Transcript Market Structure and Competition: A Cross Market Analysis

The Use of Oligopoly Equilibrium
for Economic and Policy
Applications
Jim Bushnell, UC Energy Institute
and Haas School of Business
A Dual Mission
• “Research Methods” - how oligopoly models
can be used to tell us something useful about
how markets work
– Potentially very boring
• What makes electricity markets work (or
not)?
– Blackouts, Enron, “manipulation,” etc.
– A new twist on how we think about vertical
relationships
– Potentially very exciting
Oligopoly Models
– Large focus on theoretical results
– Simple oligopoly models provide the
“structure” for structural estimation in IO
– Seldom applied to large data sets of
complex markets
• Some markets feature a wealth of detailed data
• Optimization packages make calculation of
even complex equilibria feasible
A Simple Oligopoly Model
• Concentration measures
2


q
2
i
HHI   s i   
Q
i
i  
p  MC 1 HHI
m


p
n

where m is Cournot equilibrium margin.
Surprising Fact:
Oligopoly models can tell us something
about reality
• Requires careful consideration about the institutional
details of the market environment
– Incentives of firms (Fringe vs. Oligopoly)
– Physical aspects of production (transmission)
– Vertical & contractual arrangements
• Recent research shows actual prices in several
electricity markets reasonably consistent with Cournot
prices
• Cournot models don’t have to be much more
complicated than HHI calculations
qi 
a  nqi
a  bk
,p
n  1  bc
b
Empirical Applications
• Analysis of policy proposals
– Prospective analysis of future market
– Merger review, market liberalization, etc.
• Market-level empirical analysis
– Retrospective analysis of historic market
– Diagnose sources of competition problems
– Simulate potential solutions
• Firm-level empirical analysis
– Estimate costs or other parameters (contracts)
– Evaluate optimality of firm’s “best” response
– Potentially diagnose collusive outcomes
Oligopoly equilibrium models
• Cournot – firms set quantities
– many variations
• Supply-function – firms bid p-q pairs
– infinite number of functional forms
– Range of potential outcomes is bounded by
Cournot and competitive
– Capacity constraints, functional form restrictions
reduce the number of potential equilibria
• Differentiated products models (Bertrand)
Green and Newbery (1992)
Simple Example
•
•
•
•
2 firms, c(q) = 1/2 qi 2, c = mc(q) = qi
Market supply = Q = q1 + q2
Linear demand Q = a-b*p = 10 – p
NO CAPACITY CONSTRAINTS
 a  ( qi  q j ) 
2
i  
q

.
5
q
i
i

b


 i  a  ( 2qi  q j ) 

