“Recent Electricity Auctions”
Dr. David J. Salant
Senior Vice President
National Economic Research Associates
June 2003
Agenda
 Electricity restructuring, auctions and new
markets
 New Jersey Basic Generation Service auction
 Texas Capacity auctions
 CALPX/CAISO
 Alberta
 FTR auctions
 Parallels in other states and other types of
electricity markets
2
Auctions and Regulation
Auctions have been used increasingly to replace
regulatory processes
Spectrum auctions replaced beauty contests and lotteries
Default Service Auctions
 Customers who are not being served by a competitive “third party”
supplier must be served by the electricity distribution company at
regulated rates
 Auctions now have been used for utility energy purchases to serve
these consumers
 RFPs tend to be followed by bilateral negotiations, providing more
scope for regulators to question the results.
 Entitlements and PPAs
Utilities required to divest assets
Rather than sell off entire assets, auctions now are used to
sell entitlements, strips, PPAs, as in Texas and Alberta
3
New Types of Electricity Auctions
 Simultaneous clock auctions
 Used for buying or selling multiple units of a few types of lots
or products, such as system slices or energy entitlements.
 First application for default service procurement was
Simultaneous Descending Clock Auction (SDCA) which NERA
developed for New Jersey.
 Variations of clock auctions have been used to sell energy
entitlements in Texas capacity auctions and French VPPs.
 Clock auctions are well suit for interconnection capacity.
 Simultaneous multiple round auctions has been used on
one occasion for selling PPAs in Alberta.
 Other, more traditional, auctions such as Yankee auctions
and English auctions used for energy entitlements in
Alberta and for interconnection in Netherlands did not work
well.
4
Auction Design Objectives
Auction design adopted should best match objectives
 Simultaneous multiple round/clock auctions are appropriate
for efficiently allocating multiple lots, with value
interdependencies, such as energy entitlements, capacity and
spectrum licenses
 Theory suggests that simultaneous auctions will result in
economically efficient assignments assuming substitutes and
“straightforward” bidding
 Simultaneous auctions work adequately when bidders have
similar views about complements
 Clock auctions are well suited for dividing shares of load.
 Simultaneous auctions can create some incentives for bidders
to withhold supply, but these incentives can be mitigated with
volume adjustments
 Simultaneous auctions work less well with strong, overlapping
complements
5
The New Jersey
BGS Auction
6
The New Jersey BGS Auction
 Four utilities were under legislative mandate to
purchase energy in a competitive bidding
process.
 NJ EDCs needed to secure one-year forward supplies
for approximately 17,000 MW of forecast peak load
 Auction was for one year forward contracts.
 The auction outome
 Prices appeared to be competitive, between 4.87¢ and
5.82 ¢ for entire year.
 Over 20 bidders competed to sell and there were 15
winners.
 in first ever simultaneous descending price clock
auction.
 New Jersey Board rendered decision on auction
results within 48 hours of the close of the auction
7
New Jersey Winning Bidders
BGS Winning Bidders for Year 4
Winning Bidder Number of Tranches
Won per EDC territory
ALLEGHENY ENERGY SUPPLY
AMERADA HESS CORPORATION
AQUILA ENERGY MARKETING
CONECTIV ENERGY SUPPLY INC
CONSOLIDATED EDISON ENERGY
DTE ENERGY TRADING INC
DUKE ENERGY TRADING
FIRSTENERGY SOLUTIONS CORP
MIECO
NRG ENERGY
PPL ENERGY PLUS CORP
SELECT ENERGY INC
SEMPRA ENERGY TRADING CORP
TXU ENERGY TRADING
WILLIAMS ENERGY MARKETING & TRADING
PSE&G
JCP&L
Conectiv
5.112 ¢/kWh
4.865 ¢/kWh
5.117 ¢/kWh
15
9
15
RECO
5.819 ¢/kWh
1
5
1
3
20
10
1
5
2
5
5
3
1
6
7
12
15
9
3
8
5
4
8
What was Being Purchased?
 Full requirements slices of BGS load of each EDC
 Payments will be based on load measured at the
PJM interface—no risk of losses
 Slices set at roughly 100 MW resulted in




