Master Class 2008: The Irony of Poverty (Class 5)

Sendhil Mullainathan [10.29.08]
Topic:

INTRODUCTION
CLASS ONE • CLASS TWO • CLASS THREE • CLASS FOUR • CLASS FIVE•  CLASS SIX
PHOTO GALLERY 

I want to close a loop, which I'm calling "The Irony of Poverty." On the one hand, lack of slack tells us the poor must make higher quality decisions because they don't have slack to help buffer them with things. But even though they have to supply higher quality decisions, they're in a worse position to supply them because they're depleted. That is the ultimate irony of poverty. You're getting cut twice. You are in an environment where the decisions have to be better, but you're in an environment that by the very nature of that makes it harder for you apply better decisions. —Sendhil Mullainathan

THE IRONY OF POVERTY (Class 5)
A Talk By Sendhil Mullainathan

SENDHIL MULLAINATHAN, a Professor of Economics at Harvard, a recipient of a MacArthur Foundation "genius grant", conducts research on development economics, behavioral economics, and corporate finance. His work concerns creating a psychology of people to improve poverty alleviation programs in developing countries. He is Executive Director of Ideas 42, Institute of Quantitative Social Science, Harvard University. Sendhil Mullainathan's Edge Bio Page


[10:26 minutes]

Daniel Kahneman, Paul Romer, Richard Thaler, Danny Hillis, Jeff Bezos, Sean Parker, Anne Treisman, France LeClerc, Salar Kamangar, George Dyson

A SHORT COURSE IN BEHAVIORAL ECONOMICS
Edge Master Class 2008
Richard Thaler, Sendhil Mullainathan, Daniel Kahneman
Sonoma, CA, July 25-27, 2008

AN EDGE SPECIAL PROJECT 


SENDHIL MULLAINATHAN: I told you there was a bunch of pieces to the puzzle to the poor, and I wanted to finish that argument because I got out the first piece of the puzzle, but there are two other pieces that are important and I want to kind of walk through those pieces. I'll walk through the other piece of the puzzle, and then I'll close on a slide about mental models. I realized in trying to write that up, everything can be shown in this one slide, and it's kind of a stunning fact.

To recap: The first piece of the puzzle is that scarcity buys a broader frame, the need for a broader frame. And one implication of a lot of this which is worth keeping in the back of your mind and that what should resonate for a lot of you is that even positive expenditures are tainted by the trade-offs.

If you're poor and you spend money on a luxury good, you're always at least partly thinking during the moment of consumption about what you're giving up to buy that good. If you are time scarce and you're spending time on a leisure good, you're at least partly thinking about what you're not doing during that time. That's an interesting hedonic aspect of things, which could turn out to be important.

The problem of a broader frame isn't just a planning problem, there's also a hedonic problem that it induces, which is the inability to not think about trade-offs during the consumption. That's an interesting indication. I won't talk more about it because I'm not talking that much about hedonics. But it's important.

Here's the second piece of the puzzle. And this is related to the question of, what is causing, what's preventing the vendors from saving up? Here's a little thought experiment. Eve and Ben both pass by a clothing store. A summer suit draws their eye and in a moment of weakness both make an ill-advised purchase.

Eve goes home and thinks, what a bad purchase. Ben goes home and thinks, what a bad purchase; now I won't have the money to repair my car. Why is this example is useful? It's that temptations have consequences. If you have slack, the consequences are that you've now used up some of your slack on this pointless purchase. And you may feel bad because it was a pointless purchase and it was in the spur of the moment. Of course, we're tempted ...

KAHNEMAN: To give talks.

MULLAINATHAN: To give talks, right, that's right. But we're tempted by certain consumption acts. The poor are tempted by certain consumption acts. Here's an interesting observation. Suppose that the goods that tempt us, so say I make ten times as much as somebody else, suppose the goods that tempt me are ten times as expensive as the goods that tempt the other person. Then we both face the same structural problem. We've taken one problem and scaled it up.

It's like an exchange rate. Just multiply it by ten, but everything else is still the same, so nothing has changed. If, on the other hand, the things that tempt me are only twice as expensive or three times as expensive it's interesting that now I face the fundamentally easier problem than the poorer person. Does that make sense to everybody?

