Evaluation of the 2006
Energy-Smart Pricing Plan SM
Final Report
Prepared for:
CNT Energy
(Formerly known as the Community Energy Cooperative)
2125 W. North Ave.
Chicago, IL 60647
Prepared by:
Summit Blue Consulting
1722 14 th Street, Suite 230
Boulder, Colorado
720-564-1130
November, 2007
C
CONTENTS
Executive Summary ....................................................................................................... i
1 Introduction ................................................................................................................ 1
2 Key Findings............................................................................................................... 2
3 Program Background................................................................................................. 3
4 Estimation of 2006 Program Impacts ....................................................................... 6
5 Comparison of 2006 Results to Previous Years .................................................... 10
6 Results/Conclusions............................................................................................... 11
Summit Blue Consulting
E
EXECUTIVE SUMMARY
The evaluation of the Community Energy Cooperative’s Energy-Smart Pricing PlanSM (ESPP) in
2006, its fourth and final year, found results that supported and bolstered the findings of the
evaluations conducted of the program’s first three years.
Key findings include:
·
·
·
·
·
·
Overall ESPP participants continue to respond to hourly electricity prices in a manner
similar to prior years. They reduce electric consumption during high priced hours.
Participants’ price responses vary depending upon the absolute level of prices, with a price
elasticity of -0.047 when prices are below $0.13/kWh and -0.082 when prices are above
$0.13/kWh. The first price elasticity is comparable to those found in other programs that
use price signals to motivate changes in consumer behavior. The second price elasticity
was not measured in prior ESPP analyses since prices did not significantly exceed that
price level. However, this finding is significant because it suggests that participants
increase their elasticity during times of very high prices, which are the times when
reducing demand on the electric grid will have the most positive impact.
The Energy PriceLight—a small orb that glows different colors based on the current price of
electricity, was effective in assisting customers in responding to hourly prices, with these
customers having an additional 2.4% price elasticity resulting in an elasticity of -0.067.
Automatic cycling of the central-air conditioners (turning the compressor on and off for
short periods of time via remote control) during high-price periods added to a participant’s
response to electricity prices by as much as 9.8%.
ESPP reduced summer energy (kWh) usage by about 27 kWh per month (or 3%). This
indicates that ESPP is also effective at reducing energy usage.
Elasticity of demand increased as the average neighborhood household income declined. In
other words, lower income households were more responsive to price signals than higher
income households.
The implication of this and prior analyses of ESPP is that real-time pricing (RTP) is an effective
tool to manage system electricity demand. As prices increase, participants reduce their electricity
demand. Consequently, at a peak period when the market price for electricity will, by definition,
be high, participants will reduce their demand. This response can have positive impacts on the
operation of the electric system and markets.
Summit Blue Consulting
i
1
INTRODUCTION
This report presents results from an impact evaluation of the Community Energy Cooperative’s
Energy-Smart Pricing Plan SM (ESPP) residential real-time pricing (RTP) program as
implemented during the year 2006. The ESPP program was the first large-scale residential
RTP program in the United States, and 2006 was the fourth and final year of the program. This
evaluation addresses the following key questions:
·
Will residential customers respond to hourly market-based electricity prices?
·
What is the magnitude of the effect, i.e., to what degree is electricity consumption
effected by prices?
·
How do customers respond to high-price notifications?
This report is important given that for the first time in the pilot program there was a significant
price spike during the summer of 2006. On August 1, hourly prices exceeded $0.35/kWh for
several hours (compared to a previous high of just over $0.20/kWh). Otherwise the summer of
2006 was not as hot as the severe summer of 2005, and the number of high price periods was
more like what was seen during the first two years of the pilot program (just sixty hours over
ten days).
As a result of the success of the Energy-Smart Pricing Plan, in 2006 the Illinois General
Assembly authorized the state’s two large investor owned utilities, ComEd and Ameren, to
create new residential real-time pricing programs starting in 2007. These new programs will be
assessed after four years by the Illinois Commerce Commission to determine whether they are
resulting in net economic benefits to the residential customer class. This report (and previous
impact evaluations from 2003 through 2005) can serve as a baseline for the evaluation of new
programs, both in Illinois and elsewhere.
