Index (RSI).
33
This index measures the degree to
which a single supplier, or a small group of suppliers, is pivotal in the supply of generation. Any
value below 1 indicates that after accounting for the supply capacities of all other firms, demand
cannot be met without a portion of the largest single (or multiple) supplier. Market monitors
often consider RSI
1
, RSI
2
, RSI
3
. That is, how much of the demand for electricity in a given load
32
Market power can also be exercised on generators that a firm does not own. That firm simply needs to have
scheduling authority over that generator to bid its supply strategically into the market. This occurs in the case of tolling
contracts, discussed below.
33
The Residual Supply Index is given by the formula: RSI
i
=
1
n
i
i
c c D
=
-
å
, where c
i
is the capacity of supply of firm i,
D is demand, and
1
n
i
c
=
å
is the total supply capacity of all firms. Some regulatory oversight agencies, such as Texas
ERCOT, utilize a similar measure known as the Residual Demand Index (RDI).
102
pocket for a given hour of the day (or alternative time interval for real time markets) cannot be
met without using any of the generators owned by the largest single company, two largest
companies, or three largest companies, respectively. This provides a much more appropriate
measure of market power in real world markets than simple HHI measures of concentration.
The California markets have made significant headway in recent years in implementing
additional capacity and regulatory tools to mitigate market power. In 2009 the FERC ordered a
market redesign and technology upgrade (MRTU) that includes many of the components
discussed in the above primer, such as nodal pricing, financial transmission rights
34
(FTRs) and
convergence markets which are discussed below. According to internal market monitoring
records, in 2010 the CAISO assessed an RSI
1
below the critical value of 1, (i.e., a single supplier
is pivotal over a load pocket for a given hour), 5 percent of the hours throughout the year.
35
By
contrast, this value was below the critical value during 53 percent of the hours throughout the
year in 2009. The RSI
2
dropped below the critical value 6 percent of the hours, whereas in 2009
it dropped below that value 60 percent of the hours (Hildebrandt et al., 2011).
Comparing competitiveness measures for other ISOs also provides some insights. In
New England, Dominion Energy is a dominant energy producer, and controls 16 percent of the
regional generation of electricity. And, Next Era Energy controls 11 percent. In a 5 month
assessment, ISO-NE found that the RSI
1
dropped below the critical value of 1 during 223 hours
during that 5 month period (ISO New England, 2011). They also found that the concentration of
the market increased between 2009 and 2010. There are two key load pockets in the ISO-NE grid,
one in Connecticut and one in the Boston area. In the Connecticut pocket, a single supplier was
pivotal 15 percent of the time, and in Boston, 37 percent of the time. This figure is significantly
34
These are called Congestion Revenue Rights (CRRs) in California.
35
One caveat with these assessments is that the indices are measured excluding the portfolios of IOUs.
103
higher in the Texas ERCOT market, where a single supplier was pivotal during a total of 60
percent of the total summer hours in 2010, and 53 percent of the total non-summer hours
(ERCOT, 2011). And a 2011 study for the PJM market on the East Coast assessed the markets to
be “not competitive” during hours of peak and intermediate electricity demand (PJM, 2012). In
some peak markets there is a single producer with 30 percent of the share of the market. Using a
time-dependent HHI index rather than an RSI, their analysis suggests that the market is
moderately concentrated 98.4 percent of the hours throughout the year.
5. Virtual “Convergence” Bidding
5.1 Purpose
Because of this inherent liquidity problem that occurs in load pockets or in the aggregate
during periods of intermediate and peak electricity demand, U.S. wholesale markets have begun
to utilize a market-based approach to improve the liquidity of these markets and disincentivize the
exercise of market power. One of the ways market monitors assess the efficiency and
competitiveness of these markets is by evaluating the difference in prices between forward
(DAM) and real-time markets. Scholars have noticed historical divergences in prices between
forward and real-time wholesale markets(Hadsell, 2007; Ott, 2003; Saravia, 2003). Because a
majority of the power supplied in real time is committed through the day-ahead market (DAM), a
firm’s ability to exercise market power in the DAM matters significantly for both competitiveness
and market efficiency. When forward markets fail to accurately capture real-time supply and
demand dynamics, inefficiencies occur, and firms have an incentive to arbitrage between the
markets or exercise market power.
In an effort to improve the efficiency and competitiveness of these markets, policymakers
have added an additional market-based mechanism to day-ahead markets, known as virtual
104
(convergence) bidding. Virtual bidding attempts to make these markets conform more closely to
other commodity markets by allowing virtual trading to occur, to incentivize price convergence
between the DAM and RTM markets. Virtual bidding occurs in other commodity futures markets
with multi-period settlement, and allows speculators such as banks and hedge funds, to participate
in the market without an actual physical position in the market. This means that they can bid to
supply or purchase power from the market without ever producing or consuming an actual
megawatt.
Virtual bids take the form of either a virtual demand bid or a virtual supply bid. A virtual
demand bid is a bid to buy a quantity of power at the day-ahead price and sell that same quantity
of power back in real time at the real-time price. A virtual supply bid is a bid to sell a quantity of
power at the day-ahead price and buy that same quantity of power back in real-time at the real-
time price. Virtual demand bids are considered “long” positions in the DAM, and virtual supply
bids are considered “short” positions. These bids are placed at any single node in the market, and
require no physical generation capacity. That is, in addition to market participants with a
physical position in the market, marketers, banks and hedge funds can also place these bids.
They are purely financial instruments, but they can and do affect the LMP for power in the DAM
and can impact how physical generation is supplied in the DAM. Because both long and short
positions in the DAM include the opposite position in the RTM, the bids are liquidated in real-
time and do not impact the actual supply of power that occurs in real time.
Virtual bids are only placed in the DAM, but because they are clearly identified as virtual
bids, they do not affect the scheduling feasibility (i.e., security constrained dispatch) of the
market. However, virtual bids can win in the auction against physical generation. Virtual bids are
also subject to the same bid caps as bids for physical generation. Virtual bidding offers an
incentive to arbitrage between the two markets and capitalize on any price differences between
105
the DAM and RTM. As such, these markets are also known as “convergence” markets because
they encourage arbitrageurs to take a mix of long and short positions until the two market prices
converge to the same LMP.
That is, if one market (e.g., DAM) offers higher prices than another, it creates incentives
to buy in the opposite market, thus creating an incentive for price convergence across multi-
settlement periods (Celebi et al., 2010). The NYISO and PJM-ISO began providing for virtual
bidding in their 2001 Standard Market Design (SMD), ISO-NE in 2003, and MISO in 2004.
CAISO recently began accepting virtual bids by order of FERC. Scholars have found that virtual
bidding increases price convergence across markets (Celebi et al., 2010; Hadsell, 2007; Saravia,
2003). And ISOs and FERC alike have endorsed virtual bidding as an improvement in both
liquidity and price convergence.
36
5.2 Virtual Bidding and Market Power
A firm that expects prices to be higher in the DAM than in the RTM will take a short
virtual position by placing a virtual supply bid. This firm will be paid at the day-ahead price and
charged at the real-time price. A firm that expects prices to be higher in real time will take a long
virtual position by placing a virtual supply bid. This firm will be charged at the day-ahead price
and paid at the real-time price. Firms placing virtual bids will make this determination in
response to a variety of factors, including prior beliefs about the exercise of market power.
36
Federal Energy Regulatory Commission Order, Docket No. ER03-1101-000, issued on Sept. 22, 2003. Federal
Energy Regulatory Commission Order, Concurring Opinion of Wood and Kelliher, in Docket ER04-121-000, issued on
Jan. 15, 2004. Market Surveillance Committee, Memorandum to the ISO Board of Governors, dated Jan 19, 2005,
available at http://www.caiso.com/docs/09003a6080/34/71/09003a60803471a9.pdf. NYISO, State of the Market, 2002,
available at http://www.nyiso.com/public/documents/studies_reports/market_advisor_reports.jsp. ISO-NE, Impact of
Virtual Transactions on New England’s Energy Market, issued Nov. 1, 2004, available at http://www.iso-
ne.com/pubs/spcl_rpts/2005/vrtl_trns/index.html.
106
In operating electricity markets, firms can have market power on both the supply side and
the demand side. Suppliers of power such as IPPs can have market power, and will exercise it to
increase the LMP. As well, buyers of power, which include the LSEs, can also have market
power and will exercise it to suppress the LMP. Whether firms with market power exercise
monopoly or monopsony power depends upon the conditions of a given load pocket or demand
characteristics, as discussed above. Virtual bidding can help mitigate market power on both sides
in the DAM.
In the case of a buyer exercising monopsony power, a buying firm such as an LSE can
bid strategically to withhold demand from the DAM to suppress the forward market price. Load
is often of significant size to be pivotal in setting prices. This strategy requires LSEs to suppress
the price in the DAM, from where they buy the bulk of their power, and then acquire the
remainder of that power in real-time at a moderately higher price, and it has been well
documented (Borenstein, et al., 2006; Isemonger, 2006). Virtual bidding can serve to ameliorate
this form of market power by allowing competing firms to place virtual bids and extend the
demand curve in the day-ahead market. Virtual demand bids are equivalent to purchases of
power, in a financial sense, and serve to counteract a monopsonist’s under-scheduling of demand.
In fact, virtual demand bids counteract demand reduction. It should also be noted that virtual bids
are not constrained by congestion or load pockets, because virtual bids do not induce congestion.
So, a firm exercising this form of market power in a load pocket would still face competition in
the DAM by virtual demand bids outside of that load pocket.
In the opposite case of a supplier exercising monopoly power in the DAM, an IPP or
other generating firm can strategically bid to withhold generation from the DAM to inflate the
forward market price, as discussed above. When this firm offers that generation for sale in real-
time, the real-time price will be lower than the day-ahead price, relative to real-time demand.
107
Withholding supply here shortens the supply curve. Thus, an incentive is created for a competing
firm to place a virtual supply bid, to lengthen that supply curve. The firm placing the virtual
supply bid is essentially selling a quantity of power in the DAM and then buying that quantity
back in the RTM at a lower price. In other words, that firm is selling high and then buying low,
and in so doing, suppressing the price in the DAM. Mitigation of market power through virtual
supply bids will only work, however, in the event that the withholding supplier is only
withholding its bid in the DAM, and not physically withholding generation in both the DAM and
RTM.
Given that virtual bidding provides a unique market-based approach to improving the
liquidity and competitiveness of wholesale electricity markets, virtual bidding has also provided
for some design difficulties. Early virtual bidding designs in eastern ISOs created incentives for
firms to manipulate congestion charges and transmission pricing (Celebi, et al., 2010). More
modern ISO markets include another market-based approach to pricing congestion charges
through the use of financial transmission rights (FTRs), called congestion revenue rights (CRRs) in California. These markets provide for another Coasian transferable property rights market,
entitling the owner of the right to the revenue from congestion charges on a given section of the
transmission grid. Policymakers learned that incentives were created for firms to manipulate
virtual bids to inflate FTR revenue by using virtual bids to induce congestion where FTRs were
owned. Although the details of this practice are beyond the scope of this chapter, it is mentioned
for completeness. ISOs have utilized rules and market optimization software changes to mitigate
this behavior, however.
108
6. Consignment Auctions
6.1 Design Mechanism
The design mechanism of an emissions auction can also play heavily into market power
dynamics. Analyses conducted for purposes of market design have experimentally tested
alternative auction formats (Burtraw et al., 2008; Holt et al., 2007), to test the assumptions of
revenue equivalence and bidding behavior across standard formats of auctions. California’s AB
32 auction mechanism, on the other hand, provides an altogether different structure of auction
that is yet to be tested empirically, as it uses a consignment auction mechanism.
The basic auction format of the AB 32 auction is the same uniform-price sealed-bid
auction format utilized in RGGI and under most national proposals. It is also the most commonly
utilized auction format in wholesale electricity auctions. However, because it is a consignment
auction, the use of auction revenue is routed differently, and as a result, can alter the incentive for
dominant firms to exercise market power through strategic demand reduction. The consignment
mechanism cannot however, disincentivize the exercise of market power through hoarding and
raising rivals’ costs, which is the other strategy discussed above.
37
Figure 3.4 provides a flow chart of the distribution of allowances and allowance revenue
under both the RGGI auction design and the California AB 32 consignment auction design.
Under the RGGI auction mechanism, member state governments decide an emissions cap and
allocate allowances in accordance with that cap. They then place those emissions allowances into
a centralized quarterly auction maintained by the RGGI Inc. nonprofit organization. Emitting
firms then bid in these auctions to purchase the allowances that they require, and the revenue
from the sale of the emissions allowances is directed to state governments. The state
governments then use this revenue to fund green development and renewable energy projects.
37
See Section 7 for a discussion of an approach to minimize hoarding behavior.
109
Figure 3.4 RGGI and AB 32 Auction Mechanisms
Under the AB 32 consignment mechanism, the California Air Resources Board (CARB) distributes emissions allowances for the electricity sector directly to the state’s three main
investor-owned utilities (IOUs). These include Sempra Energy (San Diego Gas & Electric),
Pacific Gas and Electric, and Southern California Edison. The IOUs are then required to post all
of those allowances in quarterly consignment auctions, in which they compete with independent
power producers(or other interested buyers), who supply approximately half of the state’s
110
electricity needs, to purchase emissions allowances. The revenue generated from the sale of the
emissions allowances is then given directly to the IOUs to use toward offsetting the costs to
businesses and households of the cap-and-trade program.
38
6.2 Market Power Incentives
Under the RGGI non-consignment design, all emitting firms, even those firms that do not
