more competitive electricity marketsidei.fr/sites/default/files/medias/doc/conf/eem/papers... ·...
TRANSCRIPT
Silvester van KotenCERGE-EI, Prague
Andreas OrtmannThe University of New South Wales, Sydney
More Competitive Electricity Markets:Structural vs Behavioral Measures
An Experimental Investigation Guided by Theory and Policy Concerns
1. Theoretical Physics
2. Bridge model experimentation
3. Bridge
1. Theory 2. Engineering 3. ImplementationTheoretical Physics Bridge Experimentation Bridge
Theoretical Economics Market ExperimentationExperimental & Computational
Economics
Competitive MarketE.g., Energy Market
Silvester van KotenCERGE-EI, Prague
Andreas OrtmannThe University of New South Wales, Sydney
More Competitive Electricity Markets:Structural vs Behavioral Measures
An Experimental Investigation Guided by Theory and Policy Concerns
Market share of the largest generator in the electricity
market in %
Concentration Generators
1. Behavioral Remedy• Introduce a forward market (Allaz & Vila, JET
1993)
2. Structural Remedy• Add one more competitor by divestiture
Introduction Setup Results
What is the most effective pro-competitive policy?
Behavioral Structural Design
Behavioral:forward market
Introduction Setup Results Behavioral Structural Design
( ) 60p Q Q
A
Spot MarketFriday
Week 1
Demand20
20Q=20+20=40P=20
B
2345Introduction Setup Results Behavioral Structural Design
Forward Market
( ) 60p Q Q
Spot Market
Monday
Friday
Trader
12
12 24
Demand12
12Q=12+12+24=48P=12
A
B
A
B
24
Week 12345Introduction Setup Results Behavioral Structural Design
( ) 60p Q Q
Spot Market
Q=40P=20
23Introduction Setup Results
( ) 60p Q Q
Spot Market+
Forward Market
Q=48P=12
Behavioral Structural Design
Introduction Setup Results
Spot MarketProfit Function
1 2 1 2[ ] 60p q q q q
1 1 2 1 1(60 )( )q q q f
1 2 160 2 0q q f
1 2 12 60q q f
Demand Schedule (Duopoly)
First Order Conditions
iq Total Production (Forward + Spot)
if Production sold in Forward Market( )i iq f Production sold in Spot Market
Price Spot Market Production
Allaz & Villa (JET 1993): Cournot competitionBehavioral Structural Design
Introduction Setup Results
1 2 12 60q q f
1 20 & 0f f
1 212 & 12f f
1 215& 0f f
1 2 400
1 2 288
1 450 2 225
Reaction function 15
Nash-Equilibrium
“Backstabbing”
“Forward market boycott”
(“Stackleberg Equilibrium”)
12
Behavioral Structural Design
• Theory shows that a forward market has a pro-competitive effect (Allaz & Villa, JET, 1993)
Introduction Setup Results Behavioral Structural Design
2 Firms
3 Firms
4 Firms
Without Forward Market
M2 M3 M4
92.7% 102.7% 102.9%
“2 are few and 4 are many” Huck et al. (JEBO, 2004)
Introduction Setup Results
Can we trust this theory?
Behavioral Structural Design
2 Firms
3 Firms
4 Firms
Without Forward Market
M2 M3 M4
With Forward Market
M2F M3F –
92.7% 102.7% 102.9%
“2 are few and 4 are many” Huck et al. (JEBO 2004)
80%? 92%?100%? 110%?
Introduction Setup Results
Can we trust this theory?
Behavioral Structural Design
Structural:One more competitor
Introduction Setup Results Behavioral Structural Design
What is the right comparison?