Summit Blue Consulting
1
2
KEY FINDINGS
This evaluation found that:
·
·
·
·
·
·
Overall ESPP participants continue to respond to hourly electricity prices in a manner
similar to prior years. They reduce electric consumption during high priced hours.
Participants’ price responses vary depending upon the absolute level of prices, with a
price elasticity of -0.047 when prices are below $0.13/kWh and -0.082 when prices are
above $0.13/kWh. The first price elasticity is comparable to those found in other
programs that use price signals to motivate changes in consumer behavior. The second
price elasticity was not measured in prior ESPP analyses since prices did not
significantly exceed that price level. However, this finding is significant because it
suggests that participants increase their elasticity during times of very high prices,
which are the times when reducing demand on the electric grid will have the most
positive impact.
The Energy PriceLight—a small orb that glows different colors based on the current price
of electricity, was effective in assisting customers in responding to hourly prices, with
these customers having an additional 2.4% price elasticity, resulting in an elasticity of 0.067.
Automatic cycling of the central-air conditioners (turning the compressor on and off for
short periods of time via remote control) during high-price periods added to a
participant’s response to electricity prices by as much as 9.8%.
ESPP reduced summer energy (kWh) usage by about 27 kWh per month (or 3%). This
indicates that ESPP is also effective at reducing energy usage.
Elasticity of demand increased as the average neighborhood household income
declined. In other words, lower income households were more responsive to price
signals than higher income households.
The implication of this and prior analyses of ESPP is that real-time pricing (RTP) is an
effective tool to manage system electricity demand. As prices increase, participants reduce
their electricity demand. Consequently, at a peak period when the market price for electricity
will, by definition, be high, participants will reduce their demand. This response can have
positive impacts on the operation of the electric system and markets.
Summit Blue Consulting
2
3
PROGRAM B ACKGROUND
The Community Energy Cooperative (Cooperative) is a nonprofit organization committed to
offering its members programs and services to help manage energy costs and improve
neighborhood electrical reliability. It was founded by the Chicago-based nonprofit organization
the Center for Neighborhood Technology in 2000. Starting in 2002, the Cooperative
collaborated with Commonwealth Edison (ComEd) and the Illinois Department of Commerce
and Economic Opportunity (DCEO) to test residential customers’ responses to day-ahead,
market-based hourly pricing (also known as real-time pricing) through a three-year pilot
program (later extended for a fourth year). The program was offered to the Cooperative’s new
and existing members through an experimental ComEd tariff. DCEO provided partial funding
for the interval meters, programmable thermostats, and the initial evaluation reports. In 2007
the Cooperative changed its name to CNT Energy and their ongoing work on residential realtime pricing is conducted under that name.
This program, the Energy-Smart Pricing Plan, was started in January 2003, and it was the first
time in the nation that residential customers were given the opportunity to pay market-based
electricity prices. By exposing residential customers to the market price of electricity,
customers were given the opportunity to make informed decisions about electricity use. This
program used ComEd’s Rate RHEP - Residential Hourly Energy Pricing (Experimental) (Rate
RHEP). The program’s pricing was initially set using prices as provided daily from Platts
Global Energy for ComEd translated into day-ahead hourly prices by using PJM West load
shapes. In June 2005, the prices were changed to the PJM ComEd Zone Day-Ahead Hourly
LMP’s, which became available due to ComEd joining the PJM Interconnection, LLC. In
addition to the hourly electricity price, participants’ bills contained an access charge that was
adjusted yearly, but was capped at 4.881 and 5.367 cents per kWh for single-family and multifamily homes, respectively. The June 2005 through May 2006 access charges were 3.307 and
3.734 cents per kWh for single-family and multi-family homes, respectively. Program
participants’ rates were reduced by the Program Incentive of 1.4 cents per kWh that was
designed to encourage participation in the program while also potentially reflecting changes in
post-transition rates.
