Can information make retail energy markets more equitable? Experimental evidence from the District of Columbia

We will administer two information interventions simultaneously among two distinct segments of DC’s electricity consumers. Both interventions will involve sending consumers information about how to make decisions that reflect their preferences in DC’s retail supply market. In the first intervention, we will send information to two treatment groups (via SMS messaging for one group and via mailed letters for the second group) of customers of retail suppliers charging more than the default utility service. In the second intervention, we will send information to three treatment groups (via SMS messaging, mailed letters, and email) of customers of the default utility service.

We will use administrative billing data to measure customers’ supplier switches one month following the interventions. All treatment and control individuals will then be invited to take a follow-up survey to measure consumers’ knowledge, preferences for different types of electricity services, and choices in hypothetical market scenarios.

2025-09-29

2025-10-06

Electricity supplier switches observed in administrative utility billing data

For the first evaluation, we will be sharing information with consumers on how to save money and find suppliers that meet customers’ desired preferences by either switching to the local utility’s standard offer service or to another retail supplier. For the second evaluation, we will be sharing preventative information with current customers of the local utility’s default service to help these customers protect themselves from high rates and/or predatory behavior from retail suppliers. Therefore, we will evaluate the impacts of treatments in both evaluations on customers’ electricity supplier switching behavior.

Consumer complaints filed and electricity suppliers’ comparison website traffic

We will measure secondary outcomes using Google Analytics website traffic data for an electricity suppliers’ contract comparison website and complaint records kept by the local advocacy office for utility customers.

This study uses a randomized controlled trial (RCT) design to evaluate the impacts from two sets of interventions, during which we share consumer protection information with randomly selected electricity customers in DC.

The randomized groups will include a pure control group of individuals across neighborhood clusters without any treatment households in these clusters, a control group of individuals across neighborhood clusters with treatment households in the clusters, a treatment group receiving information in a mailed letter, a treatment group receiving information in an email (for the second intervention among utility default service customers only), and a treatment group receiving information in an interactive text message.

We will store baseline billing data prior to the intervention for randomly assigned individuals. The intervention period will last for approximately one week. We will evaluate billing outcomes during the next monthly billing cycle to test whether treated individuals switch suppliers at a rate significantly different than the control individuals.

Following the next billing cycle after the intervention, we will administer a follow-up survey among all individuals in the control and treatment groups to evaluate outcomes that cannot be measured with billing data. Within the survey, we will include a short module of choice experiments to attempt to reveal sampled individuals’ preferences for different electricity service attributes, such as suppliers offering guaranteed renewable energy in amounts greater than the default utility service. We will offer $25 gift cards to a randomly selected group of up to 400 individuals who complete the survey.

We will randomize at two levels: the neighborhood cluster level and the individual level. We will assign neighborhood clusters to the pure control group until we reach a conservative threshold (i.e. in terms of power) of individuals within those clusters. We will use the same pure control clusters for both evaluations to prevent spillovers. Then we will randomly assign individuals within treatment clusters to the spillover control group or one of the treatment groups. The only shared attribute of randomization between the two simultaneous evaluations will be the pure control clusters, but the individuals assigned to the pure control group will belong to two distinct populations respective of each evaluation.

Neighborhood clusters and electricity account holder

No

47 (first evaluation) and 49 (second evaluation) only used for assigning pure control

14,284

Evaluation 1:

Pure control: 769
Spillover control: 864
Text treatment: 864
Mailed letter treatment: 864

Evaluation 2:

Pure control: 2,547
Spillover control: 1,117
Text treatment: 5,025
Mailed letter treatment: 1,117
Email treatment: 1,117

0.035 MDE with 0.012 SD and 80% power (expected control proportion of 0.044)