In addition to their normal utility bills, over 50 million households and businesses also receive customized energy reports several times a year from a company called Opower. These energy reports compare the amount of energy a specific household uses to the energy usage of the surrounding homes, thereby providing feedback to households about their relative energy consumption.

These reports have been highly effective at changing people’s behaviors and reducing energy consumption.  Thus far Opower’s intervention has saved eight terawatt-hours of energy worldwide, which is about the amount of energy needed to power all the homes in New Mexico for one year. But even more importantly, the Opower mailings generate a long-term reduction in energy usage. Households not only decrease their energy consumption after receiving a mailing, but they continue to use less energy for years, even after the Opower mailings stop.

This is noteworthy because many interventions that initially have positive effects do not produce long-term behavior change. For example, fad diets may cause people to lose weight initially, but dieters often gain the weight back over time. In contrast, the Opower energy reports not only reduce people’s energy consumption initially, but they also cause people to use less energy in the long-run. So why do some interventions result in persistent behavioral change, while others do not?

This is the question that Todd Rogers and I explore in our paper, “Persistence: How Treatment Effects Persist After Interventions Stop”, published in Policy Insights from the Behavioral and Brain Sciences. We propose a framework for understanding how and when interventions may lead to persistent behavior change. Specifically, we identify four “pathways”, or features of interventions, that may explain why some interventions are successful at generating persistent behavior changes. These pathways include (1) habit formation, (2) changing what or how people think, (3) changing future costs, and (4) external reinforcement

Habit formation occurs when an intervention causes people to repeat a behavior over and over again until they eventually perform the behavior automatically, without any continuing influence from the intervention. It could be that the Opower reports initially made people pay more attention to turning off lights when leaving a room, and over time they developed a habit of turning off lights, which continued even after the Opower reports stopped.

The second pathway works by changing what or how people think.  People may fail to change their behavior because they hold an incorrect belief, because they don’t see themselves in a particular way, or because of the way they interpret information. If an intervention is able to change what people believe about themselves or about the world, or if an intervention changes how people interpret their experiences, then it may lead to persistent behavior change. For example, people who received the Opower mailings may have learned that they were not as energy efficient as they initially thought, which made them take steps to use less energy. Alternatively, the Opower mailings may have made people use their air conditioning less, but instead of interpreting a warmer house as uncomfortable, these people may have interpreted the warmth as a positive sign that they were saving electricity. Thus, they may have continued to use less energy going forward.

Interventions may also generate persistence by changing future costs. Some behaviors are costly to perform, and people may need considerable time, attention, self-control, material resources, or effort to carry them out. If an intervention permanently reduces the “cost” of these behaviors, people may be more likely to perform them in the future. For example, the Opower mailings may have prompted people to purchase energy efficient appliances. Since these appliances automatically consume less energy than conventional appliances, they would reduce future energy use without requiring any additional time, effort, or attention from the owner. Thus, energy efficient appliances may have reduced the mental “cost” of saving energy in the future.

The final pathway we identify is external reinforcement, or exposing people to experiences or settings that reinforce the desired behavior. Reinforcement may occur in several ways, including through social situations. In the context of Opower, the energy reports may have made people talk about energy efficiency with their friends and families. If they made verbal commitments to these people to reduce energy, they may have felt accountable and followed through on their energy reduction. Repeatedly seeing these same friends and family members may have reinforced their commitment, thus leading to persistent decreases in energy use.

Energy efficiency is not the only domain that can harness these pathways to generate persistent behavior change. For example, interventions in personal finance already leverage the concept of reducing future costs. Brigitte Madrian and Dennis Shea (2001) found that defaulting employees into 401(k) savings plans (as opposed to making people opt-in on their own) significantly increases the number of people participating in these savings plans, which changes people’s savings behavior in the long-run. This intervention reduces the amount of mental effort that people need to put into thinking about saving, and therefore it reduces the “cost” of saving in the future and generates persistent saving behavior.

It is easy to imagine how these pathways could be leveraged to persistently change behavior in other areas as well, such as exercise. For example, interventions could be designed to make people form exercise habits. An intervention could incentivize people to take the stairs instead of the elevator, and if people take the stairs enough times in a row, they may develop the habit of taking the stairs. Alternatively, interventions could instead try to change what or how people think about exercise.  If people do not exercise because they do not view themselves as “athletic”, an intervention could attempt to make people perceive themselves as more athletic, which may make them more inclined to exercise in the future. Policy makers could also use the external reinforcement pathway by developing an intervention that encourages people to join a running club or a workout group. Once people establish friendships within the exercise group, they may continue to exercise, knowing that their friends are exercising and are expecting them to do so as well, which would be a form of external reinforcement.

Our framework emphasizes that the factors that make an intervention effective in the short-term are not necessarily the same factors that allow it to generate persistent behavior change in the long-run. If the goal is to produce long-term changes in people’s behavior, then policy makers may wish to consider which of these pathways are absent in current interventions, and to design future interventions with these pathways explicitly in mind.


Erin Frey is a doctoral student at Harvard Business School. Her research focuses on interventions, misconduct, and organizational responses to unethical behavior. The full text of the paper, "Persistence: How Treatment Effects Persist After Interventions Stop", can be found here.