Smarter Thermostats, Lower Bills, and Lower Emissions, with Casey Wichman

Notable Quotes

• Benefits of time-varying pricing: “Time-varying pricing suggests that there’s this mismatch between when electricity is cheap to produce and when people actually use it, especially as we have more and more renewable energy on the grid. We’re seeing that we have lots of clean energy during the day, and then, when individuals come home from work at the same time … that increase in demand needs to be matched with an increase in supply. Typically, that means coal or natural gas plants come online, which increases the cost of supplying electricity. So, we economists would say, ‘Let’s set an electricity price—a time-of-use price—that fluctuates throughout the day and that’s consistent with those changing costs of supplying electricity.’” (6:15)
• Consumers rarely opt in to time-of-use rates: “On average, based on the estimates, I’ve seen less than 2 percent of customers actually opt in to time-of-use rates, at least in the United States. But we are seeing an increasing push toward perhaps making time-of-use rates the default. You could always opt out, but one thing we know from behavioral economics is that default effects are very strong. We would likely see that customers would stick with the default and potentially optimize their consumption accordingly.” (24:37)
• Smart thermostats keep homes comfortable: “We find that very few people actually turn off [the smart thermostat feature]. We see an equal number of people actually increase their savings preferences, which scales how aggressively the algorithm will try to save you energy. We all started out being somewhat skeptical that these programs would work in the long run, but the data suggest that people are actually willing to tolerate some additional discomfort—and relatively small discomfort—for these small-dollar savings.” (27:47)

Top of the Stack

• “Smart Thermostats, Automation, and Time-Varying Prices” by Joshua Blonz, Karen Palmer, Casey Wichman, and Derek C. Wietelman
• “Savings Versus Comfort: How Smarter Thermostats Can Respond to Time-Varying Prices” by Karen Palmer
• The New Climate War by Michael E. Mann
• Why Fish Don’t Exist by Lulu Miller

The Full Transcript

Daniel Raimi: Hello and welcome to Resources Radio, a weekly podcast from Resources for the Future. I'm your host, Daniel Raimi. Today we talk with Casey Wichman, assistant professor in the School of Economics at Georgia Tech and an RFF University Fellow. Casey and several coauthors recently published a working paper that uses a field experiment to estimate how using smart thermostats and time-varying electricity pricing can reduce household utility bills and demands on the power sector. As more and more of us install smart thermostats, I'll ask Casey how much money these devices can help us save, how they affect the temperatures in our homes, and what they could mean for the grid's reliability and environmental impact. Stay with us.

All right. Casey Wichman from Georgia Tech down in Atlanta, welcome to the show. It's great to have you on Resources Radio.

Casey Wichman: Thanks, Daniel. I'm really excited to be here.

Daniel Raimi: Casey, we're going to talk about a recent paper that you authored with some RFF and former RFF colleagues. But before we do that—you haven't been on the show before—we always ask people how they got interested in environmental issues either as a kid or as you got older. What steered you into working on these types of things?

Casey Wichman: Sure. I've always been interested in the environment, and by that, I mean I was interested in being in the environment. When I was young, I remember spending every waking moment outside. We moved to a small town outside of Buffalo, New York, and we had several acres and a lot of woods in the back that butted up to Buffalo Creek. I spent all of my time outside in the woods, year-round in the snow, in the summer, when it was raining… apologies, this is the Georgia Tech steam whistle in the background, which occurs on the 15-minute mark of every hour, and then somewhat randomly at other times.

Daniel Raimi: That's great. You were telling me our office overlooks a power plant, right? So that's a steam whistle coming from the power plant?

Casey Wichman: That's right. It provides steam heat and cooling to buildings on campus.

Daniel Raimi: Awesome. Well, keep going. You were telling us about you growing up outside of Buffalo.

