Imperfect Markets versus Imperfect Regulation in the U.S ... Steve Cicala Background on Electricity Dispatch February 12, 2018. Imperfect Markets versus Imperfect Regulation ... Nameplate
Post on 26-Jun-2018
217 Views
Preview:
Transcript
Imperfect Markets versus Imperfect Regulation
Imperfect Markets versus Imperfect
Regulation in the U.S. Electricity
Industry
Steve Cicala
University of Chicago
February 12, 2018
Steve Cicala February 12, 2018
Imperfect Markets versus Imperfect Regulation
Research Questions
Big Picture Question: How do markets performrelative to regulated allocation mechanisms, inclusiveof imperfections?
Application to electricity: How has transition tomarkets for determining production affected theallocative efficiency of electricity generation?
Gains from Trade. Market Power problems.
Steve Cicala Introduction February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
Construct the universe costs, capacities, and hourlyoperations of nearly all electrical generating units inthe U.S. (1999-2012) ($1T).
Calculate measures of allocative efficiency Value of traded electricity Cost of out of merit generation.
Use staggered, discrete changes in how plants areselected to operate to estimate causal effect of marketintroduction on operations, net allocative efficiency.
Steve Cicala Introduction February 12, 2018
Imperfect Markets versus Imperfect Regulation
Preview of Findings
Large improvements in allocative efficiency in marketareas.
Following Market Dispatch: Gains from trade increase 20% from a 10% increase in
traded volume. Plant availability increases 10%, reducing out of merit
costs by 20%.
More than offsets potential market power losses.
Steve Cicala Introduction February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline of Talk
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Introduction February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Demand Fluctuations Over the Week350
400
450
500
GW
h
0 24 48 72 96 120 144 168
Hour of Week
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Demand Fluctuations Over the Year250
300
350
400
TW
h
Jan1998 Jan2000 Jan2002 Jan2004 Jan2006 Jan2008 Jan2010 Jan2012
MonthYear
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Generation Units: Leontief, Capacity Contrained
Outputit = min {AiHeatit , Capacityi ,Laborit}
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Agile Units for Peak Times: 50MW
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
High Capacity Units for Base Load: 1750MW
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Supply Curves Order Generating Units0
200
400
600
800
$/M
Wh
0 10000 20000 30000Cumulative Nameplate
NEISO 1aug2005 0:00
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Power Control Areas in 1997
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Brief History of the Electricity Sector
Was long believed to be a natural monopoly. Grew up as a balkanized patchwork of regulated,
vertically integrated utilities.
Rates were set by State Public Utility Commissions. Prudent cost recovery ensured regulation of
investments and operations.
Units dispatched based on engineering estimates ofcost within 100 Power Control Areas.
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Brief History of the Electricity Sector
Federal Energy Regulatory Commission (FERC)begins encouraging competitive generation as part ofbroader deregulatory push in late 1970s.
Late 1990s: Independent System Operators (ISO) /Regional Transmission Organizations (RTO) begintaking over dispatch in select areas areas.
ISO/RTOs consolidate Power Control Areas in to asingle dispatch algorithm.
Generating units are scheduled according to bids inday-ahead, uniform price auctions instead ofengineering estimates.
Real-time auctions determine who actually produceselectricity.
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities in 1999
NPCC
ERCOT
FRCC
SPP
MAPP
WECC
ECARMAIN
MAAC
SERC
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Balancing Authorities Becoming Markets
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Percent of Generation Capacity Dispatched by
Markets.1
.2.3
.4.5
.6S
har
e of
Nat
ional
Cap
acit
y
1998 2000 2002 2004 2006 2008 2010 2012Year
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Percent of Generation Capacity Dispatched by
Markets.1
.2.3
.4.5
.6S
har
e of
Nat
ional
Cap
acit
y
1998 2000 2002 2004 2006 2008 2010 2012Year
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Number of 1999 Power Control Areas Under
Market Dispatch0
20
40
60
Count
1998 2000 2002 2004 2006 2008 2010 2012
Year
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Existing Evidence on Restructuring and Markets
Restructuring papers focus on productivity changes: Fabrizio, Rose, and Wolfram (2007), Bushnell and
Wolfram (2005), Davis and Wolfram (2012), Chan,Fell, Lange, Li (2012), Cicala (2015).
