#23 Dr. Alec Brandon – When Do Nudges Persist in Energy Use: Evidence from 38 Experiments
In climate economics, the textbook answer to an unpriced externality is straightforward: price it—through a carbon tax or cap-and-trade. But over the past decade, a different approach has gained momentum: changing behavior directly through nudges.
One of the most successful examples is the Home Energy Report—a simple letter that tells households how their electricity use compares to their neighbors. The effects are modest but remarkably consistent (around 1–2% reductions), and the intervention scaled widely across the United States. But there’s also been a long-standing puzzle: most of the impact appears to persist even years after the reports stop—something that doesn’t happen nearly as often in other domains like charitable giving, voting, or savings.
In this conversation, I’m joined by Alec Brandon (Johns Hopkins Carey Business School), co-author of “Do the Effects of Nudges Persist? Theory and Evidence from 38 Natural Field Experiments.” (Link) We discuss what explains this unusual persistence—and why the answer may have less to do with habit formation than with something more durable: technology adoption that “locks in” behavior.
Once you see the mechanism, it changes how you think about nudges—not just as a way to shift short-run actions, but as a tool for triggering longer-run changes in homes, policies, and institutions.
Enjoy the conversation and let me know if you which topics you would like to see covered!
Arvid
Introduction
Arvid Viaene: Because climate change is essentially an unpriced externality, the obvious solution is to put a price on it, either through a carbon tax or cap-and-trade. We’ve discussed those solutions already on multiple episodes. However, another complementary approach that has gotten a lot of attention in the past decade is changing behavior directly through nudges.
One example is sending a home energy report that tells people how their energy use compares with that of their neighbors. Research has shown that these types of reports can reduce average energy consumption by 1% to 2% consistently.
That is a great result, because it is generally very hard to change people’s electricity use. As a result, energy providers in the United States have widely adopted these reports, and policymakers have praised them heavily. The company that developed them was even acquired for more than $500 million.
While that is a great success story on the surface, there has also been a puzzle in the literature: the majority of the short-run effects persist even two years after the energy reports are discontinued. However, that result stands in contrast to other areas like charitable giving, financial savings, voter turnout, and so on, where those nudges do not persist.
So the paper we are going to discuss today figures out why nudges in energy conservation can be so successful in homes, and why they may be persistent in ways that differ from other contexts.
Once you get the idea—which is both simple and really cool—you may start to see opportunities to apply these insights to climate policy, other policy areas, and maybe even your personal life.
I’m happy to have Alec Brandon on. Alec is an author of the paper and an assistant professor at the Johns Hopkins Carey Business School. His research focuses on behavioral economics.
Using a variety of empirical methods, Alec has studied the long-term effectiveness of nudges in different areas, as well as the extent of racial profiling in policy. His research has been published in leading journals like Science and the Proceedings of the National Academy of Sciences.
Before joining Carey, Alec received his PhD in economics from the University of Chicago. Alec, welcome to the podcast.
Alec Brandon: Thank you.
Arvid Viaene: I’m excited to have you on. I also forgot to mention the title of your paper during the introduction, so let me do that now. The title is Do the Effects of Nudges Persist? Theory and Evidence from 38 Natural Field Experiments. What is your paper about, and what would you like people to take away from it?
Alec Brandon: Big picture, our paper is about trying to understand whether this really promising line of research on nudges can produce persistent, long-term effects. That matters in general, but it is especially important for something like climate change. If nudges do persist, the next question is why. And if we can answer that “why” question, then maybe we can replicate the success of persistent nudges and design them more effectively. That is probably the multi-sentence summary.
What the paper finds
Short-run nudges versus long-run effects
Arvid Viaene: I think that’s a really good summary of why this research matters. What did you find in the paper?
Alec Brandon: One of the things that motivated us was the mixed early evidence on whether nudges continued to change behavior after people stopped receiving them. Back in the earlier period—say around 2015—there was some evidence that nudge effects could persist for a couple of weeks or maybe a few months, but they tended to fade fairly quickly.