Casey Wichman: I spent a ton of time in the woods, in the dirt playing outside, playing sports, building treehouses, hiking, things like that. I didn't really understand that making a career out of studying the environment was a thing you could do; that was never an option that I really considered as a career. When I eventually went to college in Ithaca, New York, I realized that there were these opportunities to focus on the environment in a more structured way. I started college as an English major, not really knowing what I wanted to do. Maybe I had aspirations of being a writer or something like that.

I was taking classes all over the place, and I realized that I wanted to study in the environment. I didn't really know what that meant, but I ended up taking a handful of economics classes, and I realized that studying the environment through the lens of economics just made a lot of sense. It was very intuitive; I thought it was a really nice way to think about leveling the playing field in terms of the way we talk about the environment in terms of its cost and benefits so that you can communicate the benefits of environmental protection in dollar terms that can be translated to other domains. I had this epiphany in college where I decided, "This is what I want to do. This is what I want to focus on." I graduated college in 2009, and the opportunity cost of getting a job versus going to grad school was pretty low, so I decided to enroll in economics graduate school to really study the environment.

When I reflect on my early upbringing in enjoying the outdoors, enjoying the wilderness, I also was thinking that there's this family connection I have with the environment or environmental pollution: my grandfather worked at Bethlehem Steel from the day he graduated high school to the day he retired. I lived a few miles away from Lake Erie, which has a lot of pollution problems. This relationship between industry and the environment was always connected to my livelihood, and it's an interesting thing I've been thinking about as I reflected on this.

Daniel Raimi: That is really interesting. Well, I would love to ask you more questions about that. We could have a whole deep therapy session kind of thing, but let's instead treat our listeners to this new paper that you've co-authored with Josh Blonz, Karen Palmer, and Derek Wietelman. The paper's called “Smart Thermostats, Automation, and Time-Varying Prices.” We'll have linked to it in the show notes so people can check it out and follow along if they like. Can you start us off by helping us understand what the three elements that you call up in the title—“smart thermostats,” “automation,” and “time-varying pricing,”—mean and how could they work together to potentially save people money and reduce emissions in the power sector?

Casey Wichman: The way I like to think about this paper is that we're trying to focus on how these new “smart technologies” can interact with classical economic incentives to optimize energy use better than humans do on their own. Usually, as economists, we think about getting the prices right. There tends to be this hyper-obsession that economists have because once you set the right incentives, you can let market forces play out naturally, and then you achieve some efficient solution.

Time-varying pricing, or time-of-use rates in this context, suggest that there's this mismatch between when electricity is cheap to produce and when people actually use it, especially as we have more and more renewable energy on the grid. We're seeing that we have lots of clean energy during the day, and then when individuals come home from work at the same time, they turn on their Halloween lights and decorations and televisions and start cooking, that increase in demand needs to be matched with an increase in supply. Typically, that means coal or natural gas plants come online, which increases the cost of supplying electricity. So, we economists would say, "Let's set an electricity price—a time-of-use price—that fluctuates throughout the day that's consistent with those changing costs of supplying electricity." By matching those incentives, we would see people reducing consumption when electricity is more expensive to generate because, generally, we consume less of things when their price is higher, that's the law of demand. That’s what we mean by time-varying pricing or time-of-use prices—because they allow customers to see that that price signal fluctuates, and then they can alter their consumption behavior in response to that changing price.

We are also interested in the interaction of that classical economic incentive with this new suite of technology called smart thermostats: these Nest, Ecobee, Honeywell thermostats that have features and algorithms embedded in them that learn about your preferences. They learn about how long it takes to cool your household, and they attempt to optimize your electricity use based on what they learn about your household. They can communicate two ways: you can communicate with your thermostat, and your thermostat sends information back to the mothership—which might be Nest or Ecobee or Honeywell—that they then share that information with your electric utility. The key with smart thermostats is that they're learning, and they're able to do something that can be influenced by utility incentives or something like that. The automation component of the title is really focusing on what types of software features or automated features these smart thermostats facilitate that can help individuals respond better to these neoclassical or standard economic incentives.

Daniel Raimi: Great. That's really helpful. With all of those elements in place, you and your coauthors ran an experiment. Can you tell us about the experiment?

