RFF research finds that a program that automatically adjusts a home’s temperature based on electricity prices can save consumers money without causing significant discomfort.
Economists have long advocated for electricity retailers to charge prices that respond to fluctuations in the cost of electricity over the course of the day. If the prices that consumers pay vary with the cost of supplying the electricity, then consumers can factor these variations into their decisions about how much electricity to use, when to use it, and when to conserve it. Costs get especially high during occasional peak demand hours when additional, rarely used generators are called into service; allowing consumers to see prices that reflect those high costs could reduce peak demand and reduce the need to add expensive peaking power plants to the grid.
Despite the sound economics, electricity prices that vary throughout the day are seldom used. Consumer advocates and electricity regulators are not fans of exposing consumers to prices that reflect real-time fluctuations in marginal cost, otherwise known as time-varying prices. Moreover, several economic studies suggest that consumer responsiveness to time-varying prices is more limited than might be expected, perhaps in part because paying attention to how much electricity costs in each hour of the day and adjusting appliance use accordingly can require significant time and effort.
Enter automation. If energy-using devices can be programmed to automatically adjust electricity use in response to price changes, consumers would not have to worry about the attention costs of responding to changing prices and could save money in the process. In a recent RFF working paper, Josh Blonz, Casey Wichman, Derek Wietelman, and I explore how consumers respond to a new feature that enables smart thermostats to vary their operation in response to time-of-use prices for electricity. We find reason for optimism.
In the spring of 2019, we joined forces with Ecobee, a smart thermostat company, to design an experiment. Together, we randomly encouraged Ecobee customers in Ontario, Canada, to sign up for a new software algorithm that the company rolled out to existing customers that summer, called Eco+, which adjusts temperature settings automatically as the price of electricity varies over the course of the day.
We picked Ontario because most residential customers pay time-varying electricity prices that fluctuate among peak (11:00 a.m.–5:00 p.m.), mid-peak (7:00 a.m.–11:00 a.m. and 5:00 p.m.–7:00 p.m.), and off-peak hours (7:00 p.m.–7:00 a.m. and weekends). The prices during peak periods are double those during off-peak hours. Our experiment focused on 2,133 customers who had Ecobee thermostats and were randomly selected to receive messages—both on their device and on the Ecobee app on their phones—that encouraged them to enroll in Eco+. The lion’s share of the customers whom we encouraged to do so opted in to Eco+. The rollout began in early August of 2019, so our paper compares hourly data from the summers of 2018 and 2019.
In our experiment, we sought answers to three questions:
- How does air conditioner use change in response to enabling automated smart thermostat responsiveness to time-of-use prices, and what are the energy cost savings?
- What is the impact of automated thermostat responsiveness on household comfort?
- Do discomfort costs lead consumers to abandon the automation feature?
We find that enabling automation leads to pre-cooling during the early-morning hours, when electricity is the cheapest, and substantially less air conditioner use during peak hours. Indeed, peak-period air conditioner use falls by roughly 60 percent for the customers using the automated feature. While we don’t have data on actual electricity use, we make some back-of-the-envelope calculations on energy use, and those calculations suggest average savings of about C$4.40 per summer month for those who activate this feature.
We also find that household discomfort increases. We measure discomfort by calculating the difference between realized in-home temperature and preferred in-home temperature (informed by the customers’ settings in the month prior to the experiment). To ensure that discomfort is measured only when the occupant is actually at home, we monitor occupant presence by using the thermostat’s built-in motion sensor. While discomfort does increase slightly, that increase is experienced by only a third of households in our sample who are usually home during peak hours. This discomfort effectively means that households are slightly hotter than usual for these occupants, translating to a relatively small average increase in hourly temperature of 0.75 degrees Fahrenheit. Thus, most households in our experiment experience energy cost savings with no increase in discomfort within the home. This result is illustrated in Figure 1, which indicates that two-thirds of the households in our sample show no statistically significant increase in discomfort.
Figure 1. Home Occupants Who Spend Most of the Day at Home Experience Higher Temperatures Than They Prefer
Only one-third of households in the sample who are often home experience higher temperatures than they prefer as a result of activating the automation feature (ITT: intent to treat. Off, mid, and peak refer to times of day when demand for electricity fluctuates)
We also explore if residents who experience discomfort disable the automation feature, thereby limiting its long-run effectiveness. We find that very few customers who initially turned this feature on later turn it off. We also find that those who are always home during peak hours are no more likely than other customers to disable the feature. Moreover, we find that the effects of participation in the automation feature persist beyond the month of August and into the fall.
While the savings per household that we identify are small, they can add up across households. Aggregating across the 2,133 participants in our study yields a reduction of roughly 0.56 megawatts in peak demand during an average peak-period hour, but the effect could be much larger across the full suite of Ecobee customers in Ontario. Extrapolating our findings to California, which has about 1.6 million smart thermostats installed and is moving toward widespread residential time-of-use pricing, we find a 66-megawatt to 427-megawatt reduction in demand during an average peak-period hour on a warm or hot summer day, depending on assumptions about enrollment in time-of-use pricing programs. Other devices, such as hot water heaters and pool pumps, could be adapted to optimize their operation in response to time-varying prices. Our work suggests that such programs could deliver important energy savings and be designed in a way to minimize discomfort for customers. Such automation could help pave the way for more widespread acceptance of time-varying electricity prices.
Other devices, such as hot water heaters and pool pumps, could be adapted to optimize their operation in response to time-varying prices. Our work suggests that such programs could deliver important energy savings and be designed in a way to minimize discomfort for customers. Such automation could help pave the way for more widespread acceptance of time-varying electricity prices.
Decarbonizing the electricity system to avoid the worst consequences of global climate change will require substantial reliance on variable and intermittent sources of renewable energy, such as wind and solar. In a renewables-abundant future, the efficiencies we gain from pricing dynamically, with the aim of balancing demand and supply at each moment in time, will be even greater than today. Smart devices can reduce the cost of adjusting consumption in response to price variation. Smart devices also can help overcome political opposition to more widespread implementation of time-varying electricity prices. With nearly 20 states and the federal government targeting complete or substantial decarbonization of the electricity system in the next couple of decades, smart technologies could be integral to achieving these goals at a reasonable cost.