Arvid Viaene: What would be an example of that kind of nudge?
Alec Brandon: The examples that come to mind are nudges encouraging people to exercise more, turn out to vote, or give to charities.
In those settings, it did not really seem like nudges were transforming who people were.
But then there were a few papers on water conservation and energy conservation that found really striking persistence. Some water conservation studies found that a single letter led to savings that lasted for years.
There were also industry white papers and academic papers on the home energy reports we study, showing that 80% or 90% of the effect remained even after the reports were randomly discontinued.
That raised the question for us: why would nudges persist in one area but not in others?
The core idea: technology can lock in behavior
Alec Brandon: The simple idea we had was that maybe the reason nudges persist in energy and water conservation is that those are settings where people can adopt technology.
If someone tells you, “You’re using more energy or water than your neighbors,” you can respond by making a change to the home. That technology can then keep producing conservation even if behavior itself doesn’t continue changing.
Other behaviors are not really like that. Voting, charitable giving, and exercise do not have the same kind of built-in technological fix.
That observation is what set us down this path.
Arvid Viaene: Right. With voting, for example, you might get a nudge and then go vote. But at the next election, if you no longer get the nudge, you stop going.
Whereas with an energy report, the nudge might get you to install something more efficient, and then that change remains.
Alec Brandon: Yes, that is exactly the idea.
Why the authors were skeptical of habit formation
A different explanation from the standard story
Arvid Viaene: Did you already have that idea going in, or did it develop as you looked at the data?
Alec Brandon: We more or less had that idea going in.
At the same time that nudges were becoming popular, there was also a lot of work on habit formation using experiments, often with incentives. Researchers would incentivize a behavior, remove the incentive, and then see whether the behavior persisted.
There really was not much evidence of habit formation at the level needed to explain the persistence being found in energy and water conservation.
Some papers from around 2008 or 2009 might find that 5% or 10% of the effect persisted and call that habit formation. That may be enough to reject a null of no habit formation, but it is nowhere near large enough to explain the levels of persistence people were reporting for some nudges.
So we were skeptical of habit formation as the full explanation.
To be completely honest, the habit formation model began as a way to describe things like addiction to cigarettes or heroin. That story makes sense to me because there is a chemical component involved.
But with a brief incentive to exercise, save more money, or respond to a nudge, I don’t think that is really changing who people are. And nudges, by design, are supposed to be light-touch. They are not intended to transform identity.
So one building block of the project was skepticism about the usual mechanism people were pointing to.
The importance of households moving
Alec Brandon: A second building block came from the earlier Opower papers.
Arvid Viaene: Opower, by the way, is the company that sent out these energy reports.
Alec Brandon: Yes, thank you. Opower was the company that designed and developed the Home Energy Report.
In the earlier papers, the researchers mentioned that they had to drop homes when someone moved. They were doing that for econometric reasons—they wanted a balanced panel.
We don’t need to get too deep into that. But we noticed that the company could observe when someone moved, and we realized that feature could help separate the effect on the individual from the effect on the home itself.
How home energy reports work
The design of the intervention
Arvid Viaene: I think that gets to the core of your paper. Could you explain the setup of the energy report experiment?
Alec Brandon: Sure. Let me start with the report itself.
There were at least 100 of these home energy report experiments run between roughly 2008 and 2015, and they may still be running. The basic setup was that an energy provider partnered with Opower.
To demonstrate effectiveness, Opower would create a small control group—maybe 20% to 40% of homes—and then a treatment group that received these reports monthly or quarterly.
The letters told households how much energy they used, how much their neighbors used, and how much their most efficient neighbors used. They also included a simple evaluation of your performance.
If you used less than your efficient neighbors, you might get a smiley face. If you used more than your efficient neighbors but less than all neighbors, you got a neutral face. If you used more than all your neighbors, you got a frowny face.