In 2003, more than 750 members were enrolled in the program. To aid in the first impact
evaluation of ESPP, the Cooperative randomly assigned ESPP enrollees into two groups:
participants (651 members), who were exposed to the hourly market prices, and a control group
(103 members), whose members had an interval meter installed, but did not receive any of the
ESPP educational information and continued to pay a flat rate for their electricity. In 2004, the
control group participants were transferred into the ESPP group and began paying for their
electricity using Rate RHEP, and additional customers enrolled in the program. This resulted
Summit Blue Consulting
3
in an enrollment of more than 1,000 members in the ESPP program. In 2005, additional
applicants were allowed to enroll in the program until the target participation population of
1,500 was reached. No further participants were allowed in to enroll in 2006.
A new feature of ESPP during 2004 and continuing into 2005 and 2006 was the addition of the
installation of cycling switches on the central air conditioners of 57 participants. These
switches were set to cycle the air conditioner 50% of the time during high-price periods. This
automatic cycling was an experiment to see how this type of direct intervention compares to
giving participants their choice of whether or not to use an air conditioner during the high-price
periods.
In 2006 the Cooperative added an additional experimental tool, the PriceLight. The PriceLight
is a small globe that uses a pager signal to activate LED lights that change color to reflect the
price of electricity for that hour.
One of the features of ESPP was the day-ahead notification of high price days, when the price
of electricity would be over thirteen cents/kWh1. This notification was either by telephone or email. The purpose of this notification was to provide a mechanism for participants to become
aware of relatively expensive prices and to adjust their consumption accordingly without
relying upon expensive technological monitoring systems. Other aspects of ESPP included:
·
·
·
Day-ahead prices. Prices for the next day were posted on the website or were available
by phone after 5 p.m.
Price protection cap. The Cooperative included a price limit hedge at $0.50 per kWh,
meaning that no customer participating in ESPP would ever see a net hourly price
greater than $0.50 per kWh. (The maximum actual price ever seen was $0.3655/kWh.)
Energy management and price response information. The Cooperative provided:
- Information about usage.
- Instructions and tips on how to reduce usage overall and during peak periods.
During the first year of ESPP (2003) weather conditions were relatively mild, and hourly
electricity prices rose to the high-price alert level (above 10 cents per kilowatt-hour) on only 20
days, and usually for only three or four hours each day. The summer of 2004 was even milder
and there were only 19 hours, spread over seven days, when prices triggered the high-price
alerts.
The weather conditions changed dramatically in 2005. June and July of 2005 were ranked as
the sixth warmest of all comparable months on record since 1871. This extreme heat resulted
in record peak electricity demands. In addition to the extreme weather, the prices for natural
gas, an input to the production of electricity, were also high, reflecting world events and
concern over the potential effects of hurricanes on supplies of natural gas. Higher gas prices
translate into higher electricity prices, independent of weather. Taken together, they provided a
set of market conditions that resulted in relatively high hourly electricity prices throughout the
1
Prior to 2006 the notification occurred when the price exceeded ten cents/kWh.
Summit Blue Consulting
4
summer. During the summer of 2005, there were 57 days from June through August when
prices exceeded $0.10/kWh (commodity portion only), with prices staying over this level for as
much as 13 hours per day (for a total of over 360 hours of high prices). The highest price was
just under 19 cents per kilowatt-hour for one hour. In 2006, the high-price notification was
changed to $0.13/kWh, and there were 63 hours of high prices over 10 days.
In 2006, the high-price notification point was changed to $0.13/kWh, and there were 63 hours
of high prices over 10 days. During the summer of 2006, there where thirteen hours (all of
them occurring on August 1 and August 2) when the price was above $0.20/kWh. The
maximum price was $0.356/kWh during the hour of 4 p.m. to 5 p.m. on August 1.