96 PSE&G tranches
51 GPUE/JCP&L tranches
19 Conectiv tranches
4 RECO tranches
 Starting prices set by auction manager/EDCs
 Maximum and minimum possible starting prices in the
EDC filing
 Not capped by shopping credits
 Will be based on indicative bids
 Will provide adequate risk premiums
 Auction manager had discretion to restrict
auction volume if competition proves very limited
9
Market Background
 NJ EDCs BGS requirements of nearly 17,000 MWs
represented over 95% of all NJ energy consumption
 Total native capacity of approximately 20,000 MW including
some NUGs (although 29,600 MW showed up at the start of
the auction)
-
PS Power controlled 57% of native resources
The top four firm concentration, HH4 = 76%
Limited import capacity from South and West through PJM
Energy prices North and East in NYISO tends to be higher than
in NJ
 PJM structure facilitated competition in the auction
- FTR allocation coordinated with BGS contract
- PJM spot market provided options for both buyers and sellers
10
The New Jersey Year 5
BGS Auction
11
Changes to the BGS Auction in Y5
 BGS Load Split into two groups:
 Hourly Electric Price (CIEP/HEP) for large corporate and
industrial customers
 Fixed Price (FP) for small and residential customers
 Separate but concurrent SDCAs will be held:
 HEP Auction is for capacity ($/MW-day)
 FP Auction is for all-inclusive price (¢/kWh, same as Y4)
 Regulatory approval for each auction’s result is
separate.
 Winning Bidders sign different contracts; BGS-HEP
Supplier Master Agreement and BGS-FP Supplier Master
Agreement differ.
12
Design of the New Jersey
BGS Auction
13
BGS Auction Rules—Overview
 The standard Simultaneous Multiple Round (SMR)
auction format
 Bidding in rounds
 Reverse auctions—the sellers bid
 Uniform pricing
 Form of bids—quantities instead of prices
 Total bids cannot increase
 Switching—suppliers can switch between EDCs
during auction
 Ending the auction
14
Sample Results – Start of Auction
EDC
PSEG
JCP&L
Conectiv
RECO
Round
1
$65.00
$65.00
$70.00
$62.00
Totals
EDC
PSEG
JCP&L
Conectiv
RECO
Totals
# Bid
142
84
55
6
287
Round
2
$63.44
$62.90
$63.37
$60.45
# Bid
162
86
31
6
285
# Available
96
51
19
4
Ratio
1.48
1.65
2.89
1.50
170
1.6882
# Available
96
51
19
4
170
Ratio
1.69
1.69
1.63
1.50
Price %Δ
2.40%
3.24%
9.47%
2.50%
P rice %Δ
3.44%
3.43%
3.16%
2.50%
1.6765
15
Sample Results – Near Auction End
EDC
PSEG
JCP&L
Conectiv
RECO
Round
42
$52.10
$53.10
$49.50
$56.40
Totals
EDC
PSEG
JCP&L
Conectiv
RECO
Totals
# Bid
102
62
20
4
188
Round
43
$51.94
$52.53
$49.37
$56.40
# Bid
100
63
20
4
187
# Available
96
51
19
4
Ratio
1.06
1.22
1.05
1.00
170
1.1059
# Available
96
51
19
4
170
Ratio
1.04
1.24
1.05
1.00
Price % Δ
0.31%
1.08%
0.26%
0.00%
Price %Δ
0.21%
1.18%
0.26%
0.00%
1.1
16
Exit Bids
 The following situation is possible:
Round 44
price /kWh
Tranches bid
Round 45
price /kWh
Tranches bid
PSE&G (20)
5.2¢
97
5.175¢
94
JCP&L (12)
5.1¢
51
5.1¢
51
Conectiv (5)
5.05¢
19
5.05¢
19
RECO (1)
6.2¢
4
6.2¢
4
17
Exit Bids
 When bidders reduce their quantity, they can elect
to submit an exit price
 An exit price is a final price for a slice on which a
bidder will no longer be bidding.
 Exit prices are
 EDC specific
 Required to be below the previous round’s going price
and above the current round’s price e.g., between 5.2¢
and 5.175¢
 If auction ends, slices would be allocated at the
exit price of the slice that just fills the load for the
product
18
Other Auction Features
 Additional Rules
 Size of decrement depended on excess of tranches
subscribed over number of tranches available
 If an EDC’s tranches are just subscribed or undersubscribed, bidders were not allowed to decrease