Casey Wichman: As a researcher, I'm always trying to pitch experiments to anyone who will listen because that’s really the gold standard of research. Whenever I have the ear of a utility company, someone who works at a municipal government, or in this case, a technology company that has access to hundreds of thousands of smart thermostats, my goal is always, “can I sell them on an idea, a randomized experiment so that I can learn something about the world?”

I can get a research paper out of it, and hopefully, they also learn something about their customer base as well. Karen Palmer and I visited Ecobee's headquarters in Toronto back in 2019 to share some work that we were doing with Ecobee's thermostat data already, just descriptive work trying to understand how people interact with their smart thermostats. We heard about this new program that they were about to roll out called Eco+, a software upgrade to existing thermostats that included a feature that automated responsiveness to time-of-use pricing.

We tried to sell them on the idea of running an experiment to measure the effectiveness of this Eco+ program rollout. We thought it would help them measure the effectiveness of this program and circumvent some standard selection concerns. How do you know you're measuring the causal effect of these thermostat features if you might have customers selecting in or opting in to the program? That might be correlated with some other thing. You might only be estimating the results for customers who really care about energy conservation. Tech companies tend to be pretty familiar with this type of randomization because they're constantly running A/B tests with their customers to determine what works and what doesn't for different products and services.

We worked with them to roll out this suite of features, Eco+, by withholding a random control group. In August 2019, Ecobee rolled out this new suite of features. We had selected a random control group that wouldn't get this push notification or updates on their cell phone app or their thermostat to enroll in this program. By comparing the outcomes of energy use and discomfort, we were able to, of the folks who were encouraged into the program versus the folks who were not encouraged into the program, compare the resulting energy use and changes in indoor temperature that households experienced to be able to measure the ultimate effects of this thermostat feature that automated responsiveness to time-of-use pricing. We did this in Ontario, Canada. Ontario may seem like an odd choice, but the one reason we wanted to do this in Ontario was that they have default time-of-use pricing, which is rare to find in the United States.

Daniel Raimi: Got it. Okay. That makes a lot of sense. There are tons of rich detail that I would encourage people to explore in the paper. We're going to gloss over lots of it and jump right to some of the results. Can you tell us what some of the key results were in terms of energy consumption for households?

Casey Wichman: Sure. We don't observe electricity usage at the household level, but we do observe a rich set of data in a variety of forms from the thermostat itself. The key metric that we're focusing on is HVAC compressor run time, and we use this as a proxy for energy use. What we actually see on our computers is, “how many minutes per hour is your AC system running?” This is a pretty intuitive way of thinking about your own energy use: you typically know when your AC system is running; you hear the compressor kick on and the fan kick on.

What we see with this automated time-of-use pricing feature is that right before these small increases in price that occur at two different points in the day—we have an off-peak period, a mid-peak period, and a peak period—the difference in electricity price between the off-peak period and the peak period effectively doubles. We see pretty large changes in electricity prices. The AC system begins to pre-cool the home right before these price increases go up, which suggests that the system is trying to cool the house more before the price goes up so that it doesn't have to work as hard during this high-priced period. Then, during the peak period, we see a big reduction in the frequency at which this AC system runs. We see this for all households who activated the smart thermostat feature, this time-of-use feature. That translates to about a 62 percent reduction in the amount of time that the AC system runs. And all in all, these reductions in AC system usage are notable, but the overall effect on a customer's bill is relatively modest because the AC use only comprises so much of your energy bill. These big reductions in energy use translate to about a $5 decrease in your average monthly electricity bill, which is modest.

It’s one reason why we might not expect these time-of-use prices to work on their own if the benefits of changing your behavior and changing your energy use profile throughout the day only have modest savings. That might not be enough to spur energy use changes on their own, but that's one of the reasons we think automation could increase the effectiveness of time-of-use pricing.