Those letters were sent regularly. Opower also tested different frequencies, but the exact cadence did not seem to matter much.
Why movers reveal the mechanism
Alec Brandon: There was also another quirk: once the homeowner moved, the house was removed from the experiment and the letters stopped.
I do not think that was an intentional research feature. It was just common sense. The energy provider did not want the new resident receiving a rating based on someone else’s energy use.
That ended up being very useful for us. If the letters stop after the original resident moves, but Opower continues observing energy use in the home, then we can measure how much of the effect stays with the home itself.
That lets us separate the effect on the original person from the effect that persists because of something about the house.
Arvid Viaene: Because before that, the assumption was that people received the nudge and changed their own behavior in the home.
Alec Brandon: Exactly.
The main result - More than half of the effect stays with the home
Alec Brandon: The effect of getting the home energy report on the household is about 1% to 2%. In our sample, it was 2.1%.
Then, once the initial resident moves, the report still has an effect of about 1.1% on the home. So a little more than 50% of the original effect remains after the original resident is gone.
Arvid Viaene: And that is the technology story.
Alec Brandon: Yes. We attribute that remaining effect to technology adoption.
The rest we attribute to habit formation, or at least to some role for habits.
Renters versus homeowners - Why the distinction matters
Arvid Viaene: One of the most interesting parts of your paper is the difference between renters and homeowners.
In general, you find that about half of the total effect is habit and half is technology. But the renter-owner comparison is especially interesting.
Alec Brandon: That analysis started partly as a way to convince referees that what we were doing made sense, but it turned into an interesting result in its own right.
Earlier papers on the Home Energy Report found that whether someone rented or owned their home did not affect whether the report produced energy savings.
That is interesting because, especially in the United States, there is not much of a culture of renters adopting energy-efficient technology.
Part of that is contractual. If you rent, you usually are not allowed to make major changes to the property.
But it is also financial. Why would you invest in energy-efficient technology if you are not going to live there long enough to recoup the benefit?
Renters mostly change habits, not technology
Alec Brandon: What we find is that renters respond similarly to homeowners while they are in the home. But once the renter moves out and a new renter moves in, none of the effect remains.
That suggests renters are changing something significant while they live there, but that change leaves with them. So in that group, the persistence seems to come from habits, not technology.
If someone believes in habit formation, this is a result they can really point to.
Arvid Viaene: That makes sense. Renters mostly have one way to reduce energy use: they change their behavior.
They turn off lights more consistently, adjust how they use appliances, and do all the small daily things that reduce electricity use. In that sense, their changes are cost-free.
When they leave, they take those habits with them.
Alec Brandon: Yes, though I would add that in an accounting sense they are cost-free. There may still be psychic costs or effort costs to adopting those habits.
I am not an expert on that specific question, but I agree with the general point. In terms of standard measurable outcomes, habits are appealing because you are getting something without a direct capital expenditure.
The hidden cost of persistent nudges - Technology adoption changes the cost-benefit calculation
Arvid Viaene: One of the important points in your paper is that when homeowners respond by adopting technology, that is not cost-free. It may have been overlooked in earlier cost-effectiveness discussions.
Alec Brandon: Yes, that is one of the main implications we try to emphasize.
A lot of people want to evaluate whether interventions work, whether in government or in the private sector. I think that kind of evaluation is extremely important.
But it is tricky, because the outcomes you observe may not capture everything people are doing in response to the intervention.
In our setting, if homeowners are adopting energy-saving technologies because of the reports, then those expenditures should be counted when you compare the benefits of the program to its costs.
We did not have direct information on exactly what technologies people were adopting, so we used back-of-the-envelope estimates from the literature on the cost per kilowatt-hour saved through technology adoption.
We chose a lower bound and an upper bound.
The exact estimate of net benefits depends on those assumptions. But the broader conclusion does not: the program still has positive net benefits either way.
