1

Market Structure and Competition:

A Cross Market Analysis of the U.S. Electricity Industry

Jim Bushnell, University of California Energy Inst,

Erin Mansur, Yale University, and Celeste Saravia, University of California at

Berkeley

2

3

Market Comparisons• Markets have experienced very

different levels of market power– Counter-factual competitive prices can tell

us the degree of market power

• But why have they performed differently?

• Market Structure vs. Market Design– Markets have important differences in both

structure and design

(cont.)

4

Market Comparisons• Market Structure vs. Market Design (cont)

– Much more policy emphasis on design• FERC “standard market design” (SMD) based upon

perception of a superior PJM design• FERC is willing to concede intervention in market

structure if firms are willing to submit to a SMD-like market design

• Role of Vertical Arrangements– Almost no policy discussion of impact of retail

policies on vertical arrangements – How much do spot market rules impact prices in

markets with significant forward arrangements?

5

Green and Newbery (1992)

60

$

Q

Dmax

Dmin

Cournot

competitive

Green and Newbery (1992)

7

0

$Dmax

DminBound on NCEquilibriumoutcomes

Cournot

competitive

Bounds on Non-Cooperative Outcomes

Qsupplied

8

0

$

Qsupplied

Dmax

DminBound on NCEquilibriumoutcomes

Cournot

competitive

Contracts Reduce Bounds

Contract Q

9

0

Dmax

Dmin Bound on NCEquilibriumoutcomes

Cournot

competitive

Contract QQsupplied

$

“Over-Contracting’ can drive prices below competitive

levels

10

Table 1: Panel A, California

Firm Characteristics Output Output Load Load Firm Fossil Water Nuclear Other Max Share Max Share PG&E 570 3,878 2,160 793 7,400 17% 17,676 39% AES/Williams 3,921 - - - 3,921 9% - - Reliant 3,698 - - - 3,698 8% - - Duke 3,343 - - - 3,343 8% - - SCE - 1,164 2,150 - 3,314 8% 19,122 42% Mirant 3,130 - - - 3,130 7% - - Dynegy/NRG 2,871 - - - 2,871 6% - - Other 6,617 5,620 - 4,267 16,504 37% 9,059 20% Total 24,150 10,662 4,310 5,060 44,181 45,857

11

Table 1: Panel B, New England

Firm Characteristics Output Output Load Load Firm Fossil Water Nuclear Other Max Share Max Share Northeast Util. 3,884 1,406 - 175 5,465 21% 7,560 34% PG&E N.E.G. 3,264 1,152 - 165 4,581 18% 4,440 20% Mirant 1,310 - - 16 1,943 8% - - Sithe 1,904 - - - 1,904 7% - - FP&L Energy 970 365 - - 1,335 5% - - Wisvest 980 - - - 980 4% 1,200 5% Duke Energy 474 - - 13 487 2% - - Other 2,291 1,663 4,359 819 9,132 35% 9,281 41% Total 17,276 3,180 4,359 1,013 25,828 22,481

12

Table 1: Panel C, PJM

Firm Characteristics Output Output Load Load Firm Fossil Water Nuclear Max Share Max Share Public Service 6,760 - 3,510 10,270 18% 8,947 17% PECO 3,682 1,274 4,534 9,490 17% 4,551 9% GPU, Inc. 7,478 454 1,513 9,445 17% 7,602 15% PP&L Inc. 6,102 148 2,304 8,554 15% 5,120 10% Potomac Electric 6,507 - - 6,507 11% 5,378 10% Baltimore G & E 3,945 - 1,829 5,774 10% 5,792 11% Delmarva P & L 2,458 - - 2,458 4% 3,103 6% Edison 2,012 - - 2,012 4% 0 0% Atlantic City 1,309 - - 1,309 2% 2,224 4% Other 428 439 - 867 2% 8,998 17% Total 40,681 2,316 13,690 56,685 51,714

13

Methodology

• Simulate prices under:– Price taking behavior– Cournot behavior– Cournot with vertical arraignments

, c tt i,t -i,t i,t i,t i,t i,t

, i,t

p=p (q ,q )+[q -q ]· -C (q ) 0

qi t

i tq

ci,tq

• Data on spot price, market demand and costs.

• The first order condition is:

14

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

6 1

7 24

2 2

ln( )S

fringet i it t s st

i s

j jt h ht tj h

q Month p Temp

Day Hour

• Estimates of price responsiveness are greatest

in California (>5000) relative to New England (1250) and PJM (850)

15

16

17

18

19

20

Price comparison on peak

PriceCaliforni

a

New England PJM

Actual 43 55 97

Competitive 35 41 37

Cournot 45 59 115

Cournot (“n.v.a.”)

Same 303 999

21

Price comparison off peak

PriceCaliforni

a

New England PJM

Actual 24 29 24

Competitive 26 32 28

Cournot 30 30 37

Cournot (“n.v.a.”)

Same 85 711

22

Policy and Research Implications

• Monitoring/Policy-making overly emphasizes spot market design and regulation– More focus on the contracting process is needed– This does not mean we should mandate contracting– We need to better understand how contract markets

can/do work in various markets– How do contract forms (futures, two-way “contracts

for differences”, options, etc.) matter?

• Prospective market power analysis needs to think about how vertical arrangements and contracting will effect outcomes

23

Caveats• Market structure during this period is exogenous• We measure only one aspect of market performance• We do not observe all contracts

– Have data on most long-term retail arrangements– Any ‘missing’ contracts likely to cause Eastern markets to

look worse relative to Cournot than California

• Prices and costs do not reflect transmission congestion– Not a real big issue in 1999

• Production costs ignore non-convexities– Start-up costs, ramp rates, etc.– Will bias costs upward off-peak, downward on-peak– Competitive estimates much more sensitive to cost bias than

Cournot.