What that means is now that if I cave in, I'm caving in. The psychological experience is the same. I'm giving into something I don't want to give in to, but the monetary consequence is very different. It's going to be in proportion to my slack, my income it's one tenth as big.

Let's make that very concrete. Suppose we're all tempted by fatty foods. I walk by a doughnut shop, and want to buy an apple fritter, I buy the apple fritter. Someone who is very poor—one of these vegetable vendors walks by—have you guys ever had a dhosa? There are these little stands all over Chenai that make dosas. It's a fatty food, tasty. Suppose they get tempted and they buy that. That's three rupees. Three out of fifty; there you go, six percent of your income right there for that day. In some sense the temptation is the same. You feel the desire for something tasty and fatty, but the consequence is much higher.

One way to understand this example is that on top of scarcity there's also an intra-personal tax which is the fact that any of us during the acts of consumption faces a temptation tax that other selves or moments of weakness we blow some cash. Okay? That temptation tax, and the argument I'm trying to make to you is, a temptation tax is a regressive tax. It's not flat, it's effectively very regressive because if one of these were to get tempted they're paying a six percent tax. I'm paying a minuscule, one percent tax.

ROMER: But that would be a temptation that costs dollars. And there are other goods that tempt you with the time.

MULLAINATHAN: Exactly.

ROMER: That's progressive. And the temptations in dollars are regressive. Temptations in time are progressive.

MULLAINATHAN: I've got to be honest, I haven't thought through the temptations in time.

The observation that we make here is that the temptation may not scale to income. And I want to make one important observation. If you think of the returns you might face on any investment that is return R.

In this case paying down your debt or whatever it might be. But that return is scaled by the tax that we face—this temptation tax. And there are two observations to make. The first is that this is not linear in income; the poor face it more. If we do a little calibration exercise we can see how big this thing is. Remember, this is a daily tax. If you're giving in and drinking an extra cup of tea or having a dhosa or some small temptation every day—let's agree that that's four, five percent a day—this is a tax that you're getting hit with at five percent every day. Unlike a discount rate where we think it's like an annualized discount rate, this is something that can very quickly be an annual temptation tax that is crazy big. Does that make sense to everybody?

THALER: I'm confused about this too because I'm wondering how it relates that to the fact that Anne Treisman mentioned yesterday about this finding that your temptation muscle can get tired. It seems like the peasant that's seeing dhosa all day is being tempted in two ways. Whereas you're only being tempted one way.

MULLAINATHAN: I'm definitely coming to that. I'm going to talk about the temptation muscle getting tired, getting stronger.

THALER: Okay, great.

HILLIS: But I didn't understand why it is important that it's daily?

MULLAINATHAN: Let me explain why it's important that it's daily. It's that we read that the extra five percent tax, like eating a little extra dhosa is sensitive. The reason that it's important that it's daily is—suppose that you're moving money between times. I said that in 30 days look at the huge return that you can get. But the point is—so this is Day 1, Day 2, Day 3, Day 4, Day 5, and you move along.

As I move money from Day 1 to Day 2, I save it. I also know that part of what I save, five percent of what I save, is getting burnt. Okay? But notice it's every day. Each guy in the chain is burning five percent. That, too, is cumulative.

BEZOS: But it's not compounding, unless I'm missing something. I guess I'm missing the point you're trying to make. It still adds up to be five percent of your annual income.

MULLAINATHAN: But it is—it's not exactly compounding, but it's getting close to compounding. And let me explain why it's getting close to compounding.

KAHNEMAN: Because you're borrowing, right?

PARKER: Right, because you're borrowing.

ROMER: It's like taking a dollar away from somebody who could earn a rate of return of 1,000 percent a year is much more costly to take a dollar from somebody and go on five percent.

BEZOS: But this is an equivalent statement then to what you said yesterday which is, if I skip a cup of tea for just 30 days, I can double my income.

