Chicago Booth Review Podcast Should We Pay People to Walk?

Rebecca Dizon-Ross: So in that population, then you still get these high returns from exercise even among people who are walking already a moderate amount. And so it kind of tails off less than it would in the general population. For this specific population, we're still getting a lot of bang for our buck, even among people who walk a lot.

Hal Weitzman: One way health policymakers can get people to exercise more is to pay them, but should we pay greater amounts to people who exercise more and smaller sums to those who do less?

Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I'm Hal Weitzman, and today I'm talking with Chicago Booth's, Rebecca Dizon-Ross about her research on personalizing incentive payments to try to encourage exercise. Does getting people to choose their own exercise targets and financial compensation encourage them to walk more? Rebecca Dizon-Ross, welcome to the Chicago Booth Review Podcast.

Rebecca Dizon-Ross: Thank you. Delighted to be here.

Hal Weitzman: Well, we wanted to have you on because you've done research about personalization and exercise, so let's talk about it. I mean, at a basic level, if this isn't too silly a question start with, why does personalizing policies make them more effective?

Rebecca Dizon-Ross: Yeah, I mean, so the basic starting point is just that people are all different from each other, right? And so the best policy is just like-

Hal Weitzman: We are all individuals.

Rebecca Dizon-Ross: Yes, exactly. The best policy to say get me to graduate from college might've been really different than the best one to get you to graduate from college. And so as a result, policymakers can do better if instead of just giving everyone the same thing, they personalize the specific policy they're using at the individual level based on the individual's type or how they will respond to that policy.

Hal Weitzman: Okay. But in practice, that's very challenging to do, right? So just tell us a bit about that.

Rebecca Dizon-Ross: Yeah, so the challenge is that people's types are generally not observable. We don't go around with it stamped on our forehead what type of person we are. And so the policymaker has to find some way to figure out what policy would be best for you or what would be best for me. The problem of ascertaining people's types will be particularly challenging if the policymaker wants to personalize a vertically differentiated program, meaning they want to give one person a more generous policy and one person a less generous policy. Because, of course, then when more generous incentives to save, for example.

Because then when the policymaker goes to people and asks, "Oh, what type of person are you?" Everyone will say, "Oh, I'm the type of person who should get the most generous version of this policy." And so that's where it becomes really challenging because then there's this tension between what the policymaker wants, which is to give some people a more generous thing and some people less generous thing and what people want, which is all to get the more generous thing.

Hal Weitzman: Okay. So let's talk about how you put this into practice with your research and talking about getting people to exercise. Tell us how you designed an experiment to find out, to tackle the issue that you just talked about and what you found out.

Rebecca Dizon-Ross: Yeah, so we were building on a previous project where we had evaluated incentives for exercise among people with diabetes in India. And so we had given everyone a step target of 10,000 steps and we'd paid them to meet that step target and we encouraged really-

Hal Weitzman: Walking 10,000 steps daily?

Rebecca Dizon-Ross: Yes, exactly. So we would pay them for the days that they hit that target. And so encouragingly, we saw really large impacts on exercise. And then importantly, it then led to downstream impacts on health. So people had lower blood sugar, which was our intent for the beginning is to impact health. So that was all very promising, but the challenge was 10,000 steps was not the right target for everyone.

So some people were walking almost 10,000 steps without our incentives, and so we were just wasting lots of money paying them to do what they would've done in the absence of incentives. And so it was a promising way to basically increase the cost-effectiveness of the program was to personalize the target based on how much people walk. So give some people who walk a lot a higher target, people who walk not very much a lower target. But the challenge is if you are paying the same amount, then a low step target is kind of a more generous version than a high step target because it pays more for any level of walking. Okay, we need more-

Hal Weitzman: Because you're really paying for the value added part. Is that right?

Rebecca Dizon-Ross: Well, so no, I'm just saying, let's say I'm paying you a dollar to meet a 10,000 step target versus a dollar to meet a 14,000 step target, then everyone will make weekly more money if they get the 10,000 step target, because for any level of walking, they will get paid at least as much for the 10,000 target. So that's what I mean is from a financial incentive, everyone should then want the 10,000 step target. And so it introduces that problem I was talking about before where everyone wants the low step target, but the policymaker wants to get the people who walk a lot into the high target, because they can benefit more there. So that's sort of the basic problem we were trying to solve with our personalization.