• By entry?1. Increased competition2. Cheaper aggregate production3. Capital cost of new plants
• By divestiture?1. Only increased competition
Brandts et al (EJ, 2008)Add one more competitor:
Introduction Setup Results Behavioral Structural Design
• Treatments:• M2, M2F, M3, M3F & M4
• Demand:• As in Brandts et al (2008)
• Costs: Steeply increasing marginal costs• (Newbery, EER 2002).• As in treatment M3 of Brandts et al (2008)
( ) (0,2000 27 )p Q Max Q
Introduction Setup Results Behavioral Structural Design
Introducing a forward market
Adding one more competitor by divestiture
Experiment comparing
with
Behavioral measure Structural measure
Introduction Setup Results
23( ) 2mc q q
2
13( ) 2
q
xc q x
3 22 13 3x x x
Behavioral Structural Design
M2 M3M2 M3Behavioral Structural DesignIntroduction Setup Results
M2 M3Introduction Setup Results Behavioral Structural Design
Introduction Setup Results
M2 M3Behavioral Structural Design
Introduction Setup Results
M2 M3
M3 M4
Behavioral Structural Design Predictions
Introduction Setup Results
M2 M3
M3 M4
Behavioral Structural Design Predictions
Introduction Setup Results
M3 M1
⅓q
⅓q
CM3(⅓q)
CM3(⅓q)
⅓q
CM3(⅓q)
CM1(q) = 3 CM3 (⅓ q)
Behavioral Structural Design Predictions
cost
cost
cost
cost
cost
cost
Introduction Setup Results
M3 M1
⅓q
⅓q
CM1(⅓q)
CM1(⅓q)
⅓q
CM1(⅓q)
131 3( ) 3 ( )M MC q C q
11( ) ( )xM MxC q xC q
( ) ( )yxxMy MxyC q C q
Behavioral Structural Design Predictions
Introduction Setup Results
2
13( ) 2
q
xc q x
3 22 13 3x x x
3 22 32 3
[ ] [ ]c q c q4 3
3 44 3 [ ] [ ]c q c q
M3
M2 M4
Behavioral Structural Design Predictions
Market with TWO
producers
Market with THREE
producers (original market)
Market with FOUR
producers
Total Production
2*q
Total Costs
2*TC
Total Production
3*q
Total Costs
3*TC
Total Production
4* q
Total Costs
4*TC
0 0 0 0 0 0 2 3 3 6 4 11 4 11 6 30 6 30 8 62 8 62 10 111 9 84 12 180 12 180 12 180
364 112
Adding competition byEntry
(Brandts et al. 2008)
• Ran main sessions in:– October 2009, December 2009, and April 2010
• Ran robustness tests in:– October 2010 and January 2013
• 11 independent obs (groups) for each treatment• Total of 198 subjects
– Prague business school• Average Earning 500CKZ = €20
– PPP: €34– Minimum: 330 CKZ– Maximum: 1080 CKZ
Introduction Setup Results Behavioral Structural Design Predictions
Introduction Setup Results Behavioral Structural Design Predictions
Introduction Setup Results
PredictionsBehavioral Structural Design Predictions
2 Firms
3 Firms
4 Firms
Without Forward Market
M2 M3 M4
With Forward Market
M2F M3F –
40
40/44
43
45
44
3035
4045
5055
1 4 8 12 16 20 24Period
M2_mean M2F_meanM3_mean
M2, M2F, M3
Introduction Setup Outcomes Results Conclusion
3035
4045
5055
1 4 8 12 16 20 24Period
M3_mean M3F_meanM4_mean
M3, M3F, M4
Introduction Setup Outcomes Results Conclusion
2 Firms
3 Firms
4 Firms
Without Forward Market
M2 M3 M4
With Forward Market
M2F M3F –
Averages
Standard errors based on groups (N=11)
39.4
46.1
44.1
49.4
46.1
115%105%
98.7% 102.5%
110.0%
104.9%Confirming meta-analysis Huck et al. (JEBO 2004)
Percentages of the Nash-Equilibrium prediction
Introduction Setup Outcomes Results Conclusion
8 12 16 20 24Period
M2_mean M2F_meanM3 mean
M2, M2F, M3
Introduction Setup Outcomes Results Conclusion
M2F
M2
***
39.3
46.3
8 12 16 20 24Period
M2_mean M2F_meanM3 mean
M2, M2F, M3
Introduction Setup Outcomes Results Conclusion
M3
M2**
M2F
M2