number of tranches offered
 Switching restrictions—switching was not allowed if it
would result in under-subscription
 Load caps—each EDC imposed a maximum on the
number of tranches that a bidder can supply
 Pace
 Length of rounds were kept as short as possible, but to
still allow bidders time to decide on when to reduce
eligibility or switch
 Bidders allowed limited recesses toward end of auction
19
Auction Closing Rules
 As long as there is some excess supply for at
least one product (more slices subscribed than
available)
 Bidding continues
 Prices continue to tick down from one round to the next
 At a given point in the auction there can be
excess supply for one product but not for others
 For auction to end
 Bidding on all products must have stopped
 The offered supply equals the number of available slices
for all products
 Bidders are paid the closing prices per MWh
20
Starting Prices and
Auction Volume Adjustments
 Starting prices and auction volume
adjustments are two instruments to limit
adverse consequences if there is limited
competition in the auction
 The Auction Manager announced starting
prices three days prior to auction
 Starting prices were required to be above PJM forward
curve by a percentage that was approved by the BPU
 Indicative bids were due two weeks prior to the auction
will factor into the setting of the starting prices
21
Starting Prices and
Auction Volume Adjustments
Continued…
 Target auction volume could be adjusted
based on total interest
 If supply, as expressed in the indicative bids is limited,
auction volume could have been reduced
 Target auction volume was to have been set at no more
than a multiple of initial supply
 Further auction volume adjustments permitted based on
competition
22
Auction Design Issues
 Factors affecting expected supplier costs/bidder values
 NUG contracts—tranche size depended on existing NUG
contracts
 Line losses in the distribution system
 Capacity obligations
 BGS suppliers are required to satisfy the renewable energy
(green) standards. Combined rate ranges from 3% in 2002
gradually up to 6.5% in 2012
 Starting prices
 Back-up provisions if supply was limited—auction volume
adjustments
 Minimum stay rules, customer switching, slamming
 Implementation—credit, qualification, software
23
The Texas Capacity
Auctions
24
The Texas Capacity Auctions
 The Texas Public Utility Commission (PUC)
mandated auctions of capacity by Power
Generation Companies (PGCs) beginning August
23, 2001
 All the PGCs (AEP, Reliant and TXU) have been
required to sell at least 15% of total capacity
 The PUC has imposed specific auction
mechanics—including simultaneous auctions of
all PGC capacity of a given duration:
 Auctions were initially parallel and simultaneous
auctions
 Activity rules initially were not comparable to usual SMR
auction formats
 Due to price gaps, PUCT adopted switching rule
25
The Texas Capacity Auctions
Continued…
 Sellers were provided some latitude in setting
starting prices and minimum bid increments
 Differences in approaches to starting prices and
bid increments drove auction dynamics in some
cases
26
Auction Background
 Products are 25 MW entitlements defined by
 Zone (North, South, West and Houston (except first
auction))
 Type (baseload, gas-intermediate, gas-cyclic, gaspeaking)
 Seller (AEP, Reliant, TXU)
 Term (2-year, 1-year, 1-month)
 Timing
 Quarterly auctions
 At each quarterly auction, there were:
 All 2-year contracts were auctioned simultaneously
(all types in all zones by all PGCs)
 Then, all 1-year contracts were auctioned
simultaneously
 All 1-month contracts were auctioned
simultaneously were auctioned last