have price influence, have an incentive to push emissions prices downward. Lower emissions
market prices mean lower operating costs. Firms with market power have an incentive to
exercise strategic demand reduction to suppress the price in the emissions market, as discussed
above. However, under the CARB consignment auction design, this incentive is neutralized,
because the IOUs receive the revenue from the sale of the emissions allowances. Why would
they seek to suppress the price in the emissions auction when they retain that revenue? The
routing of the auction revenue to the IOUs, who are major demanders of emissions allowances,
effectively neutralizes incentives to exercise market power through strategic demand reduction.
This neutralization comes from the direct relationship between revenue recovery and
auction performance. The main form of cost recovery to IOUs from costs associated with the AB
32 market comes from the revenue they receive in the consignment auctions. However, IOUs
still have access to alternative regulatory cost recovery mechanisms. As well, proposed rules by
the CARB would allow any additional costs to IOUs of purchasing allowances to flow through in
Energy Revenue Recovery Accounts (ERRA), which are contemporary fuel cost adjustment
mechanisms. Linking auction revenue and IOU cost recovery helps to ensure that IOUs have an
incentive to competitively bid toward an efficient market price, rather than demand reduce. This
38
AB 32 Final Regulation Order, California Code of Regulations Title 17, §95800-96023.
111
helps to ensure that the resultant market price is closer to the socially-efficient price, which will
send sustainable production signals to electricity producers.
Of course, this assertion is highly dependent upon several factors of a highly complicated
market. The key determinant of whether the IOUs will have an incentive to suppress the
consignment auction price is whether they are net buyers or sellers of emissions allowances. If
they are net sellers, they will have the opposite incentive, and that is to inflate the emissions price.
If they are net buyers, however, they will have the incentive to demand reduce and suppress the
price of allowances. Due to the complexities of the market, the IOUs cannot presently determine
at this time whether they will be one or the other.
39
Their net status depends upon their sources of
allowance demand, discussed below.
6.3 Sources of Demand
LSEs’ demand for emissions allowances comes from three main sources, which can
fluctuate from year to year. The first source is the portfolio of generation assets that they own.
These are power plants and generation facilities that they own and for which they must meet
compliance obligations. Under the original deregulatory plans of the state, the IOUs were forced
to divest much of their generation assets. In terms of aggregate greenhouse gases, retained
generation represents a relatively small portion of the greenhouse gases from the electricity sector,
because these include significant non-emitting technologies such as nuclear, hydro and
renewables. If this were the only source of demand for allowances on the part of the IOUs, then
they IOUs would be net sellers of allowances.
40
39
Dagli, Dhaval, Manager of Regulation and Compliance, Southern California Edison, March 8, 2012, personal
communication.
40
The specific allocation to all LSEs is provided in the AB 32 regulation order §95892 (see Footnote 38).
112
The second source is through contractually-owned power arrangements operated through
long term tolling contracts. These are generation assets that the IOUs do not own, but for which
they have entered into a contractual relationship similar to a lease. These typically stipulate that
the IOU will provide the generator fuel resources and that the generator must provide the IOU
with the electricity produced. The IOU typically becomes the “scheduling coordinator” and
controls the scheduling, or bidding, of that generator into the ISO markets. Tolling contracts
become part of the IOUs portfolio of generation assets. Under tolling agreements, fuel and other
generation costs, including emissions allowances, are usually required to be supplied by the IOU.
Presently, the IOUs have large open positions with tolling facilities that represent a significant
portion of allowance demand. When these tolling contracts are factored in, the IOUs may come
very close to being net buyers of emissions allowances.
The third source is through other forms of market exposure, mainly imported power from
out of the state. Imports represent more than 30 percent of California’s total electricity
production (California Energy Commission, 2010). Under the AB 32 implementation, power
imported from out of the state is assigned an “emissions factor” for generic power imports, or a
unit-specific emissions factor for long-term contractual power imports. Taken together, these
three sources of demand for emissions allowances may be substantial enough for the IOUs to be
net buyers of emissions allowances.
Demand for emissions allowances here is mentioned specifically in the context of IOUs,
rather than other LSEs such as Co-ops, POUs and competitive retailers, because the CARB
consignment auction design directly targets the incentives of IOUs to the exclusion of the other
LSEs. IOUs have an incentive and a duty to their customers and shareholders to respond to
market incentives in a way that is both socially-responsible as well as profit maximizing. Given
this, IOUs have sufficient wiggle room in the long run to adjust their tolling contract positions as
113
well as their own generation to meet these aims. In the absence of the consignment mechanism,
such as in a market like the RGGI, all generating firms, including IOUs have an incentive to
suppress the price of emissions allowances. The consignment mechanism offers a robust
improvement in the status quo of auction-based emissions allowance allocation.
7. Multilevel Accounts
7.1 Purchase Limits
The AB 32 design provides for another unique policy approach to mitigating market
power, through the use of multilevel holding accounts with both holding and purchase limits. As
mentioned above, one common strategy pursued by dominant firms is that of using market
dominance to acquire, or purchase a large sum of emissions allowances, drive up scarcity in the
market, and then either use that scarcity to raise rivals costs or profit from resale arbitrage in
periods of scarcity (Hahn, 1984; Misiolek & Elder, 1989).
One direct approach utilized by market designers has been to create rigid purchase limits
in emissions auctions. These purchase limits can have the salutary benefit of restricting any one
(non-colluding) firm from maintaining a pivotal market position. If these limits are set too tightly,
however, they can have the negative side effect of both restricting market liquidity and forcing
compliance entities onto the secondary market to acquire large quantities of allowances, which
may make them monopsonists over the secondary market.
Of the major markets considered in this chapter, the only markets that presently utilize
purchase limits are the RGGI market and the California AB 32 market, largely because they are
the only operating auction-based markets. Under a direct allocation system, on the other hand,
there are no purchase limits in the initial allocation process. The purchase limit under RGGI is
straightforward. It is relatively unrestrictive, set at 25 percent of any quarterly auction. Taken
114
together in any year, a single firm can purchase a total of one quarter of the total ten state
allowance allocation. On the other hand, the system under the California Air Resources Board’s
AB 32 program is far more complex, and provides a number of design mechanisms that have the
potential to systematically limit strategic purchases, presumably as a means for limiting the
potential exercise of market power. Figure 3.5 provides a diagram of the multilevel accounts
methodology under AB 32.
7.2 CARB’s Multilevel Accounts Approach
First, the CARB directly provides an allocation of emissions allowances to compliance
entities, where the bulk of emissions allowances for the electricity sector go to the state’s three
main investor-owned utilities (IOUs) and the larger Publicly-Owned Utilities
41
(POUs)( §95892).
CARB also provides a large sum of emissions allowances directly to the AB 32 consignment
auction, the revenue from which goes directly to the State of California’s Air Pollution Control
Fund (§95912(k)(2)(c)).
There are three main firm-level accounts: a Limited Use Holding Account, a Holding
Account, and a Compliance Account. The allocated allowances are placed into each firm’s
Limited Use Holding Account, and must be consigned to the allowance auction from that account
(§95831). POUs and cooperatives can however, choose to have their allocations placed directly
into their Compliance Account (§95892(b)(2)). The Compliance Account is the account from
which allowances are surrendered to meet compliance obligations, and once an allowance is
placed into that account, it cannot be pulled back out (§95831(a)(4)(b)). This is a key structural
feature that limits firms from buying up allowances for the purposes of large resale arbitrage into
a scarce market.
41
The largest POUs are the Los Angeles Department of Water and Power (LADWP), the Sacramento Municipal Utility
District(SMUD) and the City of Anaheim.
115
Figure 3.5 Flow Chart of CARB Multi-level Accounts
Second, each firm also has a third type of account, which is a general Holding Account
(§95831(a)(2)). Emissions allowances purchased at the auction go directly into a firm’s Holding
Account (§95831(b)(2)(b)). The holding account provides each firm with flexibility to bank
allowances across time, or trade them with other firms on the secondary trading market. In
previously established cap-and-trade programs, these secondary market trades often take place
through third party brokerages. Firms place allowances from their Holding Account into their
Compliance Account to meet compliance obligations (§95856(c)). Holding Accounts have a key
structural component that limits the exercise of hoarding behavior. That is, Holding Accounts are
116
subject to a strict holding limit, given by a complicated formula (§95920(d)(1)). This limit is
approximately four percent of the total emissions allocation for the year. I provide each year’s
Holding Account holding limit in Figure 3.6.
Non-compliance entities, such as commodities traders, banks and hedge funds who also
participate in the auctions and secondary market, are given a different type of holding account; an
Exchange Clearing Holding Account (§95831(a)(5)). These accounts are excluded from Figure
3.2, for simplicity.
7.3 Purchase Limits under AB 32
And finally, the quarterly auctions have strict purchase limits for firms, which also have
the potential to limit the exercise of market power through dominant purchases. Compliance
entities, such as Independent Power Producers (e.g., AES, Dynegy) can only purchase up to 15
percent of the total quantity of allowances sold in any auction (§95911(c)). Non-compliance
entities have a purchase limit of 4 percent. Utilities (IOUs, POUs and Coops) however, are not
subject to a purchase limit ((§95911 (c)(4)(b)). However, when the allowances that any firm
purchases enter into their Holding Account, they are then subject to the holding limit. Therefore,
unless a firm then places those allowances into its Compliance Account, from which the
allowances cannot be taken out, the holding limit directly circumscribes that firm’s ability to
hoard allowances. As such, it directly circumscribes that firm’s ability to hoard allowances and
then later sell them off into a scarce, high-priced market.
117
Figure 3.6 Single Firm Holding Limit under AB 32
Petroleum refiners covered beginning in year 2015
7.4 Limits on Associations
One potential loophole for exercising market dominance beyond the multilevel accounts
and holding limits is for firms to cooperate with each other or create levels of corporate
subsidiaries. However, the CARB AB 32 regulatory structure also possesses a strong regulatory
mechanism to limit cooperative behavior that would otherwise be designed to circumvent the
holding and purchase limits. The AB 32 regulation calls for full disclosure of both direct and
indirect corporate associations (§95833).
The auction purchase limit applies to any entities with a direct or indirect corporate
association (§95911(c)(1)). And, the holding limit, of approximately 4 percent, applies to any
entities with a direct or indirect corporate association (§95920(a) & §95920(f)(1)). As such, the
corporate association disclosure mechanism effectively circumscribes cooperative behavior
designed to circumvent either the auction purchase limit or the holding limits. This would limit
the ability for any group of two, or more, independent power producers (IPPs) to cooperatively
0
50
100
150
200
250
300
350
400
Tons of CO2e (millions) Compliance Year
Annual Allowance Budget Single Firm Holding Limit
118
purchase allowances beyond the purchase limit of 15 percent, or hold allowances beyond the
holding limit of approximately 4 percent of the total allowance budget. This would also limit the
ability for any firm to create a dummy corporation simply for the purpose of circumventing the
holding or purchase limit. This would not, however, limit fraudulent corporations created without
either disclosed or traceable association. And, it would not limit outright collusion among firms.
7.5 Structural Limits on Market Power
Taken together, the use of multilevel accounts with both holding and purchase limits
provides for a strong design limitation on the exercise of market power through hoarding
emissions allowances. Placing firm holding limits on a single account would limit the flexibility
of any firm, or utility, to meet its compliance obligations. But, creating multiple accounts in the
manner that CARB has, effectively allows firms to both respond to the inherent stochastic
demands of electricity markets, and limits their ability to hoard permits and raise rivals costs or
exclude entry by rival firms. The use of a Limited Use Holding Account that has a consignment
requirement structurally mitigates many of the allocation inefficiencies of prior markets, because
it does not allow misallocation, if any, to send any price signal to other producers. Strict purchase
limits in the auction limit monopsony or oligopsony power by dominant firms, except the IOUs
who are not subject to them. The requirement that the allowances purchased at auction be kept in
a Holding Account with a holding limit, restricts all firms, even IOUs, from hoarding allowances.
And, the creation of a Compliance Account without holding limits effectively provides firms with
a safe location to bank allowances, and the prohibition of pulling allowances out from the
Compliance Account limits firms from using their holding of allowances as a war chest.
These structural mechanisms may however, have two unintended consequences that can
be tested empirically when data becomes available. First, the holding limit for the Holding
119
Account may provide efficiency constraints on the secondary trading market. Coasian theory
suggests that market inefficiencies can dissipate by trading behavior. The four percent holding
limit may place a strict quantity limit on the secondary market and limit the development of a
robust secondary market, which may be a key component of efficient price discovery. Second,
because the IOUs do not have purchase limits in the auction and can place purchased allowances
into their Compliance Account, there is no direct mechanism to prevent them from hoarding
excluding resale. That is, they can still buy up a large chunk of emissions allowances and place
them into their compliance account, which does not have a holding limit, simply for the purposes
of raising their rivals’ costs. There is tremendous disincentive for this strategy, however. This
strategy would incur a cost to the IOUs and make them net buyers of allowances, and it would
raise costs to independent power producers from whom they purchase power.
8. Conclusion
This chapter began by providing a concise summary of the post-restructuring