Hal Weitzman: Okay. So then tell us how you designed the experiment, what you did, what you found.

Rebecca Dizon-Ross: Yeah. So our main treatment or our main way to personalize in personalize the step target in this setting was we designed these, what we call incentive compatible choice menus, or I'll just call it choice to be short. And so we are basically borrowing a strategy from what firms do when they're trying to do second degree price discrimination is what it's called.

So the idea is that firms would like to personalize the prices that they sell their goods for consumers. They'd like to charge more to the rich people and charge less to the poor people, for example, so that they can sell to everybody. But the challenge is, of course, everybody wants the lowest price. And so the firm has to find some way to keep the wrong people, say the rich people from choosing the low price option, and so they kind of find some way to degrade it.

So for example, they'll offer an early bird special where in order to get the lower price dinner, you have to eat at 5:00 PM, and so no young, rich hipster is going to be cut dead their dinner at 5:00 PM, and so they choose to go later and choose to pay a higher price. And so we borrowed that exact same idea. We thought about, "Well, how can we make it so that people who walk a lot don't want to choose this lower step target?"

And so what we did is we just made it so that lower step target paid a lower incentive amount if people reached it. So then we let people choose their target and they could choose between lower targets that paid out less or higher targets that paid out more. And so what we saw is that people who walk more chose the high target because it paid more, whereas people who walk less choose the lower target. And so hence, we are able to get everyone in the right target for them by offering this choice menu.

Hal Weitzman: Okay. So let the higher achievers get paid more, or the people who are already walking a lot who would be prepared to walk a lot more will get paid a lot more. And those who aren't walking very much normally and would do a little bit more, will get a smaller amount.

Rebecca Dizon-Ross: Exactly.

Hal Weitzman: See, the interesting thing about that is you talk about companies wanting to charge prices, it's all controlled by them, isn't it? The airlines, they traditionally have charged different prices, but they experiment trying different things and they have all these algorithms and everything. Here it's very simple. You are letting the market decide the individuals what they want to do and how they want to be compensated.

Rebecca Dizon-Ross: Yes. Well, I mean, it's a similar idea I think to the market. I think the firms in the market are using these complex algorithms, but then they're allowing the customers to choose which of the products. So it's similar to we didn't have a complex algorithm. We designed it using really simple data, because we just didn't have the data to run a complex algorithm, but I think we as the researchers are the analogs of the firms and then the people just get to choose which one they want similarly.

Hal Weitzman: Okay. So tell us about what you found.

Rebecca Dizon-Ross: Yeah, we found actually that this choice menu worked really well. Probably better than I expected. Our goal was to maximize the amount of exercise we could achieve relative to the cost of the program. The primary cost is the payments that we give to participants. We found that we increased our treatment effect on exercise by 80% relative to just using a kind of one size fits all approach, which was to give everybody the same step target or the same contract. So we got a much larger treatment effect on exercise, which was amazing, and then it didn't cost anymore. So it was just basically getting a lot more bang for our buck by using that choice menu. So that's sort of our headline result about that personalization approach.

Hal Weitzman: Because, I mean, traditionally with the steps, I think in the United States, they have said 10,000. And so you're saying that doesn't really work as effectively as your policy would work?

Rebecca Dizon-Ross: Exactly. So giving everyone 10,000 was not as effective as giving personalized targets. And in fact, we went above 10,000. We were getting people into 10,000, 12,000 and 14,000. And so what we can show you is that basically, the 10,000 step target works much better for the people who don't walk very much, but the 14,000 step target works much better for the people who walk a fair amount in the absence of incentives. And so if you just use a single one, you're getting it wrong for some people, and so our choice menu was able to get it right for a lot more people.

Hal Weitzman: Okay. Well, it makes sense that you would get that kind of result. So talk about, there's some language I wanted you to unpack a little bit from your research, which is about choice versus tagging.

Rebecca Dizon-Ross: Yes.

Hal Weitzman: Well, explain that distinction.

Rebecca Dizon-Ross: Yeah, so I think there's, broadly, when you think about personalization, there's two main kind of buckets of our approaches you can try. One is choice, which we've talked about, and the other though would be personalization based on observables. So basically, the policymaker has some measures about people, maybe they know your gender, maybe they know your blood sugar because you've taken that at the clinic. And so then they take all of the observable measurements they have, and based on that, they choose which one to assign you to. So in that case, individuals don't get to choose what they get. Instead, they're just assigned based on some of their characteristics. And so the better those measures of characteristics are, then the better the policymaker can do.

Hal Weitzman: But you have to collect a huge amount of data to make that really work properly. Right?

Rebecca Dizon-Ross: Exactly. So that's the huge downside of tagging relative to choice is that it is less scalable in that sense. It involves a lot more data, which is becoming perhaps more feasible with the rise of big data, but less so in developing countries like where we worked. And so that's why we were interested in benchmarking against tagging to see how well we could do with choice, because choice is easier to implement. You don't have to collect all this data. As you're saying, you just have to ask people what they want.

And so what we saw with tagging ... So the first thing we did with tagging is we tried to tagging approach where to solve that problem, we just used data that the policymaker would likely already have so that it wouldn't be too hard a lift for them. And so what we saw when we tagged with that-

Hal Weitzman: What kind of data are we talking about?

Rebecca Dizon-Ross: So things like gender, your diagnoses, so these would be health policymakers. So it was basically some of the things that were administrative data in the system. And so we tried to bias it as much as we could towards that approach to do as well. So we allowed it to use several different health measurements like blood sugar, BMI, blood pressure, things like that, as well as your diagnoses and all that and that type of stuff.

And so we tried seeing what would happen if we used all that data personalized and basically it did nothing. It didn't do better than the one size fits all. And it's just because those variables just don't do that well at predicting exactly how much people walk. And so to do better with tagging, you really need to measure exactly how much people walk and then you can do better.

So if you measure how much they walk for a week and then assign them a target based on that, then you can do quite well. But of course, that requires the government to be able to do that, and then it requires that when you do that, people don't then kind of manipulate how much they walk to try to game the system and get a different target.

Anyway, if you use all of those different observables, which would be sort of the best the policymaker could do, they actually only do as well as our choice menu. We thought that was a really promising thing for choice because it says it can do as well as you could do with all of these different things that would be quite costly to collect. And we can do as well with just a simple choice menu that's much more scalable and easy to deliver.

Hal Weitzman: I'm just thinking as you were talking there about how this would actually work. You are trying to help people who have type 2 diabetes, which activity is going to help them, but aren't you going to help those who are least active the most? So in other words, if I only walk 1,000 steps a day, getting me to 2,000 is a massive achievement in a way you might think you want to pay those people more to do that rather than someone who's already walking 10,000 to get them to 14,000. I don't know what the health effects would be, but I would imagine they would be less significant than the doubling the person who walks 1,000.

Rebecca Dizon-Ross: Yeah, so it's a great point. So I think there's a couple reasons why we still think our approach works really well even despite those types of concerns. So the first is that it turns out based on the measurements in the health literature that the health returns to steps are not as concave as you would think. So within the range of walking that we're getting in our sample, there's not too much of a difference, at least according to the health literature in the extra bang you get for your buck of moving someone from 10,000 to 12,000 as there is from 2,000 to 4,000. So it's a little bit less, but it's not hugely less.

So first, there are still significant returns even at that top end, which is kind of surprising to me. Once you get to marathon runners or things like that, then you're in a different ballpark. And also, actually something that's interesting is that, that's particularly true among people with diabetes. So in that population, then you still get these high returns from exercise even among people who are walking already a moderate amount, and so it kind of tails off less than it would in the general population.

So for this specific population, we're still getting a lot of bang for our buck, even among people who walk a lot. So that's one response. Yet there is still slightly more return for the people at the low end. And so if there was a way to perfectly identify who is who and design the things for them, then that would be ideal. The problem is the policymaker just doesn't know exactly which people are which, and so basically they can measure it, but then people can just manipulate it. So there's kind of no perfect way to screen out the people who are high walkers.

So I guess it's like to summarize, first, you wouldn't want to entirely screen out the high walkers, because there's actually still good returns for them. And second, even if you were actually trying to do that, it would be hard. And so what we're doing is instead trying to get as much return as we can from all of the people in the sample.

Hal Weitzman: If you're enjoying this podcast, there's another University of Chicago podcast network show that you should check out. It's called Capitalisn't. Capitalisn't uses the latest economic thinking to zero in on the ways that capitalism is and more often isn't working today. From the morality of a wealth tax to how to reboot healthcare, to who really benefits from ESG, Capitalisn't clearly explains how capitalism could go wrong and what we can do about it. Listen to Capitalisn't, part of the University of Chicago Podcast Network.

Rebecca Dizon-Ross, in the first half of the episode, we talked about your research about personalization exercise that you conducted in India and how you pay people to exercise more, and it worked. Talk about this menu of choices that you gave people, and you mentioned the term incentive compatibility, which is basically, as I understand it, offering people more money for higher targets, which we talked about. But just dig in, tell us a little bit deeper how that influenced the choices that the participants chose. And just overall, you referred a little bit to the performance of the program overall, but tell us a bit more about your results there.

Rebecca Dizon-Ross: Yeah, so basically, as you said by the incentive compatible menu, we mean we made it in people's sort of best interest to choose the high target by paying more for the high target. So because we saw these great results, it raises a question of, did we actually need to do that? Could we have just let people choose between step targets that all pay the same? The higher ones would be less generous, but people might still want them to, let's say, motivate themselves or something like that. And so maybe we could have done as well with just what we call a flat menu, which is just choose your step target, same payment for all.

And so to dig into that and figure out how important the incentive compatibility or the steep menu was, we just simply tried it. We offered people a flat menu and we looked at the impact on the choices that they made, and then we looked at then the impact on actually how much they walked. And so what we saw is that the choices were different.

So we did still see some high walkers choose the highest target even though it was kind of less generous. So it shows that the only thing that people care about is not just money. Some of them want motivation to walk, but we saw fewer people choosing the high target. And importantly, the people who stopped choosing the high target were the high walkers. So it was exactly the people we wanted to get into that high target, just many of them stopped choosing it because we didn't pay more for it. And so as a result-

Hal Weitzman: If this isn't too stupid a question, did they also walk less, or did the target affect how much they walked?

Rebecca Dizon-Ross: Yes. So we then overall saw then that there was less walking. So those results are suggestive. They're not statistically significantly different, but in terms of the numbers, the flat menu looked more like the one size fits all, just giving everyone the same versus our incentive compatible menu. We had quite a bit more steps just taken based on the numbers.

Hal Weitzman: Okay. So having this graduated system does actually make a big difference?

Rebecca Dizon-Ross: Yes, it does seem to, and a lot of it becomes from ... Well, yeah, so it just seems to be that some people choose the high target because they want it and for them we don't need to make it incentive compatible.

Hal Weitzman: But the rule of money is significant, because you're reminding me of when Fitbits came out in the United States, whatever it was five, six years ago or there was a big craze and everyone was taking walking meetings suddenly and randomly walking around office blocks just to get their steps in, but there was no money involved. They just had the goal that they had set themselves.

Rebecca Dizon-Ross: Exactly. In some sense, all of our program is built on the fact that money does matter, because we measure everything relative to a control group that has Fitbits and that we give sort of ... We say, "Hey, try to hit this step target." And we send them congratulatory texts when they do. So we basically show that it matters to give people money in the first place, because we see a big impact on walking when we just pay them to hit that step target instead of just telling them then. And so then what we're showing with this incentive of compatible role is that it also helps to use the money in kind of strategic ways to get people to sort correctly.

Hal Weitzman: That's fascinating. So just saying to people, "Given your age and whatever, profile, you should be walking 15,000 steps a day." That doesn't help as much as paying them to do it.

Rebecca Dizon-Ross: Yes, there's a significant difference from paying. So everything we look at is relative to just being on financial.

Hal Weitzman: Okay. What about if I said to you how many steps do you think you could be walking a day? How about if I just choose a target myself? Is that as powerful?

Rebecca Dizon-Ross: I'm not entirely sure. I guess it would be having to choose a target combined with the payment amount, so it gets a little complicated to think about how that's-

Hal Weitzman: And you're thinking from the perspective of a health policymaker who's trying to use the power that they have to get people to behave in slightly different way.

Rebecca Dizon-Ross: Exactly.

Hal Weitzman: Okay. So you said earlier that you were surprised about some of these findings. What surprised you most?

Rebecca Dizon-Ross: I was surprised by how large an impact we had without an impact on the cost. I think I was hoping for that result. Of course, when you design a study, you hope that you'll get a big impact from sorting people or from doing what you're doing, but the fact that we got sort of an 80% larger treatment effect was a very large effect. And so we were very pleasantly surprised that it worked so well to personalize.

And the reason is, first, what I was saying before that when we use this incentive compatible menu, different people really do make different choices. So the high walkers by and large choose the high target, the low walkers by and large choose the low target, and then we can show in our results that it makes a huge difference what target you're assigned to. So for low walkers, the low target works way better. For high walkers, the high target works way better.

So because there's just these huge differences across people, it means that personalization works really well. And so I think our intuition was that all those effects were there, but we weren't sure. And it turned out that the magnitude was quite large, meaning that there's just very large differences across people, which speaks to then the value of personalization.

Hal Weitzman: Okay. Since you published this, have you had a rush of health policymakers knocking down your door trying to get you to design experiments for them?

Rebecca Dizon-Ross: Not yet. I'm hoping for that, but we also haven't yet published it, so maybe it can still happen.

Hal Weitzman: Okay. Well, hopefully, they'll be listening to this podcast and they'll come and find you.

Rebecca Dizon-Ross: Exactly.

Hal Weitzman: This is fascinating. It seems to be, as you say, using AI might be ever more simpler way of designing these kinds of policies, but I'm interested in beyond health and the specific example that you were talking about trying to combat diabetes. What else do you think ... What are the other areas of policymaking you think this might be useful for, either public policymaking or even for companies and organizations?

Rebecca Dizon-Ross: So companies I think are actually already using these principles a lot more than policymakers. And so in some sense, one of the contributions of our work is to say, "Hey, companies personalize the prices and the packages all the time. Policymakers should be doing something similar." So I think it could be relevant in a broad range of policy domains. So the most direct or easy analogy is to any other programs where the government is paying people to do things that are in their best interest.

In developing countries, there's many incentives for attending education, conditional cash transfer programs. Many programs have incentives for savings, some sort of incentive match, incentives for green technologies, things like that. And so all of those could use basically almost the same design that we did to design menus for people to choose.

But then also in other domains, there's different ways to personalize where you might be designing slightly different types of choice menus. For example, unemployment insurance, you could be designing menus where people sort of trade off the duration of benefits relative to the benefit amount. And so you kind of get people into better contracts for them or-

Hal Weitzman: Just explain a little bit how that would work.

Rebecca Dizon-Ross: So there would be sort of a menu where you trade off. So everyone should want a longer duration of benefits and a higher benefit amount. And so if you just allow people to choose, everyone would choose sort of the thing that has the highest level of both. But instead, you could design a menu with trade-offs where one choice would have a higher benefit amount, but a lower duration. One might have a higher duration, but a lower amount. And so that would allow people to sort based on their characteristics and potentially improve the welfare of the program or things like that.

So that type of option could be at play. Or similarly, if you think of housing wait lists, where people might trade off based on their characteristics, the wait time until they get a house versus the quality of the house and things like this. So menus where you design them so that no option dominates. People have to make those trade-offs, but they make them based on their own preferences, and that can allow for sorting in ways that could potentially improve the efficiency of these programs.

Hal Weitzman: Rebecca Dizon-Ross, thank you so much for coming on the Chicago Booth Review Podcast, talk about your fascinating research about personalization and exercise. I'm going to go and get my steps in.

Rebecca Dizon-Ross: Okay, wonderful. Thank you. A pleasure to be here. Thanks so much.

Hal Weitzman: That's it for this episode of the Chicago Booth Review Podcast, part of the University of Chicago Podcast Network. For more research, analysis, and insights, visit our website at chicagobooth.edu/review. When you're there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research.

This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe, and please do leave us a five-star review. Until next time, I'm Hal Weitzman. Thanks for listening.