3,916 households serviced by Groton Electric Light Department were randomly assigned to one of five groups concerning a pilot program for a new dynamic pricing structure. Customers of the utility were informed via newsletter of the upcoming introduction of a new pricing rate that is “almost half our flat-rate 20 hours a day and more than triple our flat-rate from 4:00 pm and 8:00 pm.”
1. Control: 1016 households do not receive encouragement but can opt in to the new rate at any time.
2. Shadow Bill Opt-In: 508 households receive a direct mailing from the utility offering them the opportunity to receive a “shadow bill” of the new rate in the coming summer (2021)
3. Shadow Bill Opt-Out: 508 households receive a direct mailing from the utility informing them that they will receive a “shadow bill” of the new rate in the coming summer (2021) and offering them the opportunity to opt out of the “shadow bill”.
4. 4/40 Opt-In: 1900 households receive a direct mailing from the utility offering them the opportunity to opt-in to the new rates. Of these households, 940 are also told they will receive a “shadow bill” of the new rate in the coming summer but they may opt out of receiving the “shadow bill”.
5. 4/40 Opt-Out: 316 households receive a direct mailing from the utility informing them that they will be placed on the new rate structure in the coming summer (2021) and are offered the opportunity to opt out.
All customers of the utility were invited to participate in a customer survey collecting information on household characteristics and behavioral parameters including expectations, loss aversion, present bias and risk preferences.
We randomize along two dimensions: shadow bill and the new time of use (TOU) rate structure, henceforth 4/40 since it is $0.04 kwh off-peak and $0.40 kwh peak. Based on the agreement with our partner our experimental design cannot force any consumer to remain in a given treatment assignment. Therefore, we relied on the established default bias for our randomization. This assumes that the probability of remaining in a treatment assignment is higher for households that need to opt-out compared to households that need to opt-in.
Our first hypothesis is that consumers select rate plans that deliver electricity services at the lowest possible cost. We define the expected switching benefits E[SB] as the difference between the total costs (TC) on the uniform plan minus the total costs on the 4/40 plan conditional on electricity use staying constant. We define electricity use as e(t) to account for the full distribution of consumption over time.
E[SB] = E[TC(Uniform)|e = e(t)] - E[TC(4/40)|e = e(t)])
Our next hypothesis studies the impact of information on plan choice. We expect consumers with access to shadow bills to have more information on the expected benefits from being on the 4/40 plan. Therefore, we analyze heterogeneous effects with respect to switching benefits based on access to the shadow bill. We expect the responsiveness to expected switching benefits to be a function of information provided in the shadow bill (δ(Shadow Bill)).
H2∶ (∂δ(Shadow Bill))/(∂Shadow Bill) > 0
The goal of the new plan is to decrease peak electricity use. Our first hypothesis on electricity use is that the real time rate decreases electricity use during the peak hours e(t|t = Peak). We will exploit the random assignment to default plans to estimate H3 below.
H3∶ e(t|t = Peak,4/40) < e(t|t = Peak,Uniform)
Next, we will examine heterogeneity in electricity use based on active or passive joiners. We hypothesize that consumers who actively opt into the 4/40 plan (T3) will reduce peak consumption by more than consumers who passively do not opt out (T4).
H4∶ e(t|t = Peak,4/40,active) < e(t|t = Peak,4/40,passive)
Next we will examine two hypotheses based on how the shadow bills affect peak electricity consumption. First, we hypothesize that some households that receive shadow bills, yet remain on the uniform rates will reduce peak consumption immediately. There may be two main mechanisms behind this short-run reduction despite not facing hire peak prices. First, consumers may use the opportunity to learn how to adapt to the new rates that will be universal next year. Second, consumers may understand that peak consumption is more costly to the community and reduce consumption out of good will. This leads to our next hypothesis.
H5∶ e(t|t = Peak,Uniform,Shadow Bill) < e(t|t = Peak,Uniform,No Shadow Bill)
For the subset of consumers who completed the survey we will analyze heterogeneity in our hypotheses along several dimensions:
Financial literacy on plan choice
Present bias on plan choice
Stated elasticity on plan choice and peak use
Work from home status