Daniel Raimi: Right. Absolutely. Of course, small changes for individuals can end up adding up to be pretty significant, and we're going to touch on that in just a moment. One of the other elements that you and your coauthors measure here is how changes in electricity consumption or changes in air conditioning running affect people's comfort level. In other words, how much discomfort are people willing to endure to save some money? Can you talk a little bit more about that?

Casey Wichman: Yes. This is actually one of the most fun things about this project. I mentioned that we didn't have access to electricity data, and most research in this space tends to focus on changes in kilowatt-hours. But because we don't have that, we were able to come up with a metric that is actually closer to the energy service that people consume, which is in-home comfort. We constructed a measure that we call “discomfort,” which is defined as the temperature wedge between your preferred temperature and your experienced temperature times the number of minutes you actually experience that discomfort. If you prefer your indoor temperature to be 74 °F but you experience the temperature to be 76 °F degrees for 60 minutes, we would call that “120 discomfort degree-minutes.” That metric doesn't matter that much, but we're capturing this deviation between your preferences and what you actually experience. This is another outcome variable we look at because this is the cost of those energy savings that we saw previously.

Overall, we find very modest impacts on discomfort, which suggests that for our full sample, there is a small increase in discomfort primarily focused during the peak period. Once we start digging into that, we find that for two-thirds of our sample who are typically not home during the peak period because they're out running errands or they're working in an office, we see that this discomfort effect is concentrated among the folks who are typically home throughout the day. These are in the top tercile of motion sensing that we can observe through the thermostat. Even though we find these increases in discomfort costs for the small set of folks who are typically home during the day, the costs tend to be pretty small. On average, it translates to a 0.75-degree-Fahrenheit change per hour in the average hour in our sample. We think these are pretty small effects that people likely are willing to bear or are likely not even going to be perceptible to most households.

Daniel Raimi: Yeah. That's so interesting. Can you talk a little bit more about that difference between people who were home during peak periods and people who weren't? Were you measuring discomfort only for people who were home? I think the answer is yes, but can you clarify that?

Casey Wichman: Yeah. So, there are two things. In our measure of discomfort, if no one is home to consume that discomfort, it comes up as a zero by our definition. One of the ways we leverage the data that we have access to through Ecobee is that the thermostats have motion sensors on them, so we know when somebody walks by their thermostat or another motion sensor that they have in their home. We use that to construct a measurement of who is typically home during typical hours based on their schedule. We can segment customers or households into three different groups, which we call often “home,” “sometimes home,” and “rarely home.” I might get the names slightly off, but during the peak period in the pre-treatment period, we essentially say, “do we observe motion in the home during a typical hour?” We break households up into these three groups.

For the groups that aren't typically home during the peak period, they see energy savings with no corresponding increase in discomfort. For the folks who are typically home—maybe these are people who are working from home—we see the same magnitude of energy savings, but they do have slightly higher discomfort costs. Again, these discomfort costs are pretty modest, less than a degree on average for each hour that they're home.

Daniel Raimi: Great. Thank you for that additional explanation. That's really helpful. Let’s turn now to this issue of scale that I mentioned a few moments ago. If this program were rolled out at a large scale, maybe beyond the borders of where your study focused, can you give us a ballpark of how much energy it might be able to save? What are the implications of those energy savings for emissions and also grid reliability if we think about really hot days when the grid is under stress? What sense do you have of how big this thing could scale?

Casey Wichman: Sure. This is something we were also curious about. Within our small Ontario study, we did a little back-of-the-envelope calculation to see how much energy savings we might actually see from this experiment itself. We were interested in this because even though these are relatively modest energy savings at the household level, we expected that if we added them up across a lot of customers, they might be substantial. Within our Ontario experimental sample, which is about 2,200 households, we see that translate to a 0.56 megawatt reduction in average hourly peak demand.

Now, what is that? This is something where my coauthors are even better at describing what exactly a megawatt is, but a megawatt on average can power anywhere from 400 to 900 homes in a year. Even within our small experimental sample, we see somewhat large reductions in peak period energy use that could be valuable to an electric supplier.