MULLAINATHAN: Exactly. That's the sense in which it's compounded. The tax isn't compounded, they're not getting poorer and poorer, but on the other hand, you can observe puzzles where it looks like they're leaving crazy amounts of money on the table. And the reason they're leaving crazy amounts of money on the table is you need 30 periods of exercising self-control. And, moreover, every period thereafter to exercise self-control not to eat up the money you just made. That's the sense in which it operates differently than a discount. That's point I'm trying to make.

BEZOS: But the thing that's still a little confusing—it does seem like you can double income. Say, today you have discretionary income. Five percent of your income is discretionary income. I spend all my discretionary income every day on a cup of tea. Because my loan shark is so expensive, if I were to skip tea for 30 days, I could increase my discretion—I'd double my income in just 30 days so I've increased my discretionary income by a factor of 21. I've gone from being able to have one cup of tea a day to being able to have 21 cups of tea a day, if I did the arithmetic right. Why isn't that a fantastic trade that anybody would figure out?

MULLAINATHAN: That's exactly right. Think about how someone's income could change with their spending. Suppose there are two types of goods out there. Goods that you genuinely value that you look forward to and say, I'm so glad I'm consuming them. And goods that are incidental goods that you end up spending money on. They're temptations. Right? You don't look back and say, gosh, I had lots of cups of tea. It's wasted money. Stuff got spent, there was a momentary purchase. You don't value that prospectively or retrospectively. It just got spent.

What I want you to think about is how might consumption of these two goods change as your income changes?

When you're very poor, you have very little income to spend on anything other than food. There's very little spending on temptation goods. At some point you might start to spend more on these temptations. You can start to eat dosas, doughnuts, lots of things that you might not value so much, but they're around you when you have cash and you buy them. And of course because of the regressivity, that flattens out. Once you're over here, you can now spend on things like education, etcetera. So, we’ve got an S-shaped curve: very low spending on temptation goods at deep poverty, but an increasing amount as income increases and then it flattens out as you get wealthier.

If you have somebody like this, notice that being here is very good. Being here is very good. But moving from here to here is ultimately pointless—it looks like your income has doubled, but so much of that extra discretionary income is being spent on wasted things, on things that you don't think are that important.

BEZOS: But what's your evidence for that?

MULLAINATHAN: This is a hypothesis. We have to figure out if this is, in fact, true or false. The evidence for it is ...

BEZOS: Basically, it's people saying—if I can reword what you just said and make sure that I understand it—people say, yes, I could double my income so that I could have 21 cups of tea a day instead of ... —but I will just squander it on stupid things like tea which I don't value and so why even bother?

MULLAINATHAN: And, what's the point?

BEZOS: It's too discouraging. I'll just spend it on stupid things anyway so why should I double my income?

MULLAINATHAN: Yes, exactly. Let me articulate...

BEZOS: I'm skeptical of that.

THALER: Well, it could be your husband will squander it on beer, and then...

ROMER: Yes, but that's a different model.

THALER: Well, that's a version of what Jeff thought.

BEZOS: Then you just hide your money from your husband.

THALER: If you can.

MULLAINATHAN: Well, let me try and convince you that...it's good to be skeptical, but this is true. Let me give you three pieces of evidence.

BEZOS: You're asking for skeptical open-mindedness.

MULLAINATHAN: No, no, skeptical closed mind is good. It gives me a challenge. The first piece of evidence is true equality. These are smart women. They're business people. You can ask them, "you know you can do this, why don't you?" And they say exactly to a tee, "What's the point? I'll have a little more money, money gets wasted." It's stunning.

BEZOS: When they say a little more money do they realize their discretionary income will go up by a factor of 21? What do they mean by—that doesn't seem like a little bit of money.

MULLAINATHAN: Agreed. What they do realize is their income will go up a lot. But what they are saying is what I'm telling you.

One relation to that qualitative evidence is when you look at micro-finance institutions. They go in to give loans to this population. The way they market the loan in most countries is remarkable. They say, "Look, take a loan from us. How are you going to repay it? It's going to come from things you don't care about. One less cup of tea." This cup of tea example is something you hear MFIs say all the time. You're going to have one less cup of tea, you're going to have one less dhosa. Big deal. But in return you can get a washing machine. That's a real thing. That's one piece of evidence.