Arvid Viaene: So it remains net beneficial, but the total benefits are meaningfully smaller once you account for those costs.
Alec Brandon: Yes. It takes a chunk out of the benefits.
That is basically just accounting, and accounting points are nice because you can be fairly confident you are right.
Why earlier papers missed the technology channel
Rebate data was too narrow
Arvid Viaene: Earlier papers, such as Allcott and Rogers and another by Ferraro and coauthors, looked at technology and seemed to rule it out. Why are you more confident their conclusion was incomplete?
Alec Brandon: Those papers were broadly contemporaneous, and they looked at related interventions.
The Allcott and Rogers paper, on energy conservation, used data from utility rebate programs. The idea was to see whether people who got the Home Energy Report were more likely to participate in rebates for energy-efficient purchases.
They found basically no difference between treatment and control. Their back-of-the-envelope conclusion was that maybe only 1% of the report’s effect came through that rebate channel.
The water conservation paper used an early version of a movers design and found that no effect remained in the home.
I suspect that in both cases, these analyses were somewhat speculative and placed toward the end of the paper.
In the case of the rebate analysis, I think the issue is that many of the relevant technology changes are too small or too inconvenient to show up in rebate data.
If someone installs efficient light bulbs or adds a door sweep, there may technically be a rebate. But are people really going to file the paperwork for that? Probably not. Those programs are notoriously undersubscribed.
So unless the Home Energy Report was causing large, contractor-driven renovations, rebate participation was probably the wrong thing to measure.
Observing the home directly is different
Arvid Viaene: Right, and your design is different because you do not need to infer behavior from rebate participation. You keep observing the home after people move out.
Alec Brandon: Exactly. It is always nice to have data that directly tells you what people are doing, but we basically accepted that this might be impossible in this setting.
The kinds of changes people make may simply be too small to observe directly.
That is actually a broader challenge with nudges. Nudges usually have small effects, but they are cheap enough to implement at large scale, which makes it possible to detect those effects.
The flip side is that small effects are hard to unpack mechanistically. It is hard to run surveys with high response rates when you are just sending letters. It is hard to get the “smoking gun” evidence people would ideally like.
So our approach is to infer the mechanism from what remains in the home. That is a strength, but also a limitation.
Arvid Viaene: Right. You cannot identify exactly which technology people installed. You can only see that something about the home changed.
Alec Brandon: Exactly.
What this means for climate policy
Nudges work better when paired with something durable
Arvid Viaene: One thing I really like about the paper is that it changes how you think about nudges.
Instead of only asking whether a nudge changes behavior in the short run, you ask whether it can trigger a technological or institutional change that keeps affecting behavior over time.
For example, with voting, maybe the nudge could get someone to install an app or sign up for a system that keeps nudging them. In climate policy, maybe the key is to get people to adopt something that keeps the effect going.
Alec Brandon: I think that gets to the most important implication of the paper.
Nudging for the sake of nudging is still worthwhile. It helps us learn what works, and the things that work can sometimes be scaled.
But I think it also makes sense to ask what the underlying response actually is. What do you expect people to do in response to your nudge? And are the conditions there for that response to stick?
In energy, that is straightforward. If there is a low-cost technology you want people to adopt, then a nudge that grabs attention and gets people to consider action can complement that technology nicely.
So instead of taking a purely reduced-form approach—“I do not care how people respond, I only know the experiment works”—our paper is pushing people to think more concretely.
What technologies would be useful if people adopted them today? Can you design a nudge that gets attention and leads to adoption?
Or, alternatively, can you pair a nudge that works in the short term with something that makes long-term persistence more likely?
Pair nudges with defaults or new incentives
Alec Brandon: For example, maybe a voting nudge could have stronger long-term effects if it also included a form that helps people sign up for mail-in voting. That would reduce the cost of repeating the behavior in the future.