If we were to go out on a limb even further and extrapolate this to smart thermostats, say, in California—which we do because California is considering rolling out an opt-out time-of-use pricing program, so we think that this is a reasonable place to be studying these types of automated programs—if all California households that have a smart thermostat saw the same savings as we did in our experiment, that would translate to 427 megawatts of reduction in hourly peak demand, which is a lot larger. We also make some conservative assumptions about how many people will opt in to time-of-use (TOU) rates and opt in to the automated program. With that California sample, we find about a 66-megawatt reduction in hourly peak demand. To put this in scale, 66 megawatts of grid-scale battery storage costs around $100 million today. If we wanted to substitute that demand for grid-scale battery storage, we would see pretty big cost savings coming from this automated feature. There are a lot of assumptions and simplifications that we've had to do to make that translation, but I think it gives you a sense of the potential magnitude of these programs.

Daniel Raimi: Yeah, it absolutely does. And maybe just clarifying one acronym that you used, I think you said “TOU rates,” and that's time-of-use rates—is that right? That's the same thing as time-varying pricing basically?

Casey Wichman: That's correct. TOU rates, time-of-use rates, are a form of time-varying pricing. Not the only form, but the form that we are studying in this experiment.

Daniel Raimi: Great. Understood. You just mentioned California as one place where these types of programs might be rolled out at a larger scale. I've certainly heard about these types of programs from my utility where I live in Michigan. I have a smart thermostat, so I haven't actually opted in to these yet, but maybe I will. Can you give us a sense of how widespread these types of programs currently are? And in addition, what types of barriers you might expect, either political or social, that could impede them from being further deployed?

Casey Wichman: So, Daniel, you are not the only one who has access to TOU rates and has not opted in. I think that's one of the challenges here. I've seen varying estimates, but about 50 percent of electric utilities have some form of time-of-use rates available to customers, which means that you could opt in if you were so inclined. If you knew you had irregular hours that you could benefit from time-of-use rates, you might actually want to opt in to those types of rates.

On average, based on the estimates, I've seen less than 2 percent of customers actually opt in to time-of-use rates at least in the United States. Very few people are actually on time-of-use rates in the United States, but we are seeing an increasing push toward perhaps making time-of-use rates the default. You could always opt out, but one thing we know from behavioral economics is that default effects are very strong. We would likely see that customers would stick with the default and potentially optimize their consumption accordingly.

My sense is that these rate structures are not that popular because customers don't see very large cost savings. As I suggested, if you have this very intelligent thermostat feature optimizing for you and it’s doing probably as good of a job as it could, you're only seeing $5 savings a month—that might not be enough of a carrot to encourage customers to change their behavior in a substantial way.

But one of the nice things, and one of the things that I think we're somewhat optimistic about based on the results of this paper, is that if TOU rates were the default rate structure, and if a large number of customers were on a time-varying rate structure, we would be seeing more customers adopting smart thermostats. Some estimates of smart thermostat penetration range from 15 percent to 20 percent across the United States, so a lot of people have smart thermostats on the wall. This feature is very simple—well, it's not necessarily simple to build the software that goes into it—but you can push it out to hundreds of thousands of customers very easily with a routine software upgrade. You could actually get this automated feature operating in a lot of homes with a very little additional cost per household. We think that that's pretty encouraging, and that's one way where I think we can see some of these bigger effects scaled in a meaningful way.

Daniel Raimi: Yeah, that makes sense. Another thing I was thinking about as you were answering is the effects on the discomfort that you find: you highlighted that they're really quite small. I could imagine that some customers might be wary of opting in to these programs because they might fear some kind of scenario where they're sweltering in their house during a hot day and the discomfort effects are much larger than you actually find in your analysis.

Casey Wichman: Yeah, that's right. This is something we were suspicious of as well. I have a smart thermostat that I don't love because it thinks it's smarter than me. I always worry that it's making my dogs too hot in the summer when we're not home. I was suspicious that these programs were not going to work out because people would be uncomfortable and then just disable the feature.