Let me give you the second piece of evidence that is very related. Granularity matters here. Double your income, but daily receiving twice as much is very different in this world than if I could take all these pieces here and give it to you in one, big lump sum. Because if I can give it to you in big, lump sum then that day you're over here—sure, you take one percent or five percent of that big lump sum and you blow it on a lot of tea, a lot of dosas. But now you still have a lot of money to buy a washing machine and you can go and do things. This is the one consequence that is probably the best piece of evidence we have with this.

Remember that I told you that this population is simultaneously not taking up this opportunity, but they're saving in a ROSCA, a rotating savings and credit association. That's an intriguing piece of evidence. They're taking a zero-return savings vehicle that has an interesting feature. That zero return savings vehicle is taking small and making it big.

That is the second piece of evidence that I have. To understand the investment behavior of this population you have to not just think about returns, but you have to think about the scale of that return. Is it taking lots of small things and making one big thing? Or is it simply giving out a little more every day. Is it just adding, not a lot more, but still every day? Does that make sense to everybody?

THALER: It's essentially a mental accounting problem. Right? Putting it into a lump serves a purpose for them.

MULLAINATHAN: Putting it into a lump means that today or tomorrow, whoever gets the lump may tax a bit ... ten rupees, 20 rupees. But that's a small proportion of tax and that can then be taken to do something with it.

THALER: Right.

MULLAINATHAN: I would be at zero percent return or negative five percent return, be willing to convert five rupees a day into a big lump.

BEZOS: Why isn't the ROSCA solvent then?

MULLAINATHAN: What the ROSCA doesn't solve in this case is you'll notice that the ROSCA solves the problem of taking five rupees a day and allowing me to buy a washing machine. That's great. It won't allow me to pay off my debt. And here's why.

If I pay off my debt I've given all future selves an extra 50 rupees a day and they're all going to waste it in some piddling way, and we're going to fall back down. What would be ideal for this population would be not the ability to repay debt, but for me to come in as a lender and I say, "Look, I'm going to lend to you the same way as the money lender. I still charge you five percent a day, etcetera, etcetera, but I take that five percent a day and I put it into an account, and every two months I pay you 2,000 rupees." Which would have been the 50 rupees accumulated in 50 days. It's the same rate of return on investment, but it's not the same payout structure. I hope I'm being clear. The distinction I'm trying to make is it's not just the rate of return that matters, but is the payout structure a little bit every day or is it one big lump sum?

ROMER: When we looked at this from the perspective of the lenders, we realized that the rates of return were potentially misleading because the transactions costs could dominate. It looked like a very high rate of return and you annualize it.

MULLAINATHAN: I see what you're doing.

ROMER: Could there be something symmetrical going on from the individual point of view as well, that there's something about transactions costs here? That why big things are helpful compared to small things is that the transactions costs proportionally are just a smaller part of the big thing?

MULLAINATHAN: That's a great way of saying it. We agree the lender had a transaction cost and had to deal with you every day. The person has an intra-personal transaction cost every day.

KAHNEMAN: He's resisting.

MULLAINATHAN: He's resisting temptation.

ROMER: Right.

MULLAINATHAN: If I give you payout that's big, once every two months, then you amortize that temptation transaction cost much more effectively because I'm giving it to you every day. It's kind of, unfortunately, the paid intra-personal portion of that.

HILLIS: Isn't it also a social thing in this? Which is that ... these people have a stable pattern, they have friends, they have family, so on. And so if you're deviating from that ... there's expectation that if they get a little bit more money then they have to help solving their friends ...

MULLAINATHAN: This goes back to this issue of how much of the temptation transaction cost is my own spending and how much of it are friends and neighbors coming to me...?

HILLIS: I guess specifically I'm asking the question—would it be different if you changed this, you try to change one person versus if you go in and change everybody synchronously?

MULLAINATHAN: right

THALER: Although the fact that half of them or 40 percent or whatever are self-financing suggests that it's not like there's a hard social norm that you have to do it this way. We have to ask the question why is it that some of them get into this equilibrium where they pay and others don't? That's just part of the puzzle.