***44.2
39.3
8 12 16 20 24Period
M2_mean M2F_meanM3 mean
M2, M2F, M3
Introduction Setup Outcomes Results Conclusion
M2F
M2
***M3
M2**
M2F
M3ns
44.2
46.3
12 16 20 24Period
M3, M3F, M4
Introduction Setup Outcomes Results Conclusion
M3F
M3***
44.2
49.6
12 16 20 24Period
M3, M3F, M4
Introduction Setup Outcomes Results Conclusion
M3F
M3***
M4
M3ns
44.2
46.2
12 16 20 24Period
M3, M3F, M4
Introduction Setup Outcomes Results Conclusion
M3F
M3***
M4
M3ns
M3F
M4***
46.2
49.6
Introduction Setup Results Behavioral Structural Design
Introducing a forward market
Adding one more competitor by divestiture
Conclusions of comparison
with
Behavioral measure Structural measure
• Are equally effective in M2• Behavioral measure more effective in M3
– Contrast with Brandts et al (2008) • Are equally effective in M3 if adding one more
competitor is done by entry
Silvester van Koten,CERGE-EI, Prague
Andreas Ortmann, The University of New South Wales, Sydney
Structural versus Behavioral Measures in the Deregulation of Electricity Markets:
An Experimental Investigation Guided by Theory and Policy Concerns
Introduction Setup Outcomes Results Conclusion Complication
Observed Forward Market
Predicted
Inexperienced
Experienced
2 firms 85.7 85.6 62.5 4 firms 89.1 99.9 76.8
Ferreira, Kujal & Rassenti, 2009
Are these results robust for experienced players?
M2, M2F, M3
Introduction Setup Outcomes
Experienced
Inexperienced
3035
4045
5055
1 4 8 12 16 20 24Period
M2_mean M2F_meanM3_mean
3035
4045
5055
Aggr
egat
e Pr
oduc
tion
1 4 8 12 16 20 24Period
run4xM2_mean run4xM2F_meanrun4xM3_mean
Results Conclusion Complication
3035
4045
5055
1 4 8 12 16 20 24Period
M3_mean M3F_meanM4_mean
3035
4045
5055
Agg
rega
te P
rodu
ctio
n
1 4 8 12 16 20 24Period
run4xM3_mean run4xM3F_meanrun4xM4_mean
Introduction Setup Outcomes
Experienced
Inexperienced
M3, M3F, M4
Results Conclusion Complication
Introduction Setup Outcomes
M2 M2F M3 M3F M4 run123 (Inexperienced)
39.3 (1.5)
46.3 (2.0)
44.2 (1.2)
49.6 (0.6)
46.2 (1.0)
run4 (Experienced)
43.1 (1.5)
45.7 (2.4)
42.0 (1.6)
50.9 (0.2)
46.4 (0.9)
Effect Experience
+3.8 -0.6 -2.2 +1.3 +0.2
Significance (two-sided test)
0.08 0.85 0.25 0.03 0.89
Results Conclusion Complication
M2, M2F, M3
Introduction Setup Outcomes
Experienced
Inexperienced
3035
4045
5055
1 4 8 12 16 20 24Period
M2_mean M2F_meanM3_mean
3035
4045
5055
Aggr
egat
e Pr
oduc
tion
1 4 8 12 16 20 24Period
run4xM2_mean run4xM2F_meanrun4xM3_mean
Results Conclusion Complication
3035
4045
5055
1 4 8 12 16 20 24Period
M3_mean M3F_meanM4_mean
3035
4045
5055
Agg
rega
te P
rodu
ctio
n
1 4 8 12 16 20 24Period
run4xM3_mean run4xM3F_meanrun4xM4_mean
Introduction Setup Outcomes
Experienced
Inexperienced
M3, M3F, M4
Results Conclusion Complication
-10
010
2030
% IN
CR
EASE
1 4 8 1 2 16 20 2 4Perio d
Increase in production by Experienced SubjectsIntroduction Setup Outcomes Results Conclusion Complication
Silvester van KotenCERGE-EI, Prague
Andreas OrtmannThe University of New South Wales, Sydney
More Competitive Electricity Markets:Structural vs Behavioral Measures
An Experimental Investigation Guided by Theory and Policy Concerns
Plot & Uhl, SEJ, 1981
Plot & Uhl, SEJ, 1981
Silvester van KotenCERGE-EI, Prague
Andreas OrtmannThe University of New South Wales, Sydney
More Competitive Electricity Markets:Structural vs Behavioral Measures
An Experimental Investigation Guided by Theory and Policy Concerns