mechanisms and policy approaches that have been taken by policymakers to constrain some of
the inherent market failures of deregulated electricity markets. Today, these markets are defined
by a complex array of multi-settlement markets with additional market mechanisms to constrain
strategic behavior in transmission pricing. This chapter then provided a concise summary of cap-
and-trade programs that have added further complexity to the wholesale electricity markets.
Within the context of the potential exercise of market power, this chapter provided an analysis of
three contemporary regulatory mechanisms and policy approaches that have been under studied in
the literature on either electricity market design or market power.
Virtual (convergence) bidding has recently been implemented in regional electricity
markets as a tool to converge price differentials between multi-settlement markets and constrain
120
market power in load pockets. Under efficient virtual bidding, the incentive to raise prices in one
forward market is neutralized by the incentive to capitalize on price differentials across multi-
settlement markets. However, virtual bidding cannot constrain systemic market power that is
exercised across all multi-settlement periods.
Price convergence across markets may provide significant implications for the utilization
of renewable generation. As renewable power is intermittent and lacks sufficient ramping
response to meet diurnal load cycles, it is inherently disadvantaged in more remunerative spot and
ancillary services markets, as it may only be able to participate in day-ahead forward markets or
through long-term contracts. Price convergence across markets may effectively make renewable
participation in day-ahead markets more remunerative, and this may also send positive
investment signals to producers regarding the development of renewable power beyond
renewable portfolio mandates.
Consignment auctions can effectively neutralize the incentives of dominant firms to
exercise market power in emissions auctions through strategic demand reduction. Dominant
firms that can exercise price influence can suppress prices in emissions auctions, unless their
incentive to do so is reduced by a policy mechanism. The CARB AB 32 consignment mechanism
provides the revenue from a significant share of the emissions allowances sold at auction, directly
to the Investor-owned Utilities who would otherwise be large buyers in the market, to use toward
defraying retail costs to end users. By directly linking the cost recovery mechanism with the
consignment auction, dominant buyers have a reduced incentive to suppress prices in emissions
auctions.
Multilevel accounts with strategic holding and purchase requirements can also neutralize
the incentive to exercise market power through exercising either monopoly or monopsony power.
The CARB AB 32 design mechanism creates a multilevel set of holding accounts that each have
121
specific utilization requirements and holding limits that gives firms sufficient flexibility to meet
compliance in stochastic electricity markets, yet disincentivizes the exercise of market power
through strategic hoarding to raise rivals’ costs or to profit from significant resale arbitrage into
high-priced scarcity markets. Although yet to be tested empirically, this design may however
constrain efficient price discovery in secondary trading markets as it constrains the overall
volume of the secondary market.
122
References
Arimura, T. H. (2002). An empirical study of the SO2 allowance market: effects of PUC
regulations. Journal of Environmental Economics and Management, 44, 271-289.
Ausubel, L., & Cramton, P. (2002). Demand reduction and inefficiency in multi-unit auctions
(Working Paper). Retrieved from University of Maryland:
http://www.cramton.umd.edu/papers1995-1999/98wp-demand-reduction.pdf
Bernard, J. C., Mount, T., & Schulze, W. (1998). Alternative auction institutions for electric
power markets. Agricultural and Resource Economics Review, 27(2), 125-131.
Borenstein, S., Bushnell, J., Knittel, C., and Wolfram, C. (2006). Inefficiencies and market power
in financial arbitrage: A study of California’s electricity markets. Retrieved from
http://web.mit.edu/knittel/www/papers/arb_jie_2.pdf
Bovenberg, A. L., & de Mooij, R. (1994). Environmental levies and distortionary taxation.
American Economic Review. 84, 1085-1089.
Bovenberg, A. L., & Goulder, L. H. (1996). Optimal environmental taxation in the presence of
other taxes: general equilibrium analyses. American Economic Review, 86, 985-1000.
Brennan, T. J. (2007). Preventing monopoly or discouraging dompetition?: the perils of price-cost
tests for market power in electricity. In A.N. Kleit (Ed.), Electric choices: Deregulation
and the future of electric power(163-180). Oakland: Independent Institute Press.
Burtraw, D., Goeree, J., Holt, C.A., Myers, E., Palmer, K., & Shobe, W. (2008). Collusion in
auctions for emissions permits: an experimental analysis(Working Paper DP 08-36).
Washington, D.C.: Resources for the Future.
California AB32 Final Regulation Order, California Code of Regulations Title 17 §95800 --
96023, (2011).
California Energy Commission. (2010). Energy almanac. Retrieved from:
http://energyalmanac.ca.gov/electricity/electricity_generation.html
Cason, T. N., & Gangadharan, L. (2006). Emissions variability in tradable permit markets with
imperfect enforcement and banking. Journal of Economic Behavior & Organization,
61(2), 199-216.
Celebi, M., Hajos, A., & Hanser, P. Q. (2010). Virtual bidding: The good, the bad and the ugly.
The Electricity Journal, 23(5): 16-25.
Chavez, C. A., & Stranlund, J. K. (2003). Enforcing transferable permit systems in the presence
of market power. Environmental and Resource Economics, 25, 65-78.
Coase, R. (1960). The problem of social cost. Journal of Law and Economics, 3, 1-44.
123
Cramton, P., & Kerr, S. (2002). Tradeable carbon permit auctions: how and why to auction not
grandfather. Energy Policy, 30(4), 333-345.
Dewees, D. N. (2008). Pollution and the price of power. The Energy Journal 29(2), 81-100.
Dormady, N. (2012). The political economy of collaborative organization. Administration &
Society, forthcoming.
Dormady, N., & Maggioni, E. (2009). Climate change mitigation policy and energy markets:
Cooperation and competition in integrating renewables into deregulated market. In D.
Mazmanian (Chair), Forging closer ties: Transatlantic relations, climate and energy.
Symposium conducted at the Freie University, Berlin.
Downward, A. (2010). Carbon charges in electricity markets with strategic behavior and
transmission. The Energy Journal, 31(4), 159-166.
Ellerman, A. D., Convery, F. J. & de Perthuis, C. (2010). Pricing carbon: The European Union
emissions trading scheme. Cambridge: Cambridge University Press.
Ellerman, A. D., & Montero, J. P. (1998). The declining trend in sulfur dioxide emissions:
implications for allowance prices. Journal of Environmental Economics and
Management, 36, 26-45.
Ellerman, A. D., & Montero, J. P. (2005). The efficiency and robustness of allowance banking in
the U.S. Acid Rain Program(working paper 05-005). Retrieved from The Center for
Energy and Environmental Policy Research, MIT:
http://dspace.mit.edu/handle/1721.1/45035
Ellerman, A. D., Schmalensee, R., Bailey, E.M., Joskow, P.L., & Montero, J.P. (2000). Markets
for clean air: The U.S. Acid Rain Program. Cambridge: Cambridge University Press.
ERCOT (2010). 2010 State of the Market Report for the ERCOT Wholesale Electricity Markets.
Retrieved from Potomac Economics website:
http://www.potomaceconomics.com/uploads/ercot_reports/2010_ERCOT_SOM_REPOR
T.pdf
Fischbacher, U. (2007). z-TREE: Zurich Tookbox for Ready-made Economic Experiments.
Experimental Economics, 10(2), 171-178.
Garratt, R., & Troger, T. (2006). Speculation in standard auctions with resale. Econometrica,
74(3), 753-769.
Godby, R. (2000). Market power and emissions trading: theory and laboratory results. Pacific
Economic Review, 5(3), 349-363.
Goulder, L. H., Parry, I. W. H., Williams, R. C., & Burtraw, D. (1999). The cost-effectiveness of
alternative instruments for environmental protection in a second-best setting. Journal of
Public Economics, 72(3), 329-360.
124
Hadsell, L. (2007). The impact of virtual bidding on price volatility in New York’s wholesale
electricity market. Economics Letters, 95: 66-72.
Hagem, C., & Westkog, H. (1998). The design of a dynamic tradeable quota system under market
imperfections. Journal of Environmental Economics and Management, 36(1), 89-107.
Hahn, R. W. (1984). Market power and transferable property rights. The Quarterly Journal of
Economics, 99, 763-765.
Hildebrandt, E., Collins, K., McDonald, J., Cooper, B., Deshmukh, A., Kara, E.,…Yang, D.
(2011). 2010 Market Issues and Performance Annual Report. Retrieved from: California
ISO website:
http://www.caiso.com/Documents/2010AnnualReportonMarketIssuesandPerformance.pd
f
Holt, C. A. (1989). The exercise of market power in laboratory experiments. Journal of Law and
Economics, 32, 107-130.
Holt, C., Shobe, W., Burtraw, D., Palmer, K. & Goeree, J. (2007). Auction design for selling
CO2 emission allowances under the Regional Greenhouse Gas Initiative. Retrieved from
the Regional Greenhouse Gas Initiative: http://www.rggi.org/docs/rggi_auction_final.pdf
Hortacsu, A., & Puller, S. (2008). Understanding strategic bidding in multi-unit auctions: A case
study of the Texas electricity spot market. RAND Journal of Economics, 39(1): 86-114.
Hunt, S. (2002). Making competition work in electricity. New York: John Wiley and Sons Press.
Isemonger, A. G. (2006). The benefits and risks of virtual bidding in multi-settlement markets.
The Electricity Journal, 19(9): 26-36.
ISO New England. (2011). 2010 Annual Markets Report. Retrieved from: http://www.iso-
ne.com/markets/mkt_anlys_rpts/annl_mkt_rpts/2010/amr10_final_060311.pdf
Jing Pro [Computer Software] Okemo, MI: Techsmith.
Joskow, P. L. (1998). Restructuring, competition and regulatory reform in the U.S. electricity
eector. In, H. Chao, & H. G. Huntington (Eds.), Designing competitive electricity markets
(11-31). Boston: Kluwer Press.
Joskow, P. L., & Kahn, E. (2002). A quantitative analysis of pricing behavior in California’s
wholesale electricity market during summer 2000: The final word. The Energy Journal,
23(4), 1-35.
Joskow, P. L., & Schmalensee, R. (1983). Markets for power: An analysis of electric utility
deregulation. Cambridge: MIT Press.
125
Joskow, P. L., Schmalensee, R., & Bailey, E. M. (1998). The market for sulfur dioxide emissions.
The American Economic Review, 88(4), 669-685.
Kahn, A. E., Cramton, P., Porter, R. H., & Tabors, R. D. (2001). Pricing in the California Power
Exchange electricity market: Should California switch from uniform pricing to pay-as-
bid pricing? Blue Ribbon Panel Report to the California Power Exchange. Retrieved
from the University of Maryland: http://www.cramton.umd.edu/papers2000-2004/kahn-
cramton-porter-tabors-blue-ribbon-panel-report-to-calpx.pdf
Klemperer, P. (2004). Auctions: Theory and practice. Princeton: Princeton University Press.
Koenker, R. (2005). Quantile regression. New York: Cambridge University Press.
Kolstad, J., & Wolak, F. (2003). Using environmental emissions permit prices to raise electricity
prices: evidence from the California electricity market(Working Paper WP 113).
Retrieved from the University of California Center for the Study of Energy Markets:
http://www.ucei.berkeley.edu/PDF/csemwp113.pdf
Krishna, V. (2009). Auction theory(2
nd
ed.). Oxford: Elsevier Press.
Lange, A. (2012). On the endogeneity of market power in emissions markets. Environmental
and Resource Economics, forthcoming.
Liski, M., & Montero, J. P. (2006). On pollution permit banking and market power. Journal of
Regulatory Economics, 29(3), 283-302.
List, J. A., & Lucking-Reiley, D. (2000). Demand reduction in multiunit auctions: evidence from
a sportscard field experiment. American Economic Review, 90(4), 961-972.
Malik, A. S. (2002). Further results on permit markets with market power and cheating. Journal
of Environmental Economics and Management, 44(3), 371-390.
Maskin, E., & Riley, J. (2000). Asymmetric auctions. The Review of Economic Studies, 67(3),
413-438.
McAfee, P. R., & McMillan, J. (1987). Auctions and bidding. Journal of Economic Literature,
25, 699-738.
McAfee, R. P., & Vincent, D. (1993). The declining price anomaly. Journal of Economic Theory,
60, 191-212.
Milgrom, P. (2004). Putting auction theory to work. Cambridge: Cambridge University Press.
Misiolek, W. S., & Elder, H. W. (1989). Exclusionary manipulation of markets for pollution
rights. Journal of Environmental Economics and Management, 16, 156-166.
126
Muller, A. R., Mastelman, S., Spraggon, J., & Godby, R. (2002). Can double auctions control
monopoly and monopsony power in emissions trading markets? Journal of
Environmental Economics and Management, 44(1), 70-92.
Murray, B. (2009). Power markets and economics: Energy costs, trading, emissions. London:
John Wiley and Sons Press.
Newberry, D. M. (1999). Privatization, restructuring and regulation of network utilities.
Cambridge: MIT Press.
Ott, A. L. (2003). Experience with PJM market operation, system design, and implementation.
IEEE Transactions on Power Systems, 18(2): 528-534.
Parry, I. W. H., Williams, R. C., & Goulder, L. (1999). When can carbon abatement policies
increase welfare?: The fundamental role of distorted factor markets. Journal of
Environmental Economics and Management, 37(1), 52-84.
PJM (2012). State of the Market Report for PJM, Volume II: Detailed Analysis. Retrieved from
Monitoring Analytics website:
http://www.monitoringanalytics.com/reports/PJM_State_of_the_Market/2011.shtml
Perekhodtsev, D., & Baselice, R. (2010). Empirical assessment of strategic behavior in the Italian
power exchange. European Transactions on Electrical Power, 21(6): 1842-1855.
Porter, D., & Vragov, R. (2006). An experimental examination of demand reduction in multi-unit
versions of the uniform-price, Vickrey, and English auctions. Managerial and Decision
Economics, 27(6): 445-458.
Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel and longitudinal modeling using Stata(2
nd
ed.). College Station, TX: STATA Press.
Regional Greenhouse Gas Initiative Inc. (2008). Model Rule. Retrieved from:
http://www.rggi.org/docs/Model%20Rule%20Revised%2012.31.08.pdf
Regional Greenhouse Gas Initiative. (2010). Annual Report on the Market for RGGI CO2
Allowances: 2010. Retrieved from: http://rggi.org/docs/MM_2010_Annual_Report.pdf
Rogerson, W. P. (1984). A note on the incentive for a monopolist to increase fixed costs as a
barrier to entry. Quarterly Journal of Economics, 99, 399-402.
Rose, A., & Dormady, N. (2011). A meta analysis of the economic impacts of climate change
policy in the United States. The Energy Journal, 32(2), 143-165.
Rose, A., Wei, D., & Dormady, N. (2011). Regional macroeconomic assessment of the
Pennsylvania climate action plan. Regional Science Policy and Practice, 3(4), 357-379.
127
Ruth, M., Gabriel, S. A., Palmer, K. L., Burtraw, D., Paul, A., Chen, Y.,...Miller, J. (2008).
Economic and energy impacts from participation in the regional greenhouse gas
initiative: A case study of the State of Maryland.” Energy Policy, 36(6), 2279-2289.
Salop, S. C., & Scheffman, D. T. (1983). Raising rivals' costs. American Economic Review, 73,
267-271.
Salop, S. C., & Scheffman, D. T.(1987). Cost raising strategies. Journal of Industrial Economics,
26, 19-34.
Salop, S. C., Scheffman, D. T. & Schwartz, W. (1984). A bidding analysis of special interest
regulation: Raising rivals' costs in rent seeking society. In Federal Trade Commission
(Eds.) The political economy of regulation: private interests in the regulatory process.
Washington, D.C: Federal Trade Commission.
Saravia, C. (2003). Speculative trading and market performance: The effect of arbitrageurs on
efficiency and market power in the New York electricity market(working paper 121).
Retrieved from UC Berkeley, Center for the Study of Energy Markets at:
http://escholarship.org/uc/item/0mx44472
Sartzetakis, E. S. (1997). Tradeable emission permits regulations in the presence of imperfectly
competitive product markets: Welfare implications. Environmental and Resource
Economics, 9, 65-81.
Smith, A. E., Ross, M.T. & Montgomery, W. D. (2002). Implications of trading implementation