But we actually explore that question explicitly with the data we have. We find that very few people actually turn off the feature. We see an equal number of people actually increase their savings preferences, which kind of scales how aggressive the algorithm will try to save you energy. I think we all started out being somewhat skeptical that these programs would work in the long run, but the data suggests that people are actually willing to tolerate some additional discomfort—and relatively small discomfort—for these small-dollar savings.

Daniel Raimi: Yeah. That's really interesting. Well, Casey, this has been really fascinating conversation and I've learned a lot and I'm sure our audience has as well. Let's close it out with the same question that we ask all of our guests, which is what's at the top of your literal or metaphorical reading stack? Something that you've read or watched or heard recently that you think is really great and that you'd recommend to our listeners.

Casey Wichman: Sure. As I was thinking about this, I couldn't decide, so I'm going to offer two options. The first is going to be a safe option that I think your audience will appreciate, and that is Michael Mann's The New Climate War. Michael Mann is a climate scientist who is an excellent communicator about the climate problem. I'm teaching a course called “The Economics of Climate Change” this semester to upper-level undergraduates, and I read this to get psyched up for the semester at the end of the summer. What I found most illuminating in this book is how clearly Michael Mann describes this disinformation campaign fueled by the fossil fuel industry about the severity of the climate problem and how this disinformation campaign is attempting and succeeding in sowing division within groups that still want to tackle climate change. I found it to be very eye-opening. It confirmed some things that I already knew, but I also think that Michael Mann's policy prescription for the climate problem is very economist-friendly, which I appreciated because economists often come out as the villain. We only want to price carbon, and that's not necessarily what environmentalists' first choice of a policy would be, but I think that there's a lot that I agreed with in this book.

Daniel Raimi: Great. And what's your less safe choice?

Casey Wichman: The second book—and this is a nod to my wife, who is always encouraging me not to read books by only white men, and it's a little more outside the purview of climate policy, energy policy specifically—is a book called Why Fish Don't Exist by Lulu Miller. Lulu Miller is a co-creator of Invisibilia, the NPR podcast, and I think a former producer of Radiolab, which is one of my all-time favorite podcasts.

It’s hard to categorize what exactly it is. It's part memoir, part history, of this individual, David Starr Jordan, who was a taxonomist and the founding president of Stanford. I was looking into him, and I actually realized he grew up in western New York in a small rural town in the 1850s that is 30 minutes away from where I grew up, in some bizarre coincidence. He was a taxonomist, and it's this really nice, pleasant, and beautiful way of looking for order in life and trying to understand this early taxonomist who was the first president of Stanford labeled all of these fish and how he viewed the world. He ended up having some pretty awful eugenicist views down the line, but connecting it back to the author's personal life. I struggle to actually communicate how incredible the book was, so I’ll just leave it at that and let your listeners take that book up for themselves.

Daniel Raimi: Yeah, that sounds great. I've heard that that book is great from our producer, Elizabeth Wason, and also from some others. So yeah, I should definitely check that out too.

Well, one more time, Casey Wichman from Georgia Tech, thank you so much for coming on the show and telling us about this fascinating new work that you've done with your coauthors. We really appreciate it.

Casey Wichman: Thanks a lot, Daniel. It was great to talk about this.

Daniel Raimi: You've been listening to Resources Radio. Learn how to support Resources for the Future at rff.org/support. If you have a minute, we'd really appreciate you leaving us a rating or a comment on your podcast platform of choice. Also, feel free to send us your suggestions for future episodes. Resources Radio is a podcast from Resources for the Future.

RFF is an independent nonprofit research institution in Washington, D.C. Our mission is to improve environmental, energy, and natural resource decisions through impartial economic research and policy engagement. The views expressed on this podcast are solely those of the podcast guests and may differ from those of RFF experts, its officers, or its directors. RFF does not take positions on specific legislative proposals. Resources Radio is produced by Elizabeth Wason with music by me, Daniel Raimi. Join us next week for another episode.