KAHNEMAN: I've heard that before and I have the same question. It looks as if some cultures have solved that problem. The Chinese and the Vietnamese don't seem to be in that trap in this country. There seems to be a cultural solution to exactly this problem. And it makes me wonder how much the culture is worth.

MULLAINATHAN: It's a good point. We do the same project in the Philippines and we find exactly the same thing. We're starting to look at other countries.

KAHNEMAN: Clearly the difference between the Chinese immigrants and Vietnamese immigrants, and others, is that they don’t fall into that trap.

ROMER: You can think about this as cultures about saving, but it might be more useful to focus as just cultures about consumption. There is ... this evidence about people who are overweight and the effect of their friends ... that the more you're overweight, your friends also are overweight.

KAHNEMAN: The network effect.

ROMER: Yes, that it influences how much food you eat. If you're around a lot of other people who eat a lot of food you eat more as well.

BEZOS: If you want to be fit hang around fit people. Right.

MULLAINATHAN: Let me tell you the third argument for believing this—which is don't believe it. I don't quite believe it myself either. What we have to think about is what intervention or what experimental design would convince us of this. Let me throw one out and let me see if this can convince you guys.

Right now we've paid off these vendors. Right? And let's suppose that, in fact, they fall back. Suppose we do another treatment? We pay off the vendors, plus we give them a commitment savings account. Well, what's a commitment savings account? Something that they can put money in every day and then they can't access it until it builds up to a certain critical mass.

BEZOS: Time lock savings.

MULLAINATHAN: Time lock savings. The question is would they be much less likely to fall back into this trap?

BEZOS: It's enforced willpower.

MULLAINATHAN: It's financially enforced willpower.

THALER: Yes, a version of that that I had been thinking about is—suppose that what you do is you give the money to the money lender and what the vegetable seller is entitled to is a free thousand rupees every day. Right? They go in, they get their thousand rupees, and then they give it back. And I don't know what the price of that would be. It would be collateralized so there's no risk.

MULLAINATHAN: That's right.

THALER: And then there's no way they can fritter that away.

MULLAINATHAN: There's a very important policy implication in what you're saying. We often look at micro finance institutions giving credit to the poor. And we say, oh, look, they've come in and they've done this. What's the interest rate they charge?

You know they may start by saying 30 percent interest. You're borrowing from moneylender at 70 percent so we say, oh, great, there's a 40 percent savings there. And then we ask can we get 30 down to 12? Or 9. As this example illustrates, that's not a productive thing. By moving from 70 down to 30 you're getting a little bit more cash each week into their pockets instead of an interest payment, which we know they can make.

The more productive thing would be to charge them the higher rate, what they’re willing to pay. But take the difference between this and what you can comfortably charge and put it to some use. For example, you can offer it as a lump sum at the end of the loan term. You can place it into a savings account. The key is that a lump sum may have more value than a stream of small payments.

What this says is that when thinking about the poor you shouldn't be thinking about income you should be thinking about the temporal of distribution of that income, that small being converted to big. Many smalls to one big have a huge gain.

THALER: There's another point I wanted to make about that which is this is absolutely not limited to the very, very poor. 75 percent of Americans get tax refunds. The average tax refund is 2,500 dollars, that's the order of magnitude. Middle class Americans are choosing a ROSCA where they over-withhold on their income tax in order to get a lump sum early in the year. The interest rate is zero, but that's a bargain, as it turns out, because the alternatives are all negative interest rates. And there have been many successful interventions of the experiment with H & R Block of trying to get people the day they're doing their tax return to put some of that money into a savings account. And there's one day a year most Americans have a lump sum, and it's the day they file their tax returns. And it's a unique day that you can do some intervention.

MULLAINATHAN: The other example like this is Christmas bonus. The Christmas bonus is an explicit thing saying you like a lump. You can't convert each of your weekly ... salaries into one big lump but I'll do it for you.

THALER: What we do at Fuller & Thaler, which is a little money management firm that I'm involved with, is we pay bonuses the last week of the year like most people in our industry do. And we go around and talk individually to the half a dozen lower paid—everybody is high paid in money management firms as it turns out—but the relatively lower paid people, and ask them if they want to top off their 401K plan that week. And it's exactly the same principle.