design for equity-efficiency trade-offs in carbon permit allocations. Retrieved from
Charles Rivers Associates:
http://www.crai.com/uploadedFiles/RELATING_MATERIALS/Publications/Consultant_
publications/Smith,_A/files/carbon-permit-allocations.pdf
Stevens, B., & Rose, A. (2002). A dynamic analysis of the marketable permits approach to global
warming policy: A comparison of spatial and temporal flexibility. Journal of
Environmental Economics and Management, 44(1): 45-69.
Stranlund, J. K., Murphy, J. J., & Spraggon, J. M. (2011). An experimental analysis of
compliance in dynamic emissions markets. Journal of Environmental Economics and
Management, 62(3), 414-429.
Tietenberg, T. H.(2006). Emissions trading: Principles and practice(2nd ed.). Washington,
D.C.: Resources for the Future Press.
Tirole, J. (1988). The theory of industrial organization. Cambridge: MIT Press.
United States Department of Justice (USDOJ) (2010). Horizontal merger guidelines. Retrieved
from: http://www.justice.gov/atr/public/guidelines/hmg-2010.html#5a
Van Dyke, B. (1991). Emissions trading to reduce acid deposition. Yale Law Journal, 100: 2707-
2726.
128
Van Egteren, H., & Weber, M.(1996). Marketable permits, market power, and cheating. Journal
of Environmental Economics and Management, 30, 161-173.
Vickrey, W. (1961). Counterspeculation, auctions, and competitive sealed tenders. Journal of
Finance, 16: 8-37.
Waterson, M. (1984). Economic theory of the industry. Cambridge: Cambridge University Press.
Webber, R. J. (1997). Making more from less: Strategic demand reduction in the FCC spectrum
auctions. Journal of Economics & Management Strategy 6(3): 529-548.
West, B. T., Welch, K. B., & Galecki, A. T. (2010). Linear mixed models: A practical guide
using statistical software. New York: Chapman and Hall/CRC Press.
Williamson, O. E. (1968). Wage rates as a barrier to entry: The Pennington case in perspective.
Quarterly Journal of Economics, 82: 86-116.
Wolfram, C. (1998). Strategic bidding in a multiunit auction: An empirical analysis of bids to
supply electricity in England and Wales. RAND Journal of Economics, 29(4): 703-725.
Wrake, M., Myers, E., Mandell, S., Holt, C., & Burtraw, D. (2008). Pricing strategies under
emissions trading: An experimental analysis. Resources for the Future Discussion Paper.
DP 8-49.
Zheng, C. Z. (2002). Optimal auction with resale. Econometrica, 70(6): 2197-2224.
129
Appendix A. Market Conduct and the Herfindahl-Hirschman
Index
The Herfindahl-Hirschman Index (HHI) is directly related to another formal measure of
market power, the Lerner Index. This relationship, known as the Structure-Conduct-Performance
relationship is elaborated below. The interested reader is encouraged to see Waterson (1984) for
a more detailed treatment. Structural features of the market refer to the composition of the
market itself, such as the equality of market influence among firms within the market(e.g., HHI) or features of demand for products produced by those firms (e.g., demand elasticity). Conduct
refers to the profit maximization behavior of firms within the market (e.g., quantity competition).
And, performance refers to any deviations from a competitive market price, such as monopoly
profits.
Assume a basic firm-level profit function of the form:
()() i i i i
p Q q C q Õ = × - 1,2,..., i n =(1) Assuming each firm maximizes profits by setting output, the first order condition is:
0
i
i i
i i
d dP dQ
p q C
dq dQ dq
Õ
¢ = + × × - =(2) The third term of equation (2) can be extended further, as it is the effect of the ith firm's
output on the other firm's outputs. The ith firm, in making its maximization decision, has to come
to some assessment of this effect. It can be expanded to include the output of the other firms ( i
Q) as:
1
i i
i
i i i
dq dQ dQ
dq dq dq
 = + º +
130
Lambda is known as a conjectural variation term, as it is i's assessment, or conjecture,
regarding the relationship between its output changes and the output changes of corresponding
firms. Substituting this decomposition into equation 2 yields:
(1) 0
i i i
dp
p q C
dQ
 ¢ + × × + - =(3) Rearranging equation 3 slightly yields:
(1) i i i
dp
p C q
dQ
 ¢ - =- × × +(4) Multiplying both sides by
1
p
results in:
[ ]
1 1
(1) i i i
dp Q
p C q
p dQ Q p
 é ù
¢ × - = - × × + × ×
ê ú
ë û
(5)(1) i i
i
p C s
p
  ¢ -
= × +(6) By definition we know that
1
(1) 1
n
i i
s  × + =
å
, and thereby equation 6 can be reframed
as:
(1) i i
i
p C s
p
  ¢ -
= × +(7) where the left hand side of the equation is the Lerner Index:
1
i
i
p C
L
p  ¢ - -
= =
131
The performance of the market (the profit-revenue ratio) is determined by the structural
features of the market (HHI and demand elasticitye) by way of market conduct (maximization
behavior and conjectural variation).
132
Appendix B. Design and Operation of the Monte Carlo
Simulation Environment Oligopsony 1.0
B.1 Background
Oligopsony 1.0 is an object-oriented program (OOP) that is designed specifically to
simulate uniform-price auctions in large samples. It is designed to allow a broad variety of user-
based inputs, representing conditions specific to many auction-based markets. And, it is designed
to simulate these auctions in a Monte Carlo environment capable of producing very large sample
sizes, restricted only by the computational capacities of the user's computer.
Figure B.1 Data Input Screen
133
Moreover, Oligopsony 1.0 also allows for the simulation of two types/classes of bidders.
It is designed to carry out simulations consistent with the theoretical properties of the oligopolist-
fringe style of model, first introduced by Hahn (1984). Although the software is designed with
labels specifically for this framework, slight alternations of the user-supplied inputs would enable
the same Monte Carlo simulations with two altogether different types of bidders, such as bidders
who may have different ex ante valuations.
B.2 Sampling Methodology
The sampling methodology of Oligopsony 1.0, like many other Monte Carlo
environments, is driven by distribution-constrained random number generation guided by an
internal business logic consistent with the simulation parameters. Table B1 provides a brief
snapshot of the method through which user-supplied input parameters are utilized in order to
generate the bid matrices and profit calculations provided in Section 4. Of note, Oligopsony 1.0
generates a unique bid price and quantity, and then secondarily matches that bid to a bidder,
unless that bid is inconsistent with the ex-ante business logic structure.
That business logic sets only a few constraints on the matching of bid to bidder. First,
multiple bids per bidder are permitted. For purposes of simplicity, no more than 4 bids per bidder
are permitted by Oligopsony 1.0. For example, the number of rows in a bid matrix in a market
with 50 bidders could be as few as 50, and as many as 200 for each auction. Second, the user-
supplied maximum and minimum bid quantities are resampled against prior bids by the same
bidder. If the combination of old bid(s) and the newly generated bid do not exceed the user-
supplied maximum and minimum quantities, that newly generated bid is permitted to be matched
to that bidder. If the bid exceeds the maximum, the bid is discarded and no additional sampling
for that bidder is permitted. If, for example, the user wishes to force a specific class of bidder to
134
bid a specific quantity, as is done in Section 6, the user can supply the same values for both the
maximum and minimum, thus permitting no quantity bids for that bidder to deviate above or
below that chosen quantity.
Third, similar sets of constraints are imposed on the pricing side based upon the user-
supplied inputs. However, unlike the maximum and minimum values imposed on the quantity
side, the pricing side values are not hard and fixed constraints. Their full range is constrained by
the probabilistic capacities of a Gaussian/Standard Normal distribution. As such, larger standard
deviations, if supplied by the user, could produce significantly larger or smaller bid prices, near
the extremities of the distribution's tails.
Table B.1 Sampling Methodology for User-Supplied Market Parameters
Input Parameter Sampling Methodology
Total Quantity
to be Auctioned
The quantity of goods (e.g., emissions permits) to be auctioned can be
constrained or expanded as per the user's input.
Number of Bidders The overall size of the market (e.g., number of firms bidding) can be
constrained or expanded as per the user's input.
Number of
Oligopsonists
The share of the total market that maintains market power. Market power is a
function of the below pricing and quantity constraints imposed upon this class
of bidder, relative to those imposed upon the alternative class of bidder (e.g.,
fringe).
Max Bid Quantity
(Oligopsonist &
Fringe) Each bid is drawn separately from a uniform distribution with minimum and
maximum input parameters. Oligopsony 1.0 allows the user to distinguish the
sampling range (minimum and maximum) differently for each class of bidder.
Bid quantities are generated prior to class (e.g., oligopsonist/fringe) matching,
therefore bid quantities for multiple bids per bidder are possible.
Min Bid Quantity
(Oligopsonist &
Fringe) Secondary
Market Price
(Exogenous) Profit is calculated from the simple profit function given in Section 4.1. The
secondary market value, or utility derived from that good's (e.g., emissions
permit) use is set exogenously as per the user's input.
Pricing
(Min, Max &
Standard Deviation) Bid prices are drawn from a Gaussian/Standard Normal distribution. Two user-
supplied parameters are required to generate a number from a Gaussian
distribution; mean and standard deviation. The mean is generated from a
uniform distribution with two additional user-supplied parameters; minimum
and maximum. Oligopsony first receives these minimum and maximum
parameters and generates a mean from a uniform distribution. It then uses that
mean along with the user-supplied standard deviation to generate a bid price
from a Gaussian/Standard Normal distribution.
135
B.3 Multi-Threading
Oligopsony 1.0 is designed for smooth operation on a modern Windows- based PC. The
software enables multi-threading for faster simulation. Multi-threading enables different
subroutines, such as those generating pricing bid components to be handled by separate
processors on a modern multi-core computer. This is important because Oligopsony 1.0 is a
computation-heavy software, and can require significant computational resources if many
iterations are requested by the user.
It is important to note, that Monte Carlo environments are not typically designed for
multi-threaded applications. The combination of multi-threading and number generation requires
a carefully balanced timing procedure through which the application can determine the
appropriate time to allow an event to fire. This is because all number generators to date, utilize
the computer's internal clock to generate numbers. As the computer's clock records the passing of
milliseconds, the decimal values of that record becomes the randomly generated number. When
fast multi core computers run computation intensive operations like Oligopsony 1.0, timing is key.
For example, a multi-threaded application calling on the system clock for a random
number, operated by four separate processors on a quad core processor, would access the system
clock at such a high rate of speed, that the same decimal value would be culled from the clock
before a millisecond change, and the returned values would essentially be a repetitious list of the
exact same four numbers. To overcome this, significant trial and error went into a balanced
timing procedure. As events fire (e.g., a number being pulled from the system clock), thread
delays are introduced into the code to postpone that event for a short period of milliseconds , to
allow the system clock to advance forward. To the author's best knowledge, this innovation has
not yet been introduced in Monte Carlo methodology.
136
B.4 Use and Operation
Despite the large number of background operations, the user interface of Oligopsony 1.0
is quite simple. First, the user begins by selecting the "Set Parameters" button located on the top
left of the UI. Figure B1 provides a graphic of the input parameters available in this UI. Once
the user has entered in parameters consistent with the intended simulation, and clicks on the
"Apply" button, those parameters will be stored in the system's background configuration, and
available for later simulation. Additionally, the user may select the "Clear" button, to begin with
a fresh set of parameters.
Second, the user selects the number of iterations (auctions) that (s)he would like to
simulate. If the user selects only "1" simulation, a bid matrix will appear as a pop-up box so that
the user can interrogate the specifics of that simulation. The columns in the bid matrix are
identical to those in bid matrix B, in Section 4.1. For simplicity, entries in that box are color
coded red if they belong to an oligopsonist.
137
Figure B.2 Individual Bid Matrix
Once a simulation has been completed, the completion bar at the bottom of the UI will be
completed. The first four columns represent the market clearing price of the individual auction,
the fringe loss (quantity members of the type fringe bid for but did not receive), total oligopsonist
profit, and total fringe profit, respectively. Above those columns, the software automatically
provides three basic sets of descriptive statistics for each of the columns of auction summaries,
including the minimum value, mean value, and maximum value. For example, if the user called
for 1,000 auctions to be simulated, the descriptive statistics box located above the auction
clearing price array would show the lowest price at which the market cleared out of 1,000
auctions, the mean price, and the maximum price.
138
Figure B.3 Data Output Screen
The fifth column represents the mean price value at which members of the type
oligopsonist bid in that specific auction. This provides for a quick and easy method of verifying
class-specific sampling. Columns 6 through 8 provide each specific auction's individual bids, in
bid matrix form and rank ordered, in a rich text box so that the user can easily export those values
to a statistical analysis package (e.g., Stata) or a spreadsheet. The user can see any auction's bid
matrix by simply clicking on either of the two left columns.
139
Finally, the first two columns that represent the auction clearing price and the fringe loss
are not rich text boxes, because they are used as linking items. If the user wishes to place those
items into a rich text box for further data analysis, two additional buttons on the top of the UI
provide for this. Selecting either of those buttons will export those values into the far right
columns, and place them into a rich text box.
140
Appendix C. Experiment Environment
C.1 Treatment A-- Control Group (No Market Power) Video Instructions Transcript
42
Introduction
Today you will be participating in an economics experiment to simulate strategic
behavior in an energy market. Thank you for your time and participation!
The experiment should last approximately 3 hours. We will begin with some instructions
to help you through the experiment.
In addition to the money you receive for showing up today, you will get to keep the
money you make during this experiment. For every dollar you make during the experiment, you
will keep 10 cents. If you have any questions or if any instructions are unclear, please ask them
and do not be shy. The better you understand the experiment, the higher your profit will be and
the more you will get paid. We will take a ten minute break in the middle of the experiment so
you can use the restroom, get a drink or eat a snack outside of this room. Please note, there is no
eating within these facilities.
Please do not leave the room during the experiment. If you leave the room, you will be
excused from the experiment. If you have a cell phone or beeper, please turn it off now. You are
not permitted to communicate with other players during this experiment, unless instructed
otherwise. Please do not attempt to look onto another player's screen. If you require any specific
accommodation, please notify us by raising your hand so that we can address your need. During
the experiment, if you have any questions, please quietly raise your hand, and ask an
