A residential electricity demand response field experiment
How to cite
Bailey, Megan; Brown, David; Shaffer, Blake; Wolak, Frank. 2022. "A residential electricity demand response field experiment." AEA RCT Registry. AEARCTR-0008680. https://doi.org/10.1257/rct.8680-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Electricity retailers must ensure that system demand is less than or equal to available generation capacity at all times. One way for the sector to achieve this outcome is through demand-side management actions that incentivize or require consumers to reduce electricity demand during peak times. In partnership with a large electricity retailer, we randomly offered residential electricity customers one of several treatments composed of combinations of smart home electricity management equipment and financial incentives for reducing their demand.. About 1100 households were ultimately enrolled in the experiment across all of the treatments. Our objectives are to evaluate: (1) the relative take-up rates and characteristics of consumers that accept offers for each treatment based on take-up incentives and customer demographic characteristics; (2) the average treatment effect on the treated in terms of (a) electricity demand reduced during peak load hours and (b) the reliability of the amount of electricity demand reduced, for those who accept; and (3) intent-to-treat estimates for each offer as a program, using (a) and (b). We also plan to estimate heterogenous treatment effects along a number of household characteristics from each intervention.
Intervention
The intervention is composed of event days where customers notified that they will be compensated for demand reductions relative to a baseline level of demand during pre-selected periods of the day. These event days are randomly assigned across days of the month and customers. Event days are declared an average to three times per month for each participant in the experiment.
Experimental design
We randomly offer one of six treatment or control offers to residential electricity customers in our partner’s service area. We followed a stratified random assignment approach in which we first used a machine learning algorithm (k-means) to identify clusters of households based on answers to survey questions and geographically matched Census data. We then randomly assigned households from clusters across treatment groups to obtain balance along observables with random assignment. Households were offered randomly assigned up-front incentives to participate in addition to the benefits of the treatment/control group to which they were assigned.
Treatments consist of combinations of smart home electricity use management equipment and incentives for reductions in peak demand time electricity usage. To estimate average treatment effects, we plan to use a selection model to account for customers’ selection into treatment offers.
Design details
Cited by
No citation data available for this registration yet.