 qi  0

qi 
b

a  q j 10  q j
BRi ( q j )  qi 

2b
3
2 Cournot Firms
Best Reply Functions
BR2(q1)
BR1(q2)
10
9
8
7
Firm 1
6
5
4
3
2
1
0
0
1
2
3
4
5
Firm 2
6
7
8
9
10
Three Studies of Electricity
• Non-incremental regulatory and structural changes
– Historic data not useful for predicting future behavior
• Large amounts of cost and market data available
– High frequency data - legacy of regulation
• Borenstein and Bushnell (1999)
– Simulation of prospective market structures
• Focus on import capacity constraints
• Bushnell (2005)
– Simulation using actual market conditions
• Focus on import elasticities
• Bushnell, Saravia, and Mansur (2006)
– Simulation of several markets
Western Regional Markets
• Path from NW to
northern California
rated at 4880 MW
• Path from NW to
southern California
rated at 2990 MW
• Path from SW to
southern California
rated at 9406 MW (WO-R constraint)
• 408 MW path from
northern Mexico and
1920 MW path from
Utah
12
Cournot Equilibrium and
Competitive Market Price for Base
Case - Elasticity = -.1
June
60
50
50
30
Dem and Level
September
December
Dem and Level
450th
300th
150th
500
450
400
350
300
250
200
150
100
50
0
Peak
720th
600th
450th
300th
150th
Price ($/Mwh)
1000
900
800
700
600
500
400
300
200
100
0
Peak
Price ($/Mwh)
720th
Dem and Level
744th
Peak
744th
150th
600th
0
450th
0
300th
10
600th
20
10
600th
20
450th
30
40
300th
40
150th
Price ($/Mwh)
60
Peak
Price ($/Mwh)
March
Dem and Level
13
Table 1: Panel A,
California Firm Characteristics
Firm
Fossil
Water
Nuclear Other
PG&E
570
3,878
2,160
793
AES/Williams
3,921
Reliant
3,698
Duke
3,343
SCE
1,164
2,150
Mirant
3,130
Dynegy/NRG
2,871
Other
6,617
5,620
- 4,267
Total
24,150 10,662
4,310 5,060
HHI of 620
Output Output Load
Load
Max
Share
Max
Share
7,400
17% 17,676
39%
3,921
9%
3,698
8%
3,343
8%
3,314
8% 19,122
42%
3,130
7%
2,871
6%
16,504
37%
9,059
20%
44,181
45,857
Methodology for Utilizing
Historic Market Data
• Data on spot price, quantity demanded,
vertical commitments, and unit-specific
marginal costs.
• Estimate supply of fringe firms.
– Calculate residual demand.
• Simulate market outcomes under:
– 1. Price taking behavior:
P
= C’
– 2. Cournot behavior: P + P’ * q
= C’
– 3. Cournot behavior with vertical arraignments:
P + P’ * (q-qc) = C’
Modeling Imports and Fringe
•
•
•
Source of elasticity in model
We observe import quantities, market price, and weather conditions in
neighboring states
Estimate the following regression using 2SLS (load as instrument)
9
S
i 6
s 1
qtfringe   i Monthit   ln( pt )    sTempst
7
24
j 2
h2
  j Day jt   h Hourht   t
• Estimates of price responsiveness are greatest in
California (>5000) relative to New England
(1250) and PJM (850)
Residual Demand function
 t  Qtactual   ln( ptactual )
  t  Qt 
pt  exp

  
• The demand curve is fit through the observed
price and quantity outcomes.
mean
mean
mean
observations
Cournot
PX
Competitive
price ($/MWh) price ($/MWh) price ($/MWh)
June
699
127.93
122.29
52.67
July
704
131.84
108.60
60.27
August
724
185.02
169.16
79.14
September
696
116.26
116.64
75.12
Table 1: Cournot Simulation and Actual PX Prices - Summer 2000
0
100
200
300
400
Simulation Results
California 2000
0
5000
10000
dem_cal
cournot
competitive
PX_price
15000
AES
Dynegy
Duke
Mirant
Reliant
SDG&E units
Alamitos
Ormond
Moss
South Bay
Actual
MW
HHI
3921
550
2871
295
3343
400
2886
298
3698
489
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
16718
2032
Counter-factual
MW
HHI
1873
126
1165
49
1585
90
2886
298
2306
190
1407
71
2048
150
1271
58
1474
78
704
18
16718
1126
Table 1: Actual and Hypothetical Thermal Ownership
Impact of Further Divestiture
(summer 2000)
June
July
August
September
Total
Total Cost Total Cost Total Cost
Cournot
Divest
Savings
$ Million $ Million $ Million
1870
1410
466
1830
1420
415
2800
2190
605
1570
1230
341
8070
6250
1827
Table 1: Market Savings from Further Divestiture
September 2000 Cournot Prices
Simple Cournot
actual
non-linear cournot
300.00
250.00
Price ($/MWh)
200.00
150.00
100.00
50.00
0.00
0
2000
4000
6000
8000
10000
Residual Demand
12000
14000
16000
18000
The Effect of Forward
Contracts
• Contract revenue is sunk by the time the spot
market is run
– no point in withholding output to drive up a price
that is not relevant to you
• More contracts by 1 firm lead to more spot
production from that firm, less from others
• More contracts increase total production
– lower prices
• Firms would like to be the only one signing
contracts, are in trouble if they are the only
ones not signing contracts
– prisoner’s dilemma
Simple Example
–
–
–
–
–
2 firms, c(q) = 1/2 qi 2, c = mc(q) = qi
Market supply = Q = q1 + q2
Linear demand Q = a-b*p = 10 – p
NO CAPACITY CONSTRAINTS
Firm 2 has contracts for quantity qc2
a  (q  q ) 
  