Exhibit 3-1: ESPP Hourly Electricity Prices, Summer 2006
2006 ESPP Prices
0.4
0.35
0.3
$/kWh
0.25
0.2
Prices
0.15
0.1
0.05
08/31/06
08/24/06
08/17/06
08/10/06
08/03/06
07/27/06
07/20/06
07/13/06
07/06/06
06/29/06
06/22/06
06/15/06
06/08/06
06/01/06
0
Date
The summer of 2006 represented an opportunity to examine how residential customers respond
to hourly changes in electricity prices, including relatively high electricity prices. The results
of the estimation of ESPP participants’ response to hourly electricity prices during the summer
2006 are presented below.
Summit Blue Consulting
5
4
Estimation of 2006 Program Impact s
This section presents the results of the impact evaluation of the fourth summer of the EnergySmart Pricing Plan program.
The specific issues that were addressed in this evaluation include the following:
·
What is participants’ price elasticity, and how do they respond to high price
notifications?
·
Have differences in the responses of participants in the Central Air Conditioner Cycling
Option occurred, particularly when they have been controlled for very long periods of
time?
·
Does the Energy PriceLight—a small orb that glows different colors based on the current
price of electricity, have an effect on energy consumption (in addition to the above
demand impacts)?
As in the prior evaluations of these programs, the impact evaluation relied upon hourly
consumption data for participants, weather data, and hourly electricity prices to capture, ex
post, the response to prices. The details on how these data are combined within a regression
model to understand participants’ behavior are presented below.
HOURLY PRICE RESPONSE
The fundamental approach that this (and the previous) analysis used to determine how
participants respond to hourly electricity prices is the estimation of the price elasticity of
demand for electricity. Price elasticity is defined as the percentage change in demand
associated with a percentage change in price. For example, a price elasticity of -10% implies
that a 100% increase in price will reduce demand by 10%.
While the concept of price elasticity is straightforward, estimation of it is not. The problem is
that there are other things changing hourly beside the price of electricity, which can
significantly affect the demand for electricity. While there are many such things occurring for
any given household (such as people coming and going), the only variables that can readily be
measured are those relating to weather. In order to control for the effects of these non-price
variables, this evaluation used a regression model that statistically models hourly energy use as
a function of the hourly price of electricity and the hourly weather conditions.
There are many ways in which such a regression model can be specified. In the previous
analyses, a pooled model was used, which combines data across households (i.e., crosssectional) with data over time (i.e., time-series) into a single regression equation. Any
Summit Blue Consulting
6
differences across homes that do not vary over time were captured through house-specific
intercept terms, as discussed below.
The fixed effects model can be viewed as a type of differencing model in which all
characteristics of the home, which (1) are independent of time and (2) determine the level of
hourly electricity use, are captured within the house-specific constant terms. In other words,
differences in housing characteristics that cause variation in the level of energy consumption,
such as building size and structure, are captured by constant terms representing each unique
household.
Algebraically, the fixed-effect panel data model is described as follows:
ln( y ) = a + r ln( price ) + b x + e
it
i
t
it
it
,
(Eq. 1)
where:
ln(yit)
=
The natural log of electricity consumption for home i during hour
t
ai
=
the estimated constant term for household i
r
=
the price elasticity of electricity demand
ln(pricet)
=
The natural log of the price of electricity during hour t
ß
=
vector of estimated coefficients
xit
=
e it
=
vector of variables that represent factors causing changes in the
electricity consumption for home i during hour t (i.e., weather)
error term for home i during hour t.
This hourly demand model was estimated over all ESPP participants during the month of June
through August. Since there was little price variation in early morning hours, the estimation
included hours after 9 a.m. Thus, the model used 1,977,927 hourly observations from
1,134 participants.2 The estimated model for 2006 is presented in Exhibit 4-1. The resulting
price elasticity estimate for prices below $0.13/kWh is -4.7% (with a t-value of -21.2,
suggesting that this response is statistically significant at the 95% confidence level). Thus, a
100% increase in the hourly price of electricity would produce, on average, a 4.7% decrease in
hourly energy use across all participants. This result was very robust across different
specifications and different subsets of participants.3
2
3
This is fewer than the total number of participants due to missing data from sources such as estimated bills.
This model also incorporates corrections for heteroskedasticity and autocorrelation.
Summit Blue Consulting
7
Exhibit 4-1: Elasticity Model.
Dependent variable is log of hourly kWh. (Summer 2006)
Independent Variable
Coefficient
t-value
Natural log of hourly price when
price <= $0.13/kWh
-0.047
-21.2
Natural log of hourly price when
price > $0.13/kWh
-0.082
-30.7
Natural log of hourly price and
customer has a PriceLight
-0.067
-13.9
Customer has A/C cycling and it is a
high price hour
-0.098
-9.7
0.10
14.9
Temperature lagged one hour
Sample Size
R-squared
Partial (ignoring customer effects)
Full
1,977,927 (1,134 homes)
1%
45%
This estimated elasticity is similar to the elasticity found in the 2003 impact evaluation 4.2% with a t-value of 12.6. Thus, residential customers clearly respond to hourly changes in
electricity prices, and this response is consistent over time. The unique result of this analysis is
that for prices above $0.13/kWh (the high-price notification point), the elasticity increases (in
absolute terms) to 8.2%, indicating that customers become more responsive when they are
given the high price notification. 4 The next section investigates whether this price
responsiveness lead to any energy savings.
ENERGY EFFECTS
The analysis presented above concentrated only on the response to hourly electricity price
changes. Thus, the impacts are only demand (average kW). Another aspect of electricity usage
is energy (kWh). So while it is clear that the ESPP reduces demand when prices are high, the
next question that will be addressed is how the ESPP program affects energy usage, more
specifically monthly kWh usage.
In order to determine the effect of this program on energy (kWh) consumption, a regressionbased model was estimated that used monthly electricity consumption for all Cooperative
members (thus all ESPP participants and a large sample of non-participants) going back to
2003 (so as to include consumption for both before and after participation in the program) as
the dependent variable. The independent variables include the monthly weather conditions and
two program effect variables: one variable that is equal to 1 for every month the person has
been a participant in ESPP, and a second variable that is equal to 1 if the person is a participant
4
We were unable to address price elasticity changes in the past under high price notifications because there
were not sufficient variations in prices above the high-price point. This is the first year where prices
exceeded the high price point by such a large degree.
Summit Blue Consulting
8
and the month is June, July, or August.5 A fixed-effects specification for the regression model
was used, which pooled across customers and over time. The model was also corrected for
autocorrelation and heteroskedasticity. The results are presented in Exhibit 4-2.
Period
Participating in ESPP
Participating in ESPP
Summer Months
Sample Size
R-Squared
Exhibit 4-2: ESPP Energy Impacts
(Dependent variable, monthly energy use, 2003 to 2006)
Estimated Coefficient
T-Value
-16.7
-14.5
-10.0
-9.3
1,253,593 (7,567 customers)
1.00
The estimation results show that ESPP participants consumed -26.7 kWh less per month during
the summer months relative to individuals not on the ESPP rate (-16.7 kWh year round plus 10.0 kWh specific to the summer months). This represents a savings of around 3% of summer
electricity usage. This result is statistically significantly different from 0 at the 95% confidence
level. Therefore, ESPP results in a net decrease in energy consumption.
THE IMPACT OF INCOME ON PRICE ELASTICITY
An ongoing issue in the assessment of the potential of real-time pricing for residential
customers is how it will impact different sub-sectors. An analysis of the ESPP data using
neighborhood income as an added variable found that, in fact, elasticity of demand varied by
neighborhood. Elasticity decreased as income increased. While the amount was small, it shows
that lower income individuals are more price sensitive in their energy use than higher income
households.
Because income data can be hard to collect and verify, this analysis used census data at the zip
code level to assign ESPP participants into various income ranges. Due to the high degree of
residential segregation in the Chicago area, this method was considered an acceptable proxy for
actual individual income levels.
While this preliminary result is very encouraging in regard to the potential value of real-time
pricing for low-income households, more research on this issue is needed, and additional
research goals and objectives should be defined.
5
The model also includes monthly indicator variables to capture seasonal effects. For clarity, these
coefficients, as well as the coefficients on the weather and annual variables are not included in the table.