experimenter to answer your question. Please do not ask another player.
42
Subjects received instructions via MP4 video. These are the transcripts of the subject instructions. The video also
included a walkthrough of the software environment, not provided here. Subjects also received three handouts,
including a single page summary of the experiment parameters, a payment sheet to record earnings at the end of the
experiment, and a cartoon diagram depicting the market matching of subjects to production units (i.e., energy
generators/products).
141
Production Auctions
You will be participating in two auctions. In one auction, you will be selling products
that you have produced, to the government who needs these products. And in return, the
government will pay you for selling those products. You will be paid based on the rules of the
auction, which we will discuss in a moment.
License Auction
In the other auction, you will be buying licenses from the government, for permission to
produce those products. If you sell products to the government without licenses, you can be
charged a penalty. Your objective is to profit and receive the largest payment as possible.
Two Products
Your products will be two products, Product X and Product Y. The government does not
have a preference between Product X and Product Y, except that it will give preference to the
lowest bids. In other words, just like you when you go to the store, the government wants to buy
its products at the lowest cost possible.
Ten Players
There are 10 players total. All players have one of each product to produce, which means
that there are a total of 20 products in the market overall; 10 Product Xs and 10 Product Ys.
Price Cap
Your bids will indicate the price that you wish to be paid for producing that product.
Bids above $10.00 will not be accepted. This means that the highest price the government is
142
willing to pay is $10.00 for either Product X or Product Y. For example, if you bid $5.99 for
Product X and $8.99 for Product Y, you are indicating to the government that you are willing to
sell your products at or above those prices.
The auction is a uniform-price auction, which means that the price you are paid for your
product will be determined by the auction and will be the same for all other players. You will
never be paid less than your bid, and oftentimes you will receive a higher price. If your bid is too
high, above the clearing price, your bid will not be accepted.
Periods
You will be participating in many periods of bidding. In each period there are two
rounds, and in each round there will be both auctions. Each period the government will indicate a
different demand for your products. That demand will either be 10, 15 or 20. It will be the same
for both rounds, and only change across periods.
When the demand is 10, the government will buy 10 products from those of you who
have the lowest bid, and the 10th lowest bid price will set the price for the bids that are included.
When the demand is 15, the 15th lowest bid will set the price, and when the demand is 20, the
highest bid (last bid) will set the price. For example, if the government specifies that it will buy
15 products this period, and you bid $5.99 for your Product X and $8.99 for your Product Y, and
the 15th lowest bid is $9.25, you will be paid $9.25 for both of your products, and receive a
revenue of $18.50 (e.g., $9.25 * 2).
If there are ever any ties between bids, the winner will be decided by the computer
randomly.
143
Production Costs
When the government buys your products, you have produced that product. However,
you will be required to pay a cost for producing your products. For each Product X you produce,
you will have to pay a $1.00 production cost. For each Product Y you produce, you will have to
pay $2.00. Be careful; if you bid lower than your production cost you can lose money. If your
total profit for any period is negative, you will not lose money. Instead, you will receive a profit
of zero dollars ($0.00).
Example
Consider the example below (next page) to help you understand how the uniform-price
auction works. The table on the left shows bids placed by players #1-10 for each product X and
Y. The government wishes to buy its products from players at the lowest cost possible, so it will
rank your bids from lowest to highest. (Notice the column indicated “Rank”, which places the
lowest bid of $1.10 for the top rank of #1, and the highest bid of $9.99 as the bottom rank of #20.) In the table on the right, you will see the same bids placed in order of their rank, from lowest bid
to highest bid. When the demand for products is set at 15, the government will buy the 15 lowest
cost products, and the price that clears the market will be $5.50, in this example. Those bids that
are less than the clearing price are winning bids, and those players win the right to produce those
products and sell them to the government. Notice that players #6 and #10 do not get to sell any
products, and will receive revenue of $0.00, and Player #4 wins only the right to sell his Product
X. Because $5.50 is the price that clears the market, all players will receive $5.50 per
product that they sell. Player #8, for example, will sell 2 products (both Product X and Product
Y) at $5.50, and receive revenue of $11.00. Player #4, on the other hand, sells only one product,
and will receive revenue of $5.50.
144
Remember that the demand for products will not always be 15; sometimes it will also be
10 or 20. If the demand for products in this example were 10, only the top ten bids would win the
right to produce products, and they would be paid a uniform price of $4.44 per product produced.
If the demand for products in this example were 20, all bids would win the right to produce
products, and all bids would be paid a uniform price of $9.99 (e.g., all players would receive
revenue $19.98).
License Auction
For the License auction, you will be bidding to buy licenses from the government. For
this auction, the government will give preference to the highest bids. At the end of each period
(which is the end of Round #2), you must hold 1 license for every product you have produced.
The license auction works the same as the production auction (from the example above), except
the ranked preference will be given to the highest bids, because the government wishes to receive
the highest payment possible. If you were selling something of yours (e.g., computer, cell phone) in an auction, wouldn’t you want to be paid the highest price possible?
Compliance
For example, if you produced Product X and Product Y in both rounds, you will need to
hold at least 4 licenses at the end of Round #2. And, if your bid for Product Y was too high and
you did not win production of that product, and only produced Product X in Round #1, but won
production of both products in Round #2, you will have produced three (3) products, and will
require at least 3 licenses at the end of Round #2.
145
How to Bid
For the License auction, you will specify both a price and a quantity. For example, if you
wish to buy all 4 licenses in the second round, you may place a bid such as 4 licenses for $3.99
per license. This indicates to the government that you are not willing to pay more than $3.99 per
license for 4 licenses. You can place bids at any price that you think is appropriate to the
conditions of the market. You can buy your licenses in any round you wish, but you must have 1
license for every product you have produced by the end of round #2.
Example
Let’s go through another example of how the auction works, but this time let’s do it for
the license auction.
The license auction is similar to the production auction, except that bids are placed in
reverse order- that is, the highest bids are most highly ranked. Of course, this makes sense,
because the government wants to sell its licenses at the highest cost possible, just like you would
if you were selling something of yours.
Consider in this example, ten players who each place a bid for a price and a quantity of
licenses. Player 1 places a bid for 2 licenses at a price of 2 dollars and 99 cents per license. He is
telling the government that he is not willing to pay more than 2 dollars and 99 cents for each
licenses, and wishes to buy two of them this round.
Player number 2 places a bid for 1 license, at a price of 3 dollars and 74 cents. Player 4
places a bid for 4 licenses at 1 dollar and 95 cents. The highest bid is from player 7, who placed a
bid for 8 dollars and 2 cents. Player 7 receives the highest rank of 1. The lowest rank goes to the
lowest bid, who in this example is Player 4. 1.95 is the lowest bid.
146
Looking over to the table on the right, here is what the auction mechanism does with
those bids. Those bids are now ordered by rank, and Player 7, who received the highest rank of
Rank 1, is at the top and receives the highest priority to receive licenses from the government.
Then, player 5, player 3, and so on.
In this example, 15 licenses are to be sold. You will see that the bids are counted by their
bid quantity, and when the cumulative bid is equal to the amount for sale, or in other words, when
supply meets demand, that bid price becomes the market clearing price. In our example, it is
Player #1’s bid, that sets the price. Everyone who receives licenses, receives them at his bid price
of 2 dollars and 99 cents, in this example.
Everyone who had a rank above that player who sets the price will receive their full
quantity of licenses at $2.99. Everyone below him will receive zero licenses. Player 1 however,
will only receive 1 license. Even though he bid for 2 licenses, he will only receive 1, because
there are only 15 licenses for sale. Everyone who has priority above him received their bids, and
that leaves only 1 more license left for him. So, player 7 receives 2 licenses, and pays 2 dollars
and 99 cents per license. Player 5 receives 1 license and pays 2 dollars and 99 cents. And so on
and so on.
Again, if there is a tie between any two players, the bid will be awarded randomly. And,
if the bid quantities do not meet the supply of licenses for sale, the price will become zero. In
other words, if there are 15 licenses for sale, but after adding up everyone’s bid quantities, the
most everyone bid for was 14 licenses, the supply will not meet demand, and the price will fall to
zero. If this occurs, you will receive licenses for free.
147
Non-Compliance
At the end of round #2, if you do not have enough licenses to cover your production,
you will be fined $5.00 per missing license.
For example, if you produced a Product X and a Product Y in both rounds, you produced
two (2) products both rounds, for a total of 4 products for that period. In this case, you will be
required to have at least 4 licenses by the end of Round #2. If you only receive 3, as in the
example shown here, you will receive a penalty for being short 1 license, of $5.00.
In the second example here, if you produced both Product X and Product Y in Round #1,
but only produced Product X in round #2, you have produced a total of 3 products this period.
You will need at least 3 licenses. If, as in the example, you only received 1 license by the end of
Round 2, you will be short 2 licenses, and have to pay 2 $5 penalties, or $10.00.
In order to avoid paying non-compliance penalties, you may wish to purchase enough
licenses to cover your production.
Supply of Licenses
The quantity of licenses that the government will sell is always the same. In round #1,
the government will always sell 20 licenses. In round #2, the government will always sell 15
licenses. Overall, there will be 35 licenses every period. Again, ties for licenses will be awarded
randomly.
Bank Account
Your license bid cannot exceed your current bank account. You will begin with $20 in
your account in round #1. If you profit in round #1, your bank account for round #2 will increase.
If you lose money in round #1, your bank account in round #2 will decrease. Your bank account
148
will be your limit to purchase licenses. For example, if you have $20 in your bank account, you
will only be able to buy $20 worth of licenses, such as 4 licenses for $5.00 each.
You will profit from selling your products to the government at a price higher than it cost
you to produce them. You will also profit from reselling licenses to other players who need them,
at a higher price than you paid for them.
Resale
After you place your bids for round #1, you will see a screen that indicates whether you
won the right to produce products and how many licenses you received. This will help you make
decisions for how to bid in the second round. On that screen, you will be given the opportunity to
resell licenses that you won in round #1, into round #2. You may find it profitable to resell
licenses to other players. If you or other players post licenses for resale, the quantity of licenses
to be sold in round #2 will increase. For example, if all 20 licenses are sold in Round #1, and you
post 3 of your licenses for resale, the total quantity of licenses for sale in Round #2 will be your 3
licenses, plus the government’s 15, for a total of 18 licenses.
The quantity of licenses for sale in round #2 will also include any licenses that the
government did not sell in round #1. At the end of Round #1 you will be provided with the
number of unsold liceneses, to help you make your resale decision.
Note, the government will give preference to itself. This means that even if you attempt
to resell licenses, you will not resell any licenses unless all of the government’s licenses are sold
first. For example, if 18 of the original 20 licenses are sold in Round #1, and other players post 5
licenses for resale, the total quantity of licenses for sale in Round #2 will be the government’s 15,
plus the 2 the government did not sell in Round #1, plus the 5 licenses posted for resale. In this
case, 22 licenses will be for sale in Round #2.
149
Please note, there is a possibility that you will resell only a fraction of the licenses you
posted for resale, which can occur if there is a tie for a quantity of licenses less than the total for
sale. For example, it is possible that you post 3 licenses for resale, but only resell 1.5 of them.
At the end of each period, you will see a screen that indicates all of your revenues and
costs. This screen indicates both your profit from Round #2, and your total profit from both
rounds.
C.2 Treatment B Treatment Group (Market Power) Video Instructions Transcript
Introduction
Today you will be participating in an economics experiment to simulate strategic
behavior in an energy market. Thank you for your time and participation!
The experiment should last approximately 3 hours. We will begin with some instructions
to help you through the experiment.
In addition to the money you receive for showing up today, you will get to keep the
money you make during this experiment. For every dollar you make during the experiment, you
will keep 10 cents. If you have any questions or if any instructions are unclear, please ask them
and do not be shy. The better you understand the experiment, the higher your profit will be and
the more you will get paid. We will take a ten minute break in the middle of the experiment so
you can use the restroom, get a drink or eat a snack outside of this room. Please note, there is no
eating within these facilities.
Please do not leave the room during the experiment. If you leave the room, you will be
excused from the experiment. If you have a cell phone or beeper, please turn it off now. You are
not permitted to communicate with other players during this experiment, unless instructed
150
otherwise. Please do not attempt to look onto another player's screen. If you require any specific
accommodation, please notify us by raising your hand so that we can address your need. During
the experiment, if you have any questions, please quietly raise your hand, and ask an
experimenter to answer your question. Please do not ask another player.
Production Auction
You will be participating in two auctions. In one auction, you will be selling products
that you have produced, to the government who needs these products. And in return, the
government will pay you for selling those products. You will be paid based on the rules of the
auction, which we will discuss in a moment.