Or in energy, maybe the nudge could encourage enrollment in critical peak pricing, which changes the incentives people face over time.
That is still a conjecture, but that is where we hope people take the paper: thinking more clearly about what exactly is going to produce the long-run effect.
Arvid Viaene: Yes, being more rigorous about what is actually changing.
I also like your point that behavioral economics is very good at grabbing attention. The smiley-face report gets people to notice something. Then you can use that moment to connect them to a pricing program, a technology, or a more durable change.
Alec Brandon: Exactly.
Behavioral economics offers a very useful toolkit for getting people’s attention. The field has a more uneven track record when it comes to very long-run effects.
So one promising direction is to combine those tools with more traditional economics—changing incentives, reducing friction, encouraging technology adoption.
That could let you get the best of both worlds: the immediate change in behavior and the more durable effects that come from structural changes.
Examples beyond the paper - Heat pumps, solar panels, and framing
Arvid Viaene: I think that is especially relevant for climate and energy policy.
Take heat pumps in Europe. Energy prices are high, and for many homes heat pumps can be a very good option. But a lot of people are not aware of what they are or how they work.
Behavioral economics seems useful here because it can grab attention and help people recognize that something could genuinely benefit them.
Another example is a paper by Verboven and De Groote on solar panel adoption in Belgium. Unlike in the United States, Belgium had a very complicated subsidy system.
It was not something simple like “you buy this and get 50% back.” Instead, you bought something for €10,000 and got the rebate gradually through electricity credits over many years. People just did not realize how favorable the deal was.
That seems like a case where behavioral tools could help people understand how valuable the policy actually is.
Alec Brandon: I agree.
One way to put that is that prices are always framed in some way. There is no such thing as a frame-free price.
Behavioral economics gives us tools to think about which frames are better and which are worse.
What you are describing in Belgium sounds like a case where the policy may have been framed in a way that made it much harder for people to understand how attractive it actually was.
Additional result from the paper - Homes with electric heating show more persistence
Arvid Viaene: Is there anything else from the paper that we have not covered and that you would like to add?
Alec Brandon: The only other thing I would emphasize is a result that is sort of a counterpart to the renter finding.
We also had a similarly sized sample of homes that used electric heating rather than natural gas. Since our outcome is electricity use, those homes had more scope to generate savings through technology adoption.
And that is exactly what we see. In those homes, around 60% to 65% of the initial effect persists, compared with about 50% on average and zero for renters.
So the story I tell myself is that people may prioritize technology adoption because it is the easiest way to get a better report card from the utility.
But if that option is not available, and you keep sending the letters long enough, maybe people do adopt habits instead.
That is a conjecture, but I think it is a useful way to connect the different results.
Related research on energy technology
Smart thermostats may not save energy
Arvid Viaene: Are you doing any further research in the energy nudges space?
Alec Brandon: Not at the moment. But I am finishing another paper that your listeners might find interesting.
It is a more conventional technology-evaluation paper in which smart thermostats were randomized.
In the United States, there is a lot of subsidy money available for people who replace their thermostat with a smart thermostat. These thermostats connect to the internet and can be controlled from your phone.
We studied a field experiment in California where people were recruited and, if randomized into treatment, got a smart thermostat installed for free.
And we do not find any energy savings from the people who received the smart thermostat compared with those who did not.
So I am not claiming that technology is always the answer. Smart thermostats do not seem to deliver the kind of energy savings that households in the home energy report study were apparently finding through other changes.
Arvid Viaene: That also shows the kind of critical mindset you need as a researcher.
Even if you are interested in technology as a mechanism, you still need to test whether a specific technology actually works.
So if anyone listening works at an energy provider, has a promising intervention, and wants to run a serious field experiment, Alec is clearly someone worth contacting.
Alec Brandon: Yes, feel free to email me.
Arvid Viaene: Alec, thank you so much for taking the time.
Alec Brandon: My pleasure.
