how do consumers respond to water and electricity pricing? · city limits laguna woods san diego...
TRANSCRIPT
How Do Consumers Respond to Water and Electricity Pricing?
Evidence from Recent Empirical Studies in Economics
Koichiro ItoSIEPR Postdoctoral Fellow
Stanford [email protected]
Assistant Professor, Boston University (from July 2013)
Two Common Assumptions in Demand Estimation and Forecasting
1) Consumers fully understand their price schedule
2) Economic theory tells us that consumers respond to marginal price
2Example: Nonlinear Residential Water Pricing in Irvine Ranch Water District in CA
Figure 3: Five-Tier Increasing Block Price Schedule in IRWD in August 2002
0
1
2
3
4
5
6
Mar
gina
l pric
e ($
) per
CC
F
0 40 100 150 200 250 300Monthly consumption relative to baseline allocation (%)
Notes: This figure shows the five-tier increasing block price schedule in IRWD in August 2002. IRWD
allocates“baseline allocation”to a customer, and the customer’s marginal price depends on consumption
relative to the baseline allocation. The marginal price equals the first tier rate up to 40% of the baseline,
the second tier rate up to 100%, the third tier rate up to 150%, the fourth tier rate up to 200%, and
the fifth tier rate over 200% of the baseline.
26
Potential Problems about the Two Assumptions
3
Consumers Are Not Well Informed about their Water or Electricity Prices
4
Consumers Are Not Well Informed about their Water or Electricity Prices
Easy to see what gas price you are paying
5
Consumers Are Not Well Informed about their Water or Electricity Prices
Easy to see what gas price you are paying
6
Hard to see what water price you are paying
Typical Utility Bills: Difficult to Understand
7
Example: An Electricity Bill in California
5
Typical electricity bills are unclear and difficultto get the information about prices
MOTIVATION 3)
Ɣ Electricity bills – Hard to get the information about prices
Typical Utility Bills: Difficult to Understand
8
Example: A Water Bill in California
Given this environment,Do Consumers Respond to Correct
Marginal Price?
9
I Examine How Consumers Actually Respond to Water and Electricity Prices
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Partnered with Water Utility (Irvine Ranch Water District) in CA
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Partnered with Electricity Utilities in CA
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Household-level monthly billing data for about 10 years
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Electricity consumers: 40,749 households
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Water consumers: 64,601 households
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Quasi-experimental research design
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Use policy changes as natural experiments
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Exploit spatial discontinuities to create “treatment” and “control” groups
10
In Orange County CA, Households in the Same City Have Different Power Companies
11
Nonlinear electricity pricing Randomized Field Experiments on Dynamic Electricity Pricing
I exploit a nearly ideal research environment in electricity markets in California
Edison (Southern California Edison) provides electricity for the north side
LagunaBeach
AlisoViejo
LagunaNiguel
Laguna Hills
MissionViejo
Cote deCoza
LasFlores
RanchoSanta
Margarita
Border of Electric Utility Service Areas
City Limits
Laguna Woods
San Diego (San Diego Gas & Electric) provides electricity for the south side3 / 8
They Experience Very Different Pricing
12
Nonlinear electricity pricing Randomized Field Experiments on Dynamic Electricity Pricing
Households experience substantially different nonlinear pricing
Edison and San Diego: Cents per kWh in 2002
San Diego
Edison
10
15
20
25
Monthly Consumption
4 / 8
Average Price = Total Bill Payment / Total Usage
Similarly, Exploit Policy Changes in Irvine Ranch Water District (IRWD)
13
Figure 2: Service Areas of Irvine Ranch Water District (IRWD)
Notes: This figure shows service ares in Irvine Ranch Water District (IRWD). The District serves the
City of Irvine and portions of the Cities of Costa Mesa, Lake Forest, Newport Beach, Tustin, Santa
Ana, Orange and unincorporated Orange County. Santa Ana Heights service area was consolidated in
1997 and Lake Forest service are was consolidated in 2001. The original map is provided by IRWD.
25
Residential Water Pricing Changed from Flat Pricing to Nonlinear Pricing
14
Figure 3: Five-Tier Increasing Block Price Schedule in IRWD in August 2002
0
1
2
3
4
5
6
Mar
gina
l pric
e ($
) per
CC
F
0 40 100 150 200 250 300Monthly consumption relative to baseline allocation (%)
Notes: This figure shows the five-tier increasing block price schedule in IRWD in August 2002. IRWD
allocates“baseline allocation”to a customer, and the customer’s marginal price depends on consumption
relative to the baseline allocation. The marginal price equals the first tier rate up to 40% of the baseline,
the second tier rate up to 100%, the third tier rate up to 150%, the fourth tier rate up to 200%, and
the fifth tier rate over 200% of the baseline.
26
(1) Before Policy Change (Flat)
(2) After Policy Change (Nonlinear)
What Do I Find?
15
Findings: Both Water and Electricity Consumers Respond to Average Price (not Marginal Price)
16
Introduction Research Design Estimation Welfare Conclusion
Research question: How do consumers respond to nonlinear price schedules?
1 Standard economic theory predicts:
Consumers respond to Marginal Price
2 Laboratory experiments find:
Many individuals respond to Average Price = ( Total payment / Quantity )
Demand Curve
MP
AP
10
15
20
25
Ma
rgin
al P
rice
(ce
nts
pe
r kW
h)
Monthly Consumption
) Standard theory and laboratory evidence provide different predictions
6 / 81
Why Do We Care about the Findings?
1) Responding average price weaken the incentive for conservation
2) Forecasts based on marginal price might be biased
17
Introduction Research Design Estimation Welfare Conclusion
Research question: How do consumers respond to nonlinear price schedules?
1 Standard economic theory predicts:
Consumers respond to Marginal Price
2 Laboratory experiments find:
Many individuals respond to Average Price = ( Total payment / Quantity )
Demand Curve
MP
AP
10
15
20
25
Ma
rgin
al P
rice
(ce
nts
pe
r kW
h)
Monthly Consumption
) Standard theory and laboratory evidence provide different predictions
6 / 81
So, What Can We Do?
18
The Key is:Providing Better Price Information
19
In My Other Research, I Provide “In-home display” for Electricity Consumers
20
Figure 2: Information on In-Home Display
&),"���01
% ������� ���,$�����-����� �#�+*!���'��!�����.���0(/���
Notes: This figure shows an example screenshot of the in-home display that are installed for bothcontrol and treatment consumers in the experiment.
16
Using the “In-home display”, Consumers Can See Real-Time Information about Price and Usage
21
Figure 1: Variable Critical Peak Pricing (VCPP)
15
50
75
100
150
0
25
50
75
100
125
150
Price
pe
r kW
h (
JPY
)
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of the day
Notes: This figure shows the price schedule of the variable critical peak pricing in the experiment.The marginal price of electricity is 15 yen for every hour in non-event days. When a CPP event dayis announced, consumers are informed about their critical peak price one-day ahead, which is either50, 75, 100, or 150 yen.
15
Randomized Field Experiment
Households
22
Treatment Group
Control Group
Random Assignment
When Consumers Have Clear Price Information, They DO Respond to their Price Incentives Correctly
23
Figure 3: Mean Consumption of Treatment and Control Groups by Peak-time Price
!2
!1.5
!1
!2
!1.5
!1
!2
!1.5
!1
10 15 20
10 15 20
15 50
75 100
150
Log o
f E
lect
rici
ty U
se p
er
30!
min
ute
inte
rval
Time of the dayGraphs by tp
Notes: This figure shows the mean of log consumption per 30-minute interval for the control (� ) andtreatment (⇧) consumers. I use data from all CPP event days to calculate the means. The verticallines show the critical peak time period, 1 pm to 5 pm.
17
Dynamic Pricing Group (Orange) and Control Group (Green)
Similar Findings from Other Studies: Providing Better Information is the Key
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Teaching Income Tax Code:
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Chetty and Saez (2013)
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Other Evidence from Electricity Pricing
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Wolak (2011), Jessoe and Rapson (2012), Kahn and Wolak (2013)
24
Summary: What Can We Learn from Recent Economic Studies?
1) Not-clear price information --> Consumers don’t get right price signals
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Evidence: both water and electricity consumers respond to average price
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Wrong price signals --> weaken incentives for conservation
2) Providing better price information is the key
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Evidence from electricity pricing and income taxation
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
Consumers respond to price signals correctly when they receive clear info
3) Discussion
Introduction Research Design Estimation Welfare Conclusion
2) Changes in price schedules can be used to estimate demand
4lnxit = ↵+ �4lnput(xit) + "it
Previous studies use simulated instruments (policy-induced price changes):
4lnpPIut (xit) = lnput(xit0)� lnput0(xit0)
Typically, the first stage is very strongAn identification assumption: a parallel trend between A and B
Consumption
Price
Household A Household B
35 / 82
How can we improve the clarity of water price information for consumers?
25