License Auction
In the other auction, you will be buying licenses from the government, for permission to
produce those products. If you sell products to the government without licenses, you can be
charged a penalty. Your objective is to profit and receive the largest payment as possible.
Two Products
Your products will be two products, Product X and Product Y. The government does not
have a preference between Product X and Product Y, except that it will give preference to the
lowest bids. In other words, just like you when you go to the store, the government wants to buy
its products at the lowest cost possible.
151
Ten Players
There are 10 players total in the market. Among those players, there are a total of 10
product Xs, and 10 product Ys. Some players will have more products than others. As in the
picture shown here, two players will have 6 products total, 3 product Xs and 3 product Ys. The
rest of the players will each have 1 product, either product X or product Y. The number of
products you have depends upon your status, which will change periodically throughout the
experiment.
Price Cap
Your bids will indicate the price that you wish to be paid for producing that product.
Bids above $10.00 will not be accepted. This means that the highest price the government is
willing to pay is $10.00 for either Product X or Product Y. For example, if you bid $5.99 for
Product X or $8.99 for Product Y, you are indicating to the government that you are willing to
sell your products at or above those prices.
The auction is a uniform-price auction, which means that the price you are paid for your
product will be determined by the auction and will be the same for all other players. You will
never be paid less than your bid, and oftentimes you will receive a higher price. If your bid is too
high, above the clearing price, your bid will not be accepted.
Periods
You will be participating in many periods of bidding. In each period there are two
rounds, and in each round there will be both auctions. Each period the government will indicate a
different demand for your products. That demand will either be 10, 15 or 20. It will be the same
for both rounds, and only change across periods.
152
When the demand is 10, the government will buy 10 products from those of you who
have the lowest bid, and the 10th lowest bid price will set the price for the bids that are included.
When the demand is 15, the 15th lowest bid will set the price, and when the demand is 20, the
highest bid (last bid) will set the price. For example, if the government specifies that it will buy
15 products this period, and you bid $5.99 for your Product X, and the 15th lowest bid is $9.99,
you will be paid $9.99 for your product X. Because the auction is a “uniform price” auction, all
winning bids will receive the highest winning price. So, even though you only asked to be paid
$5.99, you will get $9.99, in this example.
If there are ever any ties between bids, the winner will be decided by the computer
randomly.
Your “status”, or, how many products you get to product, as I mentioned previously, will
change every period, randomly. In one period, for example, you may have only one product, but
in the next period, you may be one of the two players who has 6 products.
Production Costs
When the government buys your products, you have produced that product. However,
you will be required to pay a cost for producing your products. For each Product X you produce,
you will have to pay a $1.00 production cost. For each Product Y you produce, you will have to
pay $2.00. Be careful; if you bid lower than your production cost you can lose money. If your
total profit for any period is negative, you will not lose money. Instead, you will receive a profit
of zero dollars ($0.00).
153
Example
Consider the example below (next page) to help you understand how the uniform-price
auction works. The table on the left shows bids placed by players #1-10 for each product X and
Y. Notice that the last two players each have six separate bids, because they each get six separate
products to sell to the government.
The government wishes to buy its products from players at the lowest cost possible, so it
will rank your bids from lowest to highest. (Notice the column indicated “Rank”, which places
the lowest bid of $1.10 for the top rank of #1, and the highest bid of $9.99 as the bottom rank of
#20.) In the table on the right, you will see the same bids placed in order of their rank, from
lowest bid to highest bid. When the demand for products is 15, the government will buy the 15
lowest cost products, and the price that clears the market will be $5.50, in this example.
Those bids that are less than the clearing price are winning bids, and those players win
the right to produce those products and sell them to the government, because those players were
willing to sell their products to the government more cheaply.
Notice that the bottom five bids are losing bids. Player #8 placed a bid for $6.10, that
was higher than the clearing price, and he does not win the right to sell his product to the
government. He receives a revenue of zero dollars. Players #9 and #10 also each have two losing
bids, and as a result, will not be selling those products to the government.
Because $5.50 is the price that clears the market, all players with winning bids receive
$5.50 for each product they sell. In other words, Player #1 was willing to sell his Product Y to
the government at $1.99, but because the auction uses a uniform price, he will actually be paid
$5.50.
Remember that the demand for products will not always be 15; sometimes it will also be
10 or 20. If the demand for products in this example were 10, only the top ten bids would win the
154
right to produce products, and they would be paid a uniform price of $4.44 per product produced.
If the demand for products in this example were 20, all bids would win the right to produce
products, and all bids would be paid a uniform price of $9.99.
License Auction
For the License auction, you will be bidding to buy licenses from the government. For
this auction, the government will give preference to the highest bids. At the end of each period
(which is the end of Round #2), you must hold 1 license for every product you have produced.
The license auction works the same as the production auction (from the example above), except
the ranked preference will be given to the highest bids, because the government wishes to receive
the highest payment possible. If you were selling something of yours (e.g., computer, cell phone) in an auction, wouldn’t you want to be paid the highest price possible?
Compliance
For example, if you produced Product X in both rounds, you will have produced two
products, and will need 1 license for each product. Or, as in the second example, if you produced
2 Product Xs and 1 Product Y, you have produced a total of 3 products and will be required to
have a total of 3 licenses.
At the end of Round #2, you will need 2 licenses. You can buy those licenses in Round
#1, or in Round #2, or half in one and half the other, whichever you like, but by the end of Round
#2, you must have 1 license for each product you have produced.
155
How to Bid
For the License auction, you will specify both a price and a quantity. For example, if you
wish to buy 4 licenses in the second round, you may place a bid such as 4 licenses for $3.99 per
license. This indicates to the government that you are not willing to pay more than $3.99 per
license for 4 licenses. You can place bids at any price that you think is appropriate to the
conditions of the market. You can buy your licenses in any round you wish, but you must have 1
license for every product you have produced by the end of round #2.
Example
Let’s go through another example of how the auction works, but this time let’s do it for
the license auction.
The license auction is similar to the production auction, except that bids are placed in
reverse order- that is, the highest bids are most highly ranked. Of course, this makes sense,
because the government wants to sell its licenses at the highest cost possible, just like you would
if you were selling something of yours.
Consider in this example, ten players who each place a bid for a price and a quantity of
licenses. Player 1 places a bid for 2 licenses at a price of 2 dollars and 99 cents per license. He is
telling the government that he is not willing to pay more than 2 dollars and 99 cents for each
licenses, and wishes to buy two of them this round.
Player number 2 places a bid for 1 license, at a price of 3 dollars and 74 cents. Player 4
places a bid for 4 licenses at 1 dollar and 95 cents. The highest bid is from player 7, who placed a
bid for 8 dollars and 2 cents. Player 7 receives the highest rank of 1. The lowest rank goes to the
lowest bid, who in this example is Player 4. 1.95 is the lowest bid.
156
Looking over to the table on the right, here is what the auction mechanism does with
those bids. Those bids are now ordered by rank, and Player 7, who received the highest rank of
Rank 1, is at the top and receives the highest priority to receive licenses from the government.
Then, player 5, player 3, and so on.
In this example, 15 licenses are to be sold. You will see that the bids are counted by their
bid quantity, and when the cumulative bid is equal to the amount for sale, or in other words, when
supply meets demand, that bid price becomes the market clearing price. In our example, it is
Player #1’s bid, that sets the price. Everyone who receives licenses, receives them at his bid price
of 2 dollars and 99 cents, in this example.
Everyone who had a rank above that player who sets the price will receive their full
quantity of licenses at $2.99. Everyone below him will receive zero licenses. Player 1 however,
will only receive 1 license. Even though he bid for 2 licenses, he will only receive 1, because
there are only 15 licenses for sale. Everyone who has priority above him received their bids, and
that leaves only 1 more license left for him. So, player 7 receives 2 licenses, and pays 2 dollars
and 99 cents per license. Player 5 receives 1 license and pays 2 dollars and 99 cents. And so on
and so on.
Again, if there is a tie between any two players, the bid will be awarded randomly. And,
if the bid quantities do not meet the supply of licenses for sale, the price will become zero. In
other words, if there are 15 licenses for sale, but after adding up everyone’s bid quantities, the
most everyone bid for was 14 licenses, the supply will not meet demand, and the price will fall to
zero. If this occurs, you will receive licenses for free.
157
Non-Compliance
At the end of round #2, if you do not have enough licenses to cover your production,
you will be fined $5.00 per missing license.
For example, if you produced a Product X and a Product Y in both rounds, you produced
two (2) products both rounds, for a total of 4 products for that period. In this case, you will be
required to have at least 4 licenses by the end of Round #2. If you only receive 3, as in the
example shown here, you will receive a penalty for being short 1 license, of $5.00.
In the second example here, if you produced your product X in both rounds, you have
produced a total of 2 products, and will be required to have 2 licenses. If you have zero (0) licenses by the end of Round #2, you will be short two licenses, and have to pay two non-
compliance penalties of $5 each, for a total penalty of $10.
In order to avoid paying non-compliance penalties, you may wish to purchase enough
licenses to cover your production.
Supply of Licenses
The quantity of licenses that the government will sell is always the same. In round #1,
the government will always sell 20 licenses. In round #2, the government will always sell 15
licenses. Overall, there will be 35 licenses every period. Again, ties for licenses will be awarded
randomly.
Bank Account
Your license bid cannot exceed your current bank account. You will begin with $20 in
your account in round #1. If you profit in round #1, your bank account for round #2 will increase.
158
If you lose money in round #1, your bank account in round #2 will decrease. Your bank account
will be your limit to purchase licenses.
If you are one of the players who has only 1 product to produce, you will begin with a
bank account of $10 in Round #1, and your bank account in Round #2 will increase or decrease
based on your profit from Round #1.
If you are one of the players who has 6 products to produce, you will begin with a bank
account of $60 in Round #1, and your bank account in Round #2 will increase or decrease based
on your profit from Round #1.
For example, if you begin with a bank account of $10, and make a profit of $8 in Round
#1, your bank account in Round #2 will be $18, or $10 plus $8.
The money you have in your bank account is your limit for purchasing licenses. If you
have $10 in your bank account, the most you can spend to buy licenses is $10. If you have $18 in
your bank account, you can only spend $18.
You will profit from selling your products to the government at a price higher than it cost
you to produce them. You will also profit from reselling licenses to other players who need them,
at a higher price than you paid for them.
Resale
After you place your bids for round #1, you will see a screen that indicates whether you
won the right to produce products and how many licenses you received. This will help you make
decisions for how to bid in the second round. On that screen, you will be given the opportunity to
resell licenses that you won in round #1, into round #2. You may find it profitable to resell
licenses to other players. If you or other players post licenses for resale, the quantity of licenses
159
to be sold in round #2 will increase. For example, if all 20 licenses are sold in Round #1, and you
post 3 of your licenses for resale, the total quantity of licenses for sale in Round #2 will be your 3
licenses, plus the government’s 15, for a total of 18 licenses.
The quantity of licenses for sale in round #2 will also include any licenses that the
government did not sell in round #1. At the end of Round #1 you will be provided with the
number of unsold liceneses, to help you make your resale decision.
Note, the government will give preference to itself. This means that even if you attempt
to resell licenses, you will not resell any licenses unless all of the government’s licenses are sold
first. For example, if 18 of the original 20 licenses are sold in Round #1, and other players post 5
licenses for resale, the total quantity of licenses for sale in Round #2 will be the government’s 15,
plus the 2 the government did not sell in Round #1, plus the 5 licenses posted for resale. In this
case, 22 licenses will be for sale in Round #2.
Please note, there is a possibility that you will resell only a fraction of the licenses you
posted for resale, which can occur if there is a tie for a quantity of licenses less than the total for
sale. For example, it is possible that you post 3 licenses for resale, but only resell 1.5 of them.
At the end of each period, you will see a screen that indicates all of your revenues and
costs. This screen indicates both your profit from Round #2, and your total profit from both
rounds.
Chat Room
Those two players who have 6 products, will also have the ability to communicate with
each other using a live chat feature.
Those players can type messages to each other during each round. Your messages will
only be seen by the other player with 6 products, and the other players will not have access to the
160
chat feature until they become one of the players with 6 products, in another period of the
experiment.
C.3 Treatment A Control Group (No Market Power) Subject Interface
Figure C.1 Round #1 Bidding Screen
161
Figure C.2 Round #1 Summary Screen
162
Figure C.3 Round #2 Bidding Screen
163
Figure C.4 Round #2 Summary Screen
164
C.4 Treatment B Treatment Group (Market Power) Subject Interface
Figure C.5 Round #1 Bidding Screen for Type C (Oligopolist)
165
Figure C.6 Round #1 Bidding Screen for Type B (Fringe Player)
166
Figure C.7 Round #1 Summary Screen for Type C (Oligopolist)
167
Figure C.8 Round #1 Summary Screen for Type B (Fringe Player)
168
Figure C.9 Round #2 Bidding Screen for Type A (Fringe Player)
169
Figure C.10 Round #2 Summary Screen (All Types)