27
Implications of Texas Experience for
FTR/CRR and Capacity Auctions
 FTR auctions in CA and elsewhere do not allow
switching between substitute paths
 As in Texas, equivalent FTRs can and are likely to
sell for much different prices
 Lack of switching implies
 Bidders need to guess when developing bidding
strategy
 Allocative inefficiency when higher value bid on one
FTR loses to lower value bid on a near or perfect
substitute FTR (or CRR)
 NYISO, ISO-NE, PJM are also planning to run
parallel auctions for each zone
28
Additional Factors in CRR/FTR Auction
Design
 Kirchoff’s law implies possible non-convexities in set of
feasible power transfers
 Non-convexities of feasible set will imply that support
prices (LMP) will result in corner solutions
 Implications for auction is that strong incentives can exist
for bidders to strategically under or over report net
demands
 Package bidding and contingent bidding can be useful in
finding optimal allocations
 Testing can involve simulation and experiments
 Recent developments in package bidding include SAAPB
adopted by FCC and Ausubel-Milgrom proxy bidding
procedures
29
Alberta PPA Auctions
30
Alberta PPA and MAP Auctions
Alberta conducted two sets of auctions
 PPA auction for 12 PPAs in late 2000
 MAP auction for strips from unsold PPAs in
2001
PPAs for energy entitlements of generating
units
Auction was a variation SMR auction
format used in NJ BGS and developed for
FCC spectrum auctions
31
Alberta Power Purchase
Arrangement:
Auction Results
Winning
Bidder
PPA
Battle RiverEPCOR
Clover Bar
Genesee
Keephills Enmax
Rainbow Engage
Rossdale Engage
Sheerness
Sturgeon
Sundance ATransCan.
Sundance BEnron
Sundance C
EPCOR
Wabamun Enmax
Fuel
Coal
Gas
Coal
Coal
Gas
Gas
Coal
Gas
Coal
Coal
Coal
Coal
Capacity
(MW)
PPA
Term
Minimum
Opening
Bid (C$)
663
n/a
n/a
766
93
203
n/a
n/a
560
706
710
548
2020
2010
2020
2020
2005
2003
2020
2005
2017
2020
2020
2003
$50
($96)
($300)
$50
($21)
$0
($200)
$0
$50
$50
$50
$25
Amount
Bid (C$
Million)
Merit Order
$84.90
Baseload
did not sell
Peaker
did not sell
Baseload
$240.70
Baseload
($21)
Peaker
0
Peaker
did not sell
Baseload
did not sell Not Running
$211.90
Baseload
$294.80
Baseload
$268.50
Baseload
$75.10
Baseload
Note:
32
CALPX/CAISO
33
California Energy Markets
Three utilities – PGE, SDG&E/Sempra and Southern California
Edison
Peak demands of approximately 45K MW in summer months
 18% of supplies from imports
 Five main generators controlled nearly half of fossil fuel generation within
CA
Summer 2000 crisis began and continued past the close of the
CALPX in Feb. 2001 with
 Stage 2 emergencies declared on scattered days in May – September
2000
 Stage 3 emergencies in Dec. 2000 (1 day), January 2001 (18 days),
February 2001 (16 days), March 2001 (2 days) and May 2001 (2 days).
 Average loads of < 33K MW through the latter part of the crisis
AB 1890 mandate IOUs purchase energy through CALPX/CAISO
34
Some Explanations of What Happened?
 Strategic withholding of supplies (JK & BBW) and less than
perfectly competitive behavior (Puller)
 Lack of forward contracts
 CALPX had block forward market,
 IOUs were permitted to purchase much more block forward
than they did and they could hedge with derivatives
 Ex post regulatory review of forward/derivative purchases
placed all the downside risk with IOU, so this was not a factor
 Ability of other WECC buyers to be active in forward
markets and inability of CA IOUs placed burden on latter
(Allaz and Vila/Salant&Loxley – NJ experience)
 Lack of capacity credits/markets or regulation
 Nature of equilibrium in CALPX/CAISO SFE type auctions
35
NJ vs. California
NJ
CA