Allocative efficiency losses in markets: Borenstein, Bushnell, and Wolak (2002), Borenstein
(2002), Bushnell, Mansur, and Savaria (2007), Joskowand Kahn (2002), Mansur (2001, 2008), Puller (2007),Ryan (2013), Wolfram (1999), among others.
Allocative efficiency gains of market dispatch versusbilateral trades:
Mansur and White (2012)
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Existing Evidence on Restructuring and Markets
Even the Cato Institute has lost patience withcompetitive reforms in electricity and appears tosee merit in returning to the good old days ofregulated vertically integrated utilities. -Joskow(2006)
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Existing Evidence on Restructuring and Markets
...there has been a significant decline in coalusage at Dairylands facilities due to decreasedoperating hours since joining the MISO.-Dairyland 2011 Annual Report
Steve Cicala Background on Electricity Dispatch February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Merit Order Example0
200
400
600
800
$/M
Wh
0 10000 20000 30000Cumulative Nameplate
NEISO 1aug2005 0:00
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Merit Order Example0
200
400
600
800
$/M
Wh
0 10000 20000 30000Cumulative Nameplate
Installed Capacity Operating
Net Generation: 11783
NEISO 1aug2005 0:00
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Merit Order Example0
200
400
600
800
$/M
Wh
0 10000 20000 30000Cumulative Nameplate
Installed Capacity In Merit Order
Out of Merit
Net Generation: 11783
NEISO 1aug2005 0:00
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
What are Out of Merit Costs? Cost of Production, Npt MW capacity:
Cpt(Qpt) =
Npti=0
cpt(i)qpt(i)
where Qpt =
Npti=0
qpt(i); qpt(i) [0, 1] i
Merit Order Cost of Production:
C pt(Qpt) =
Qpti=0
cpt(i)
Out of Merit Costs:
Cpt(Qpt) C pt(Qpt)
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Losses from Out-of-Merit Generation
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Trade: Fixed-Factor Hecksher-Ohlin Model
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Trade: Fixed-Factor Hecksher-Ohlin Model
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Trade: Fixed-Factor Hecksher-Ohlin Model
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Production Cost Decomposition Total Costs:
p
Cpt (Qpt) =p
[Cpt (Qpt) C pt (Qpt)
]
Within PCAs
+ ...
+p
{C t (Lpt)
[C pt (Lpt) C pt (Qpt)
]}
Across PCAs
Gains from Trade:p
C t (Lpt) C pt (Qpt) + cpt (i = Qpt) [Qpt Lpt ] + ...
p
[cpt (i = Qpt) ct (i = Qt)] [Qpt Lpt ]
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Local Gains from Imports
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Local Gains from Exports
Steve Cicala Welfare Measurement February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Data
Hourly Demand for 1999 footprints (FERC-714,Regional Markets and Authorities)
Merit Order / Supply Curve Generator Nameplate (EIA-860) Monthly Unit-Level Heat Rates using Net Generation
(EIA-767/923) Combined Cycle Heat Rates, Hydro Output (EIA-906) Market Price of Fuel (Daily Regional Gas Hub prices,
FERC/EIA-423/923) Emissions Prices
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Data
Hourly Operational Supply Curve EPA CEMS Gross Load, Emissions Boiler-Generator Crosswalks (EIA-767/860) Combined Cycle Scaling for Missing Cycles (EIA-906) NRC Power Reactor Status Reports for Nuclear USGS Stream flows, USBR, TVA
Power Control Area Configuration as of 1999 (EPAseGRID, FERC Form 714)
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Hourly Operations at Brayton Point0
50
100
150
200
250
MW
h
30dec2004 04:26:40 24apr2005 22:13:20 18aug2005 16:00:00 12dec2005 09:46:40Time
Nameplate Capacity: 243 MW (1963)
Brayton Point Unit 1
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Supply and Demand Data0
100
200
300
400
TW
h
Jan1998 Jan2000 Jan2002 Jan2004 Jan2006 Jan2008 Jan2010 Jan2012MonthYear
Nuclear Hydro
Wind CEMS Hourly
Monthly Load
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Imports and Exports20
25
30
35
40
TW
h /
Month
Jan1998 Jan2000 Jan2002 Jan2004 Jan2006 Jan2008 Jan2010 Jan2012MonthYear
Sum of Hourly PCA Imports Sum of Hourly PCA Exports
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
050
100
150
200
250
$/M
Wh
0 5000 10000 15000 20000 25000Cumulative Nameplate
Out of Merit Cost Gains From Trade
Load Generation
20aug2010 12:00:00
Measuring Welfare from Hourly Operations
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Hourly Operations Animation
Steve Cicala Data February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Estimation Strategy
Difference-in-Differences Design:
ypt = Dpt + p + tr + pt
Treatment on the Treated:
ypt(0) Dpt |Xi
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Potential Confounding with DD Approach
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Fuel Prices Over Time0
510
15
20
25
$/M
MB
TU
1998 2000 2002 2004 2006 2008 2010 2012Year
Gas
01
23
45
$/M
MB
TU
1998 2000 2002 2004 2006 2008 2010 2012Year
Coal
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Policy Function Approach
A Policy maps states to (probabilistic) actions:
py (i ,Xipt ,Dpt) = Pr [qpt(i) = 1|Xipt ,Dpt ]
Counterfactual outcomes based on Behavior predicted by policy under control Observed production costs
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Policy Function Approach: Machine Learning
Random Forest: nonparametric regression withcross-validated tuning parameters
Predict behavior in year t based on policy estimated inyear t 1
Y 0pt = F(q0pt,Xpt
)= F
(0p,y1,Xpt
)+ F
(q0pt,Xpt
) F
(0p,y,Xpt
) Contemporaneous Error
+ ...
+ F(0p,y,Xpt
) F
(0p,y1,Xpt
) Change in Policy Function
= F(0p,y1,Xpt
)+ tr + p + pt
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Policy Function Approach: Machine Learning
DD on the fit of the predicted outcome:
Ypt F (0p,y1,Xpt) = Dpt + tr + p + pt
All predictions are out of sample: How well does last years operations predict this
years operations? Avoids in-sample/out-of-sample changes in fit.
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Relative Perfomance of Random Forest.6
.7.8
.91
RS
S R
elat
ive
to S
imple
OL
S
0 20 40 60 80 100Percentile of Merit Order
Flexible OLS Random Forest
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Intuitive View of Estimation Strategy3
.54
4.5
55
.56
Gen
erat
ion
(G
W)
01apr2004 01oct2004 01apr2005 01oct2005 01apr2006Date
Northern States Power
33
.54
4.5
Gen
erat
ion
(G
W)
01apr2004 01oct2004 01apr2005 01oct2005 01apr2006Date
WAPA ColoradoMissouri
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
A Sampling of Issues / Limitations
SUTVA Measurement Error
No Labor, O&M
Heterogeneous institutions Short-run (infrastructure limitations) Perfectly Inelastic Demand: No Demand-Side Loses
(Gains)
No Pre-period for Markets before 1999 (Early CAISO,PJM).
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
DD by Months to Market:
Log(Gains from Trade)
1
.50
.51
20 10 0 10 20
Ave. Treatment on the Treated
95% Confidence IntervalSteve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Policy Function by Months to Market:
Log(Gains from Trade)
.50
.51
20 10 0 10 20
Ave. Treatment on the Treated
95% Confidence Interval
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Average Effect of Market Dispatch:Log(Gains from Trade)
(1) (2) (3) (4)Market Dispatch 0.312*** 0.350*** 0.258*** 0.191***
(0.045) (0.042) (0.041) (0.048)Log(Load) Yes Yes YesPCA Trend Yes YesPolicy Function YesClusters 16424 16424 16424 15814PCAs 98 98 98 98R2 0.494 0.579 0.617 0.125Obs. 8671235 8671235 8671235 8098935
Note: All specifications include PCA and Region-Date-Hour FixedEffects. Demand controls are PCA-specific. Standard errors clusteredby PCA-Month in parentheses. * p
Imperfect Markets versus Imperfect Regulation
DD by Months to Market:
Log(Out of Merit Costs)
.50
.5
20 10 0 10 20
Ave. Treatment on the Treated
95% Confidence Interval
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Policy Function by Months to Market:
Log(Out of Merit Costs)
.50
.51
20 10 0 10 20
Ave. Treatment on the Treated
95% Confidence Interval
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Average Effect of Market Dispatch:Log(Out of Merit Costs)
(1) (2) (3) (4)Market Dispatch 0.186*** 0.164*** 0.009 0.179***
(0.034) (0.033) (0.033) (0.033)Log(Load) Yes Yes YesPCA Trend Yes YesPolicy Function YesClusters 16450 16450 16450 16444PCAs 98 98 98 98R2 0.775 0.793 0.812 0.302Obs. 11648731 11648731 11648731 11427266
Note: All specifications include PCA and Region-Date-Hour FixedEffects. Demand controls are PCA-specific. Standard errors clusteredby PCA-Month in parentheses. * p
Imperfect Markets versus Imperfect Regulation
Average Effect of Market Dispatch:Log(MWh Traded)
(1) (2) (3) (4)Market Dispatch 0.125*** 0.118*** 0.095*** 0.098***
(0.021) (0.020) (0.021) (0.029)Log(Load) Yes Yes YesPCA Trend Yes YesPolicy Function YesClusters 16464 16464 16464 15910PCAs 98 98 98 98R2 0.557 0.595 0.613 0.099Obs. 12001882 12001882 12001882 11352820
Note: All specifications include PCA and Region-Date-Hour FixedEffects. Demand controls are PCA-specific. Standard errors clusteredby PCA-Month in parentheses. * p
Imperfect Markets versus Imperfect Regulation
Average Effect of Market Dispatch:Log(MWh Out of Merit)
(1) (2) (3) (4)Market Dispatch 0.111*** 0.113*** 0.061*** 0.117***
(0.018) (0.017) (0.017) (0.019)Log(Load) Yes Yes YesPCA Trend Yes YesPolicy Function YesClusters 16448 16448 16448 16430PCAs 98 98 98 98R2 0.841 0.852 0.863 0.157Obs. 11648909 11648909 11648909 11428353
Note: All specifications include PCA and Region-Date-Hour FixedEffects. Demand controls are PCA-specific. Standard errors clusteredby PCA-Month in parentheses. * p
Imperfect Markets versus Imperfect Regulation
Treatment Effects by Day of Year:
Log(Gains from Trade)
1
.50
.51
0 100 200 300 400
95% Confidence Interval Ave. Treatment on the Treated
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Treatment Effects by Day of Year:
Log(Out of Merit Costs)
.8
.6
.4
.20
.2
0 100 200 300 400
95% Confidence Interval Ave. Treatment on the Treated
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Summary of Causal Estimates
Short run estimates show substantial netimprovements in allocative efficiency:
Gains from trade in market areas: 20% ($1B/year) Traded Quantities: 10% Out of Merit Quantities: 10% Out of Merit Costs: 20% ($3B/year)
Increased availability has swamped withholding.
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Results: Discussion
Large for an institutional change Smaller relative to longer-run changes in sector Fuel price variance reduces precision, creates
heterogeneous effects
Steve Cicala Counterfactual Estimates February 12, 2018
Imperfect Markets versus Imperfect Regulation
Outline
1. Background on Electricity Dispatch
2. Welfare Measurement
3. Data
4. Counterfactual Estimates
5. Conclusion
Steve Cicala Conclusion February 12, 2018
Imperfect Markets versus Imperfect Regulation
Conclusion
Electricity markets currently dispatch about 60% ofgenerating capacity
Short run estimates show substantial netimprovements in allocative efficiency
Steve Cicala Conclusion February 12, 2018
Background on Electricity DispatchWelfare MeasurementDataCounterfactual EstimatesConclusion
fd@rm@0:
top related