Abstract (if available)

Abstract Chapter 1: A Monte Carlo Approach ❧ The use of auctions to distribute tradeable property rights to firms in already heavily concentrated markets may further exacerbate the problems of market power that exist within those markets. This chapter provides a model of a two-stage emissions market modeled after a contemporary regional permit trading market in the United States, the Regional Greenhouse Gas Initiative, Inc. (RGGI). It then introduces Oligopsony 1.0, a C# software package constructed in the .NET environment that simulates uniform-price auctions using stochastic Monte Carlo simulation for modeling market power in tradeable property rights auctions. Monte Carlo methods add a probabilistic element to standard auction theoretic equilibria. The results of these simulations indicate that there can be significant non-linearities between profit and market power as exercised through strategic demand reduction. This analysis finds the optimum point of strategic demand reduction that enables the firm to exploit these non-linearities, and it determines the probability distributions of these optima using kernel density analysis. ❧ Chapter 2: An Experimental Approach ❧ How will emerging auction-based emissions markets function within the context of today’s deregulated auction-based electricity markets? This chapter provides an experimental analysis of a joint energy-emissions market. The impact of market power and collusion among dominant firms is evaluated to determine the extent to which an auction-based tradeable permit market influences performance in an adjacent electricity market. The experimental treatment design controls for a variety of real-world institutional features, including variable demand, permit banking, inter-temporal (multi-round) dynamics, a tightening cap, and resale. Results suggest that the exercise of market power significantly increases electricity auction clearing prices, without significantly increasing emissions auction clearing prices, and in some cases, even significantly suppresses them. The institution of auction-based carbon markets in the already-concentrated energy sector can further strengthen the market position of dominant firms who can leverage energy-emissions market linkages to their operational advantage. ❧ Chapter 3: Regulatory Mechanisms and Policy Approaches ❧ Contemporary deregulated electricity markets are defined by a complex array of multi-settlement markets, with additional market-based mechanisms designed, to a large extent, to limit the exercise of market power by dominant firms. On top of the already complex nature of these markets, policymakers are also adding market-based mechanisms to curtail greenhouse gases. Key linkages exist between electricity and emissions markets that may be utilized by dominant firms. This chapter provides an analysis of three specific policy mechanisms that are utilized in contemporary markets to effectively reduce the incentive of dominant firms to exercise market power. These include convergence bidding, consignment auctions and multilevel holding accounts.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Processes, effects, and the implementation of market-based environmental policy: southern California's experiences with emissions trading