Summit Blue Consulting
9
5
Comparison of 2006 Results to
Previous Years
The results of the impact evaluation of the Energy-Smart Pricing Plan in 2006 show levels of
response that are consistent with what was found in previous years. The following table is a
summary of results from each year.
Year
Exhibit 5-1: Annual Elasticities, 2003 - 2006
Overall elasticity
Other key elasticities
2003
-0.042
2004
-0.080
2005
-0.047
-0.067 for air conditioner cycling
2006
-0.047 when prices below $0.13/kWh
-0.098 for air conditioning cycling
-0.082 when prices above $0.13/kWh
-0.067 for PriceLight
While not part of this impact evaluation, the Community Energy Cooperative modeled the
overall impact of participation in this pilot program on the bills of participants by comparing
what they paid to what they would have paid with the same energy use billed at the traditional
flat rate that all other ComEd residential customers paid. The following table is a summary of
the savings/losses of participants.
Exhibit 5-2: Annual Bill Impacts and Weather
Average
Average
Year
Monthly
Bill
kWh
Savings
Average
Energy Price
(¢/kWh)
Maximum
Summer Price
(¢/kWh)
Cooling Degree
Summer Days with
Days
High Price
Notifications
(Avg. is 799)
2003
$51.10
630
20.1%
3.2
12.4
659
9
2004
$56.99
648
11.3%
3.8
12.5
574
7
2005
$77.82
758
-6.3%
5.7
19.1
1,087
57
2006
$56.32
679
15.5%
5.1
36.6
937
10
Summit Blue Consulting
10
6
Results/Conclusions
The results of the fourth year of this program continue to be generally positive overall. During
a relatively warm summer with periods of high electricity prices, participants continue to have
a strong response to hourly electricity prices. Two price elasticity estimates were calculated for
the 2006 evaluation. The first price elasticity estimated was for prices below the high-price
notification level (i.e., below $0.13/kWh). This “low-price” elasticity was estimated as -4.7%
(t-value of -21.2). This compares favorably to the elasticity estimate found in the evaluation of
ESPP 2003 (-4.2% with a t-value of 12.6). For prices above the high-price notification, the
elasticity was estimated as -8.2% (t-value of -30.7) indicating that customers become more
price responsive with the high-price notification. This model also indicated that:
·
Cycling the air conditioners during high-price periods increases the elasticity, adding
9.8% on high-price days. This is consistent with the expectations for an automated
control system.
·
The PriceLight increases (in absolute terms) participants elasticity by 6.7%.
·
The high-price notifications resulted in an increase in price elasticity from -4.7% to
-8.2%. Thus, the high price notification nearly doubles customers’ responsiveness to
electricity prices.
·
ESPP reduces summer energy (kWh) usage by about 27 kWh per month (or 3%). This
indicates that ESPP is also effective at reducing energy usage.
·
Elasticity of demand increased as the average neighborhood household income
declined. In other words, lower income households were more responsive to price
signals than higher income households.
These results demonstrate that residential customers do indeed respond to hourly electricity
prices, and overall this response is consistent across years, despite great variation in weather
and energy prices. This analysis has also shown that the degree to which residential customers
respond to prices varies by the time of day, as well as by several observable characteristics of
the customer. One of the key aspects of ESPP is the personal notification (via phone or e-mail)
sent to participants when the price the next day is expected to be greater than or equal to 13
cents per kWh. The hourly price elasticity model described above revealed that these highprice notifications now result in an increase in price elasticity. This increase was not addressed
in the past because there was not sufficient price variation above the high price point to develop
a price elasticity estimate.
Summit Blue Consulting
11
As was observed in the previous evaluations, the experience of ESPP shows that real-time
pricing for residential customers can be an effective approach to establishing a demand side
management program. The ESPP program, with its low tech design and correspondingly lower
implementation costs, succeeds in enabling participants to understand and respond to the price
patterns without placing an undue burden on them. Even under conditions with consistent
periods of high prices, customers are able to respond to prices.
Summit Blue Consulting
12