Post on 17-Mar-2020
Does Increasing Block Pricing Decrease
Energy Use?
Evidence from the Residential Electricity Market
Becka Brolinson
Georgetown University
November 4, 2019
Introduction
Introduction
RQs: Do IBPs decrease total electricity use relative to a flat price? Do
IBPs help low-income households?
In This Paper
Four tasks:
1. Estimate price elasticities of demand by income
2. Calculate a flat price that would raise the same revenue as the
current IBP
3. Compare aggregate energy use under the flat schedule and the IBP
schedule
4. Compare electricity bills by income under both pricing schedules
In This Paper
Four tasks:
1. Estimate price elasticities of demand by income
2. Calculate a flat price that would raise the same revenue as the
current IBP
3. Compare aggregate energy use under the flat schedule and the IBP
schedule
4. Compare electricity bills by income under both pricing schedules
In This Paper
Four tasks:
1. Estimate price elasticities of demand by income
2. Calculate a flat price that would raise the same revenue as the
current IBP
3. Compare aggregate energy use under the flat schedule and the IBP
schedule
4. Compare electricity bills by income under both pricing schedules
In This Paper
Four tasks:
1. Estimate price elasticities of demand by income
2. Calculate a flat price that would raise the same revenue as the
current IBP
3. Compare aggregate energy use under the flat schedule and the IBP
schedule
4. Compare electricity bills by income under both pricing schedules
The Data
Residential Appliance Saturation Study
• 11,745 single-family households in CA in 2003 and 2009
• Household characteristics
• Monthly electricity consumption, 191,851 household-month pairs
Historical rates for CA utilities
• Rates for 2 CA utilities from 2000-2009 PGE & SDGE Rates
The Data
Residential Appliance Saturation Study
• 11,745 single-family households in CA in 2003 and 2009
• Household characteristics
• Monthly electricity consumption, 191,851 household-month pairs
Historical rates for CA utilities
• Rates for 2 CA utilities from 2000-2009 PGE & SDGE Rates
Empirical Challenges
1. Simultaneity
2. Price is a function of quantity:
Identification Strategy
Two Sources of Variation:
1. CA climate zones
2. Changes in price
Two Steps:
1. Compare households across neighboring climate zone borders Maps
2. Simulated instrument for the change in price
Identification Strategy
Two Sources of Variation:
1. CA climate zones
2. Changes in price
Two Steps:
1. Compare households across neighboring climate zone borders Maps
2. Simulated instrument for the change in price
CA Climate Zones
Utility Induced Price Changes
Estimating Equation: 2SLS
Simulated Instrument:
∆ln(kWhit) = β0 + δ∆̂ln(Pit) +10∑j=1
Ditj + β1∆Xit + γi + ηit (1)
• ∆ln(kWhit) = ln(kWhit) − ln(kWhi0)
• ∆ln(P̃it) = ln(Pt(kWhit0 )︸ ︷︷ ︸Price in t ifkWht = kWh0
) − ln( P0(kWhit0 )︸ ︷︷ ︸Price in period 0
) First Stage
• Ditj - Dummy variable equal to 1 if energy use is in decile j , 0
otherwise
• Xit - Weather controls
• γi - border-zone fixed effect
• Similar instrument used Koichiro Ito (AER, 2014)
• Alternative Instruments: Middle Period Consumption & Mean Consumption
Price Elasticities
Income Group Avg Price Elasticity N
Full Sample -0.016 27,144
$0–$49,999 -0.100 7,801
$50,000–$74,999 -0.132 5,913
$75,000–$149,999 -0.165 9,695
>$150,000 -0.427 3,735
More Details: Zone by Zone kWh by Income Main Strategy: Method Comparison
Middle Period: Method Comparison Mean Consumption: Method Comparison
Price Elasticities
Income Group Avg Price Elasticity N
Full Sample -0.016 27,144
$0–$49,999 -0.100 7,801
$50,000–$74,999 -0.132 5,913
$75,000–$149,999 -0.165 9,695
>$150,000 -0.427 3,735
More Details: Zone by Zone kWh by Income Main Strategy: Method Comparison
Middle Period: Method Comparison Mean Consumption: Method Comparison
Finding the Flat Price
Holding R̄ constant:
R̄ =N∑i
(Di (p̄)) ∗ p̄
=N∑i
[Di (pi ) + (p̄ − pi )εiD(pi )
pi]p̄
(2)
• R̄ Revenue
• Di (p) Demand
• p̄ Flat Rate
• εi Price Elasticity for Income Group i
Example
Example
Example
Aggregate Demand
Year Average Marginal
2003 0.01% -3.43%
2009 0.86% -4.91%
Weighted Average 0.38% -4.12%
Aggregate Demand
Year Average Marginal
2003 0.01% -3.43%
2009 0.86% -4.91%
Weighted Average 0.38% -4.12%
Using Other Elasticities
Income and Electricity Use
Consumer Surplus
$
kWhD
CS
p̄
q̄i
Consumer Surplus
$
kWhD
p1
p2
qi
p̄
q̄i
∆ Use from flat to block
CB
A
Consumer Surplus
Income Group Change in CS ($) Pct. Change N
$0–$49,999 1.22 2.64% 39,540
$50,000–$74,999 4.83 7.48% 60,446
$75,000–$149,999 3.25 4.15% 63,343
>$150,000 -2.02 -1.98% 26,771
Conclusion
• First estimates of price elasticity of energy demand by income group
under average price response assumption
• Price elasticities increase with income from -0.100 to -0.427
• Demand increases by 0.38% for average price response
• Evidence IBP may not be meeting goal of decreasing aggregate
consumption while protecting low income households from increasing
energy prices
Historical Rates
Data
Geography
• Limit the sample to households who live in a zip-code whose
centroid is within 10 km to the nearest climate zone border
Geography
• Limit the sample to households who live in a zip-code whose
centroid is within 10 km to the nearest climate zone border
Geography
• Limit the sample to households who live in a zip-code whose
centroid is within 10 km to the nearest climate zone border
• Generate the same simulated instrument and run the main
estimating equation including border fixed effects
Geography
• Limit the sample to households who live in a zip-code whose
centroid is within 10 km to the nearest climate zone border
• Generate the same simulated instrument and run the main
estimating equation including border fixed effects
Income Group Avg Price Elasticity N
Full Sample -0.100 112835
$0-34,999 -0.043 21466
$35,000- 74,999 -0.0956 35205
$75,000-149,999 -0.0977 39697
$>150,000 -0.114 16467
Geography
• Limit the sample to households who live in a zip-code whose
centroid is within 10 km to the nearest climate zone border
• Generate the same simulated instrument and run the main
estimating equation including border fixed effects
Year Average Marginal
2003 0.17% -0.40%
2009 0.52% -0.69%
Weighted Average 0.33% -0.53%
Back to Empirical Strategy
First Stage Regression
∆ln(Pit) = π0 + π1∆ln(P̃it) + Djt + βXit + γi + τt + ηit (3)
• ∆ln(kWhit) = ln(kWhit) − ln(kWhi0)
• ∆ln(P̃it) = ln(Pt(kWhit0 )︸ ︷︷ ︸Price in t ifkWht = kWh0
) − ln( P0(kWhit0 )︸ ︷︷ ︸Price in period 0
)
First Stage Regression
∆ln(Pit) = π0 + π1∆ln(P̃it) + Djt + βXit + γi + τt + ηit (3)
Dependent Variable is: ∆ln(Pit)
T/X X/R R/S S/X
∆ln(P̃it) 0.836*** 0.681*** 0.507*** 0.749***
s.e. (0.024) (0.016) (0.020) (0.014)
Households 1,719 2,356 1,968 2,766
Observations 27,656 38,124 31,829 44,809
F-Stat 378.33 365.3 303.36 618.13
R-Squared 0.8517 0.7973 0.7742 0.8271
Standard errors in parentheses *p<0.05, **p<0.01,
***p<0.001, s.e. clustered at the household level
Back to Main Strategy
Avg. Price Response Elasticities
Dependent Variable is: ∆ln(kWhit)
Climate Border: T/X X/R R/S S/X
Full Sample -0.144*** -0.248*** -0.300*** -0.203***
(0.0415) (0.0256) (0.0402) (0.0233)
$0-34,999 0.120 -0.110 0.0824 0.0388
(0.0680) (0.103) (0.103) (0.0610)
$35,000- 74,999 -0.265*** -0.253*** -0.315*** -0.218***
(0.0629) (0.0379) (0.0543) (0.0367)
$75,000-149,999 -0.0378 -0.157*** -0.278*** -0.164***
(0.0850) (0.0338) (0.0628) (0.0358)
$>150,000 -0.184** -0.312*** -0.0885 -0.247***
(0.0560) (0.0511) (0.239) (0.0551)
Standard errors in parentheses *p<0.05, **p<0.01,
***p<0.001, standard errors clustered at hh-level
Elasticities
Method Comparison
Income First Diff Sim. Inst. Sim. Inst.
+ Match
Sim Inst +
Geo
Full Sample -0.101*** -0.133*** -0.201 -0.100***
(0.00749) (0.0105) (0.0127)
$0-34,999 -0.0183 -0.0291 0.006 -0.0430
(0.0234) (0.0379) (0.0517)
$35,000- 74,999 -0.101*** -0.135*** -0.230 -0.0956***
(0.0126) (0.0179) (0.0218)
$75,000-149,999 -0.100*** -0.124*** -0.144 -0.0977***
(0.0126) (0.0161) (0.0208)
$>150,000 -0.106*** -0.123*** -0.205 -0.114***
(0.0161) (0.0198) (0.0242)
Standard errors in parentheses *p<0.05, **p<0.01, ***p<0.001,
s.e. clustered at the household leve
Elasticities
Other Elasticities
Brolinson R&W Ito
Year Average Marginal Average Marginal Average Marginal
2003 -0.28% -0.77% 3.25% -4.74% 0.51% -1.19%
2009 0.23% -1.29% 8.14% -5.05% 1.07% -1.50%
Avg. -0.05% -1.01% 5.51% -4.89% 0.77% -1.33%
Back to My Results
kWh Distribution by Income
Elasticities
Appendix A: Ito Simulated Instrument
Simulated Instrument:
∆ln(kWhit) = α0 + δ∆ln(P̃it) + Djt + βXit + γi + τt + εit (4)
• ∆ln(kWhit) = ln(kWhit) − ln(kWhit−12)
• ∆ln(P̃it) = ln(Pt(kWhit−6)︸ ︷︷ ︸Price in t if
kWht = kWht−6
) − ln( P0(kWhit−6)︸ ︷︷ ︸Price in period 0with use kWhit−6
)
• Djt - Dummy variable equal to 1 if energy use is in decile j , 0
otherwise
Back to Main Strategy
Appendix A: Ito Simulated Instrument Results
Income First Difference Sim. Inst. Sim. Inst.
+ Matching
Sim. Inst.
+ Geo
Full Sample 0.829*** -0.209*** -0.302 -0.151*
(0.0685) (0.0563) (0.0714)
$0-34,999 -0.161 -0.228 -0.371 0.00950
(0.331) (0.253) (0.278)
$35,000- 74,999 0.814*** -0.126 -0.228 -0.182
(0.121) (0.115) (0.158)
$75,000-149,999 0.981*** -0.106 -0.189 -0.0668
(0.0909) (0.0915) (0.114)
$>150,000 0.970*** -0.289* -0.241 -0.0831
(0.154) (0.146) (0.166)
Standard errors in parentheses *p<0.05, **p<0.01, ***p<0.001,
standard errors clustered at the household level to adjust for serial
correlation Elasticities
Appendix B: Mean Cons. Simulated Instrument
Simulated Instrument:
∆ln(kWhit) = α0 + δ∆ln(P̃it) + Dtj + βXit + γi + τt + εit (5)
• ∆ln(kWhit) = ln(kWhit) − ln(kWhi0)
• ∆ln(P̃it) = ln( Pt(kWh)︸ ︷︷ ︸Price in t ifkWht = kWh
) − ln( P0(kWh)︸ ︷︷ ︸Price in period 0
with use kWh
)
• Djt - Dummy variable equal to 1 if energy use is in decile j , 0
otherwise
Back to Main Strategy
Appendix B: Mean Cons. Simulated Instrument Results
Income First Difference Sim. Inst. Sim. Inst.
+ Matching
Sim. Inst.
+ Geo.
Full Sample 0.829*** -0.442*** -0.522 -0.336***
(0.0685) (0.0394) (0.0568)
$0-34,999 -0.161 -1.142*** -1.371 -1.036**
(0.331) (0.246) (0.378)
$35,000- 74,999 0.814*** -0.486*** -0.558 -0.306**
(0.121) (0.0700) (0.112)
$75,000-149,999 0.981*** -0.399*** -0.458 -0.263**
(0.0909) (0.0605) (0.0827)
$>150,000 0.970*** -0.441*** -0.512 -0.425***
(0.154) (0.0825) (0.110)
Standard errors in parentheses *p<0.05, **p<0.01, ***p<0.001,
standard errors clustered at the household level to adjust for serial
correlation Elasticities