PDF

Household carbon footprints: how to encourage adoption of emissions‐reducing behaviors and technologies

PDF

The smart grid network: pricing, markets and incentives

PDF

Assessing the implementation of the RECLAIM cap-and-trade market for pollution: measurement issues in counterfactuals, goal attainment, and command-and-control alternatives

PDF

Why go green? Cities' adoption of local renewable energy policies and urban sustainability certifications

PDF

The economic and political impacts of U.S. federal carbon emissions trading policy across households, sectors and states

PDF

Life without nuclear power: a nuclear plant retirement formulation model and guide based on economics: San Onofre nuclear generating station case: economic impacts and reliability considerations ...

PDF

Smarter markets for a smarter grid: pricing randomness, flexibility and risk

PDF

A joint framework of design, control, and applications of energy generation and energy storage systems

PDF

Electric vehicle integration into the distribution grid: impact, control and forecast

PDF

The dynamic interaction of synchronous condensers, SVC, STATCOM and superconducting magnetic energy storage on electric vehicles

PDF

Geothermal resource assessment and reservoir modeling with an application to Kavaklidere geothermal field, Turkey

PDF

Avoiding middle-class planning 2.0: media arts and the future of urban planning

PDF

Advancing computational methods for free energy calculations in proteins and the applications for studies of catalysis in nucleotide-associated proteins

PDF

Three essays on the causes and consequences of China’s governance reforms

PDF

Effects of economic incentives on creative project-based networks: communication, collaboration and change in the American film industry, 1998-2010

PDF

Quango reforms and challenges in South Korea: social relations, informal networks, and hidden actions

PDF

Detecting joint interactions between sets of variables in the context of studies with a dichotomous phenotype, with applications to asthma susceptibility involving epigenetics and epistasis

Asset Metadata

Creator Dormady, Noah Christopher (author)

Core Title Emissions markets, power markets and market power: a study of the interactions between contemporary emissions markets and deregulated electricity markets

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Policy, Planning, and Development

Publication Date 07/02/2012

Defense Date 04/27/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cap and trade,cap-and-trade,economics experiments,electricity markets,emissions markets,energy economics,environmental economics,experimental economics,human experiments,laboratory experiments,market power,monopoly,monopsony,Monte Carlo,OAI-PMH Harvest,oligopoly,oligopsony,power markets

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Mazmanian, Daniel A. (committee chair), Jurewitz, John (committee member), Rose, Adam Z. (committee member), Wilkie, Simon J. (committee member)

Creator Email dormady.1@osu.edu,dormady@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-50005

Unique identifier UC11290297

Identifier usctheses-c3-50005 (legacy record id)

Legacy Identifier etd-DormadyNoa-911.pdf

Dmrecord 50005

Document Type Dissertation

Rights Dormady, Noah Christopher

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA