California Experiment: Dynamic Pricing for the Mass Market

Will the state launch a full-scale rollout of dynamic tariffs?

A pilot program in California is putting dynamic pricing and advanced metering to the test.

The California Public Utilities Commission (CPUC) approved a Statewide Pricing Pilot (SPP) in March,¹ at a cost of approximately $10 million, including metering, project planning, management, evaluation, and concurrent market research on non-pilot participants focused on customer preferences for rate options.²

The SPP has the following objectives:

Estimate demand curves for electricity consumption by time-of-use for dynamic tariffs, and derive the associated price elasticities of demand;

Gather information on customer acceptance of dynamic tariffs, control technologies, and information treatments;

Forecast the impact of a full-scale rollout of dynamic tariffs; and

Provide input into a cost-benefit analysis of universal deployment of advanced metering infrastructure.

Building on the past quarter century of research on time-differentiated pricing, the SPP breaks new ground in a number of areas. First, it will test time-varying rates against a backdrop of an inverted five-tier rate structure that was created in response to the energy crisis of 2000-01. Second, it will use an integrated sample design across three utility service areas. Third, it will use similar rates across the three utilities.

The SPP includes a traditional time-of-use (TOU) rate and two types of critical peak pricing (CPP) rates that feature a substantially higher peak price (about 50 to 75 cents/kWh) for 15 days of the year. One type of CPP rate, CPP-F, will feature a fixed peak period identical to the one in the TOU rate, and day-ahead customer notification. The other type of CPP rate, CPP-V, will feature a variable-length peak period, which can be called on the day of an "emergency."

The rates are designed to provide a significant price signal to customers, and to generate meaningful savings on their utility bills. For example, if a typical customer were to reduce/shift 30 percent of his peak-period load, he would see a 10 percent bill reduction. At the same time, if a customer does not change her usage pattern, the impact on her bill will be plus or minus 5 percent. Since the primary objective is to obtain econometrically sound estimates of own-price and cross-price elasticities, the SPP includes multiple rate levels for each rate type. The SPP also will measure the overall conservation-total usage reduction-effect of various dynamic pricing options.

An illustrative residential CPP rate is shown in Figure 1. The current average rate, which is a function of a complex, five-tiered rate structure, is approximately 12.7 cents/kWh in the summer. Under one of two experimental rate options, on most summer days, the CPP-F rate will charge roughly 24 cents/kWh during the peak period from 2 p.m. to 7 p.m. on weekday afternoons, and 7.2 cents/kWh the rest of the day and on weekends. However, on up to 12 summer days, the charge could be as high as 73 cents/kWh during the peak period.

The SPP includes a gross sample of 2,575 customers across six segments. This is expected to yield a net sample of 2,060 customers, assuming that 20 percent of the customers will choose to opt-out of the voluntary experiment. The net sample of 2,060 is expected to include 1,500 residential customers and 560 small commercial customers. Four residential segments capture the variation in customer price response across the state's climate zones; the two commercial segments capture the variation by size (0-20 kW, 20-200 kW). About 100 residential PG&E customers will be given information on how electricity costs vary by time-of-use and will be notified when critical peak days are encountered. But they will stay on standard rates, enabling an assessment of the impact of pure information treatment. Customers on the variable CPP rate will be provided with enabling technologies to help them automate a change in energy use. Customers will be offered a choice of air conditioner, water heater, and pool pump controls.

Three Tracks

The experimental design encompasses three tracks, reflecting the diverse interests of the members of the CPUC's third working group, WG3. Track A includes about 1,500 customers selected through a stratified random sample. It is designed to provide both statewide and climate-zone specific price elasticity estimates for TOU and CPP rates. Track B includes about 200 customers selected through a stratified random sample in the Bay Area. It will measure the effects of increased awareness of local environmental and reliability issues on price elasticities for the CPP-F tariff. Track C includes about 360 residential and small commercial customers, selected from the existing population of SCE and SDG&E customers, who are participating in the AB 970 pilot program featuring smart thermostats.

The SPP is designed as a social experiment, with random assignment of customers to control and treatment groups. Unlike traditional pilot programs designed to "get the bugs out" of a new product prior to its market launch, an experiment needs to have internal and external validity. Internal validity requires that the experiment be capable of establishing a cause-effect relationship between its "treatments" and outcomes. External validity requires that the findings be capable of being applied to outside the participant population.

SPP has been designed to avoid several of the design flaws that are commonly in experimental research. Social experiments are subject to a variety of flaws, including:

Failure to include a control group, which makes it impossible to measure cause-effect relationships;

Use of a non-comparable control group, which converts the experiment into a quasi-experiment;

Use of nonrandom sampling methods, which makes it difficult to extrapolate results to the population;

The lack of pre-treatment measurements, which makes it difficult to eliminate the effects of weather and other confounding variables on the observed relationship;

Insufficient number of treatments, which makes it difficult to estimate price elasticities of demand;

Insufficient sample size by treatment group, leading to imprecision in the estimated elasticities;

Insufficient variation in the average daily price, making it impossible to separate own-price elasticities from cross-price elasticities;

Voluntary recruitment of participants, which can introduce self-selection bias in the estimated model parameters;

Making compensatory payments to participants, which could lead to biased estimates; and

The Hawthorne effect, which arises from the impact of experimental observation on participants.

The gold standard of experimental designs includes before/after measurements on treatment/control groups. In practice, trade-offs must be made in all experimental designs because of budget, time, and political constraints. The SPP will include statistically comparable control and treatment groups and will feature random assignment of customers to these groups. However, because of the time constraints under which it is being conducted, the SPP will not have "before" measurements on the treatment groups. Nor will it obtain separate measurements on the effects of enabling technologies and information treatments. It will not test hourly prices or extreme day prices, nor will it test the effect of varying the length or timing of the peak period.

Decisions, Decisions

The SPP sample was developed through a Bayesian approach, based on decision analysis.

Given the wide range in potential net benefits of dynamic pricing noted earlier, California could make the wrong decision and implement a particular dynamic pricing option that is not cost-effective. The Bayesian approach uses estimates of the potential net benefits of the SPP treatments based on a priori information about price elasticities and the costs of sampling to determine sample size. For example, if Treatment A is likely to generate large net-benefits but there is great uncertainty in that result, it is appropriate to draw a large sample for A. Similarly, if Treatment B is always cost effective and Treatment C is never cost effective, there is no economic reason to sample either B or C.

SPP will use regression analysis to estimate the impact of time-varying tariffs on customer usage. This will involve the estimation of various demand equations that express usage in a time period as a function of prices in the various pricing periods and other causal variables such as climate, appliance ownership, and household demographics.

A typical demand equation for on-peak usage would look like this:

Ln (On-Peak kWh) = A + B1*Ln (On-Peak Price)
+ B2*Ln (Off-Peak Price)
+ B3*(A/C Ownership)*(CDH)*
Ln(On-Peak Price)
+ C1*(A/C ownership)*(CDH)
+ C2*(Persons Per Household)
+ C3*(Dwelling Type)
+ C4*(Pool/Jacuzzi)
+ C5*(Electric clothers dryer)
+ C5*(Income)

Various mathematical functions will be fitted to daily and monthly data. The best functional form will be chosen by reviewing a diagnostic battery of statistical tests.

Once the models have been estimated, they will be used to test a variety of hypotheses about how customers respond to time-varying rates (see companion article by King and Chatterjee, p. 27).

The SPP will go live in July and run through December 2004. The initial demand models will be estimated late this year. Their results will feed into a preliminary cost-benefit analysis of dynamic pricing, which is expected to be conducted in December.

Decision 03-03-036, Interim Opinion in Phase 1 adopting pilot program for residential and small commercial customers.

An additional budget of $2 million has been approved for testing (a) the impact of providing customers a choice of control technologies and (b) gateway systems that are always in two-way contact with the utility and can simultaneously control a variety of appliances in customer homes.

How the SPP Came To Be

The California Public Utilities Commission (CPUC) initiated a proceeding in July 2002 designed to introduce demand response in California's power market.¹ Three working groups were charged with developing specific tariff proposals to achieve increased demand response in the state. The mission of the third working group (WG3) was to develop a dynamic tariff (or set of tariffs) for residential and small commercial customers with demands of less than 200 kW. WG3 included representatives from the state's three investor-owned utilities,² commissions, equipment vendors, The Utility Reform Network (TURN), and other interested parties. As part of the WG3 deliberations, Charles River Associates (CRA) conducted a preliminary analysis of dynamic pricing for Pacific Gas & Electric Co. It showed a wide range of net benefits from the implementation of dynamic pricing. The net present value ranged widely, depending upon assumptions about meter and rate deployment strategy and costs, the level of customer demand response, and the magnitude of avoided energy and capacity costs. Analysis also indicated that conducting an experiment with a few thousand customers could reduce this uncertainty.

Based in part on this preliminary analysis, on Dec. 10, 2002, WG3 recommended that the state conduct a carefully designed social experiment with different pricing options in connection with making a decision on full-scale deployment of an automated metering infrastructure. -A.F. and S.G.

Order instituting rulemaking on policies and practices for advanced metering, demand response and dynamic pricing, Docket 02-06-001.

Pacific Gas & Electric, San Diego Gas & Electric, and Southern California Edison.

Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.