 q  q   .5q
b
2
2

c
1

2
2
2
2
 2  a  ( 2q2  q1 )  q c 2 

  q2  0
q2 
b

BR2 ( q1 )  q2 
a  q1  q c 2 10  q1  q c 2

2b
3
2 Cournot Firms
Best Reply Functions
BR2(q1)
BR1(q2)
10
9
8
7
Firm 1
6
5
4
3
2
1
0
0
1
2
3
4
5
Firm 2
6
7
8
9
10
2 Cournot Firms
Best Reply Functions
BR2(q1)
BR1(q2)
10
9
8
7
Firm 1
6
5
4
3
2
1
0
0
1
2
3
4
5
Firm 2
6
7
8
9
10
Green and Newbery (1992)
Bounds on Non-Cooperative
Outcomes
$
Dmax
Dmin
Cournot
Bound on NC
Equilibrium
outcomes
competitive
0
Qsupplied
Contracts Reduce Bounds
$
Dmax
Dmin
Cournot
Bound on NC
Equilibrium
outcomes
competitive
0
Contract Q
Qsupplied
“Over-Contracting’ can drive
prices below competitive levels
$
Dmax
Dmin
Cournot
Bound on NC
Equilibrium
outcomes
competitive
0
Contract Q
Qsupplied
Vertical structure and forward
commitments
• Vertical integration makes a firm a player in two
serially related markets
• Usually we think of wholesale (upstream) price
determining the (downstream) retail price
– Gilbert and Hastings
– Hendricks and McAfee (simultaneous)
• In some markets, retailers make forward
commitments to customers
– utilities – telecom services – construction
• In these markets a vertical arrangement plays the
same role as a forward contract
– a pro-competitive effect
Retail and Generation in PJM, 1999
100%
Other
90%
Other
Baltimore Gas & Electric
80%
Potomac Electric Power
70%
Baltimore Gas & Electric
PP&L Inc.
60%
Potomac Electric Power
50%
PP&L Inc.
PECO Energy
40%
PECO Energy
30%
Public Service Electric & Gas
Public Service Electric & Gas
20%
GPU Inc.
10%
GPU Inc.
0%
Retail
Generation
Retail and Generation in New England 1999
100%
90%
Other
80%
Other
70%
60%
Wisvest
FP&L Energy
50%
Sithe
Mirant
40%
PG&E N.E.G.
PG&E N.E.G.
30%
20%
Northeast Util.
Northeast Util.
10%
0%
Retail
Generation
Retail and Generation in California 1999
100%
90%
Other
Other
80%
70%
60%
SCE
Dynegy/NRG
Duke
50%
Mirant
40%
Reliant
30%
20%
AES/Williams
PG&E
10%
SCE
PG&E
0%
Retail
Generation
Methodology
• Use market data on spot price, market
demand and production costs.
• Simulate prices under:
– Price taking behavior
– Cournot behavior
– Cournot with vertical arraignments (integration or contracts)
• The first order condition is:
 i ,t
pt
=pt (qi,t ,q-i,t )+[qi,t -q ]·
-Ci,t (qi,t )  0
qi ,t
qi,t
c
i,t
qi,tc
Methodology
• Data on spot price, quantity demanded,
vertical commitments, and unit-specific
marginal costs.
• Estimate supply of fringe firms.
– Calculate residual demand.
• Simulate market outcomes under:
– 1. Price taking behavior:
P
= C’
– 2. Cournot behavior: P + P’ * q
= C’
– 3. Cournot behavior with vertical arraignments:
P + P’ * (q-qc)
=
C’
Summary
• Oligopoly models married with careful empirical
methods are a useful tool for both prospective and
retrospective analysis of markets
• Careful consideration of the institutional details of the
market is necessary
• In electricity, vertical arrangements (or contracts)
appear to be a key driver of market performance
– The form and extent of these arrangements going forward
will determine whether the “success” of the markets that are
working well can be sustained