Extreme temperatures have a way of making people appreciate their utilities. Nobody wants to lose air conditioning during a heatwave or home heating during a cold spell. But Michigan residents faced a one-two punch on January 30, 2019. A fire broke out at the utility station that was set to provide more than half of the state’s natural gas, while a polar vortex caused temperatures to plunge to -13°F (-25°C).
When Dylan Brewer, an assistant professor of economics at Georgia Tech, learned of the availability of smart thermostat data before, during and after this energy emergency, he knew he had a perfect research project.
“Smart thermostats made by the Canadian company Ecobee are common in the Midwest, and many customers sign up for its ‘Donate Your Data’ program,” says Brewer, an environmental economist and EPIcenter faculty affiliate. “This voluntary program allows researchers to analyze anonymized home thermostat data, which can help utilities evaluate and improve their strategies for managing energy emergencies.”
To this end, Brewer and PhD student Jim Crozier compared Ecobee thermostat data in ~3,000 Michigan households with ~9,200 households in four neighboring Great Lakes states: Ohio, Indiana, Illinois and Wisconsin. Since all households responded similarly to an earlier cold wave in January 2019, the four states formed an appropriate “control” group for the “treated” Michigan households.
Following the fire outbreak, the utility company—Consumers Energy—purchased additional natural gas and asked customers to reduce their thermostat settings, using social media and local news stations. But this did not prevent the system from approaching a high risk of total failure as evening temperatures began to drop. At the utility’s request, the Governor of Michigan activated the FEMA Wireless Emergency Alert system at 10:30 pm, 12 hours after the fire outbreak. This system—the same that issues AMBER Alerts—sent the following text message to all of the state’s cell phone users: “Due to extreme temps Consumers asks everyone to lower their heat to 65 or less through Friday.”
Households in Michigan and the four control states had similar proportions of ≤ 65°F thermostat settings and similar average furnace fan run times—a proxy for natural gas consumption—during the 38 hours before the fire outbreak. Thus, the post-fire request to reduce home temperatures was the only difference between the two groups, allowing the researchers to establish a causal link between the text alert and the observed changes in thermostat settings.
The emergency request caused an average 1.1-degree reduction in Michigan households and a 45% increase in the number of households with ≤ 65°F settings, compared to control households. It also reduced furnace fan run times by 1.5 minutes per hour, a 6% reduction. These adjustments were sufficient to prevent a total failure of the natural gas facility.
Households responded to the emergency text alerts much more strongly than to the initial request, which only induced an extra 0.4% of households to conserve energy. The peak compliance rate after the cell phone alert was 20%.
“The salience of the request had a significant impact on household responses,” says Brewer. “The Governor’s decision to activate the Wireless Emergency Alert system was an effective strategy for preventing the worst outcome, in which 75% of households would have been unable to heat their homes in extremely cold weather.”
Response rates differed by political affiliation, based on 2018 county-level gubernatorial election results. Households in counties where the Democratic Party’s candidate received more than 70% of the vote reduced their thermostats by about twice as much as counties with less than 40%. This was a correlation rather than a causal link, notes Brewer.
While the Governor’s request prevented the worst outcome, the management of future energy emergencies could be improved. For example, critical peak pricing programs with a financial enrollment bonus allow utilities to increase electricity prices during short periods of extremely high demand, with a limit of 10 to 15 days per year.
Another financial incentive-based option is a peak saver rewards program offered to consumers at the time of their smart thermostat purchase. It gives the utility permission to remotely override consumer thermostat settings during an emergency. Such monetary incentives tend to be more effective than appeals, or moral nudges, to voluntarily conserve energy, says Brewer.
In an ongoing project similar to the Michigan study, he is analyzing flex alerts, or requests for voluntary energy conservation during heatwaves to prevent grid blackouts in California. More work is needed, he says, to compare different strategies for managing energy emergencies. For example, it would be useful to study whether a request for uniform five-degree reductions is more effective than a specific temperature target.
“We do not yet know which strategies are more cost-effective than others,” says Brewer. “I hope my involvement in EPIcenter will help me build partnerships with utilities in order to develop a suite of tools that decision-makers can use to avoid costly blackouts with dramatic consequences for consumers.”
News Contact
Published on: October 2, 2025
Story Written by: Silke Schmidt
Priya Devarajan || Research Communications Program Manager