3 large utilities

Moderate HHI in generation

Capacity and other reserves
needed to respond to demand
peaks no longer mandated

CAISO/WECC – provides links
between zones

High price spikes

CALPX/CPUC did not give
utilities much of an out

Theoretically unsound Supply
Function Auction format







3 large utilities
High HHI > 3200 in generation
Significant transmission constraints
to the remain of PJM
Capacity credit market mandates
reserve margins
Outcome competitive average 5.1¢
for entire year
Combination of primary long term
contracts and options for spot and
short term contracts gave more
flexibility for buyers to get a more
competitive price
Clock auction designed to attenuate
market power of the sellers
36
Example: Price Competition with
Capacity Constraints
 One buyer who demands D units at any price up to
R.
 Two sellers. Each seller can produce up to its
capacity K at zero marginal cost.
 K < D, so neither seller can supply the whole market.
 Sellers compete in prices – each seller names a
single price at which it will sell up to K units.
 Supply is dispatched in order of increasing price,
and sellers are paid as bid.
37
No MC Equilibrium
Price
Demand
p2 = R
Profmin
p1 = 0
Supply
Units
K
D
If both set p = MC, then each sells D/2 at price 0, earning a profit of 0
If Seller 2 charged R instead, it would sell D – K earning a profit of Profmin =
R(D – K).
38
Applicability of SDCA in
Other States
39
US Experience with Competition in
Generation Services
Customer Participation in Retail Access
Customers of Total Customers
Migration
State
Population Competitive Eligible for Retail
Rate
Suppliers
Choice
Arizona
5,456,453
n/a
n/a
n/a
California
35,116,033
72,422
10,580,906
0.68%
Connecticut
3,460,503
n/a
n/a
n/a
District of Columbia 570,898
25,115
197,359
12.73%
Delaware
807,385
n/a
n/a
n/a
Illinois
12,600,620
27,896
687,980
4.06%
Maine
1,294,464
8,713
Maryland
5,458,137
74,870
2,074,243
3.61%
Massachusetts
6,427,801
84,532
2,544,495
3.32%
Michigan
10,050,446
n/a
n/a
n/a
New Jersey
8,590,300
2,573
3,651,148
0.07%
New York
19,157,532
388,308
7,279,618
5.33%
Ohio
11,421,267
747,951
4,681,053
15.98%
Oregon
3,521,515
36,503
1,213,858
3.01%
Pennsylvania
12,335,091
278,429
n/a
n/a
Rhode Island
1,069,725
2,132
468,015
0.46%
Texas
21,779,893
469,106
n/a
n/a
Virginia
7,293,542
2,584
1,300,763
0.20%
#Large Commercial; *Small Commercial; n/a = Not Available; n/o = Not Open
Report
Date
n/a
2-Dec
n/a
2-Sep
n/a
2-Oct
3-Jan
2-Dec
2-Nov
n/a
2-Dec
2-Nov
2-Sep
2-Dec
3-Jan
2-Dec
2-Sep
2-Aug
Competitive Competitive Both Competitive
Total
Report
Residential Commercial C&I
Industrial Competitive
Date
Load
Load
Load
Load
Load
0.00%
0
0.00%
0.00%
2-Oct
0.80%
1.4*/13.9#
35.70%
13.30%
2-Aug
1.30%
0.20%
1.20%
2-Sep
11.10%
59.60%
49.40%
2-Aug
n/a
n/a
n/a
2-Sep
0.00%
26.90%
50.10%
25.80%
2-Jul
1.60%
30.40%
72.10%
34.80%
2-Sep
3.90%
29.10%
17.20%
2-Aug
2.20%
11.4*/17.4#
43.50%
21.60%
2-Aug
n/a
n/a
n/a
7.30%
2-Jul
Jun/02n/a
n/a
1.80%
Aug/02
5.50%
26.20%
18.90%
2-May
13.90%
15.20%
11.70%
12.90%
2-Jun
n/a
0.00%
0.00%
2-Sep
5.60%
10.70%
11.00%
8.70%
2-Oct
n/a
n/a
n/a
12.90%
2-Jun
4.80%
27.20%
81.10%
38.20%
2-Jul
n/a
n/a
n/a
n/a
n/a
n/a
40
Division of Load and Contract Terms
 Many options for dividing up load
 Uniform slices of system
 Division by customer segment
 Firm and non-firm tranches
 Metered vs. non-metered customer segments
 Contract terms
 Wholesale vs. resale
 Fixed price service vs. variable (market) price
 Seasonal adjustments
 Duration
 Uniform vs. non-uniform
 Length
41
Contract Duration
SDCA can permit contracts of different
durations – e.g., Texas capacity auctions
 Some longer term contracts can facilitate
bidder financing
 Some shorter term contracts can allow
bidders to adjust portfolios, and can be
especially useful for bidders with plants
coming on or going off line.
42
CA: Demand and Supply Concentration
 As in NJ experience, CA has three main utilities
 Supply in California is less concentrated than in
NJ
 Import capacity comparable in both regions
 Affiliation of PS Power with PSEG for
approximately 50% of native energy was potential
limitation of benefits of the auction
 PJM FTR approach facilitated competition in
auction for default service load procurement
 CAISO only control intra-state transmission, interstate transmission with rest of WECC could limit
competition in an auction
43
Forward vs. Day Ahead Markets
 Auction volume adjustments left open possibility
of load procurement in day ahead or hour ahead
markets
 Suppliers apparently wanted to avoid having to
sell in what could be a competitive day ahead
market, and so had incentives to bid aggressively
in BGS auction
 Starting prices and process were reviewed and
approved by BPU in advance of auction
 Auction process required BPU review within 48
hours of auction close
 EDC and supplier exposure to risks were both
mitigated by prior approval and transparency of
process
44