BEZOS: It's a good time to ask.

THALER: It's a good time to ask...you have four thousand left that you could top up right now, do you want to do it? And most of them do.

BEZOS: For the moneylenders the daily transaction costs are a killer. You said they're not making a lot of money. Why don't they take their substantive customers that they trust, that have great track records and say I want to make weekly loans to you instead of daily? I'm going to cut my transaction costs by a factor of seven and... I'll charge you a three percent interest rate.

MULLAINATHAN: There is an analogy to what you suggested that you see everywhere and this rationalizes that behavior which is there is another very important small to big conversation which we see. Even probably as much as tax refunds where we see poor people with big amounts of cash. It's lotteries. Not these million-dollar pay off lotteries, but go to any grocery store or liquor store or community store in a poor neighborhood and watch the people buy lottery tickets.

They don't buy 100-million dollar pay offs. You see 100-million dollar pay off tickets being bought, but you also see these 200-dollar pay offs being bought. That's interesting. Here are these guys going and buying a one dollar ticket whose maximum pay out is 200 dollars with four percent probability which has a real flavor that I would rather have 200 dollars than lots of little one dollars. That's one observation.

If you want to buy a DVD player—this is not a bad strategy for buying a DVD player because when you win you get your DVD player. It's like a savings account.

THALER: Well, it is a bad strategy…

MULLAINATHAN: Well, 25 percent, it's not bad. 25 percent? Look, look ... say you borrowed.

THALER: Right, right, right. It's better than rent to own.

MULLAINATHAN: It's better than rent to own and it's better than credit, because you get screwed if you miss a payment. This is when you've got an extra dollar. I just put it in. Granted, the government is charging you 25 percent for the savings account

THALER: Well, it's 50 percent, but anyway ...

BEZOS: Every cup of tea should come with a chance to win.

MULLAINATHAN: Exactly, like a little scratch off, right? Okay, we're going back to your question. And that's an interesting question. There's an answer to it in the logic of what I just said which is that if I'm a money lender one of the things that I'm also doing is that I know if you walk home with 1,000 rupees that, too, has a chance to be frittered away over the one week period.

You dip into the thousand and it goes down to 850-900, etcetera. You see, as a lender—normally when you're thinking of lenders as suffering only from the moral hazard. I give you the cash and I'm worried that you might do something with it that benefits you, but hurts me. In this case, we have mutuality. If I give you the cash all at once and I say I'll check back in a week and you fritter it away over these five days so the thousand has now been whittled down to 800, you're screwed and I'm screwed because I need that thousand back plus this amount.

BEZOS: But there would be a subset of borrowers. If I'm a rational moneylender I would know my customers well enough and there would be a subset of borrowers. And if it's true that my transaction costs are 80 percent of my costs I would be very focused on my best customers figuring out how to reduce my transaction costs.

MULLAINATHAN: I'm on board with that. There are two dimensions at best—those who are trustworthy and not going to run away with the money. And the second dimension of the best are those who are not going to be tempted to dip into that cash for that one week period.

BEZOS: But I can find that subset of customers.

MULLAINATHAN: Let me articulate why that second group doesn't exist for me as a moneylender. Because organized people who aren't tempted, are also not borrowing.

THALER: There's another less interesting explanation of this which is you already paid these guys to be there every day. And so it doesn't save you any money because the incremental costs—right?... you're paying that moneylender to go every day.

BEZOS: But if it's a selection problem, which is what you've just hypothesized then I'm very pessimistic about your experiment where you're just about to get the data on. If it's a selection problem basically you've selected for a group of people who, no matter what you do, are going to fritter the money away.

MULLAINATHAN: Well, I shouldn't say no matter what you do. Just without a commitment account or without ... that is ... what's interesting about this is ...

BEZOS: Well, you've just said that there are 40 percent of the people who even under the same set of circumstances have figured out how to get themselves out of this problem. And there are 60 percent of people who, looking at these other 40 percent they're saying, "Oh, look at Jim here. Jim figured out how to get himself out of this problem, but I can't do it."

MULLAINATHAN: Right.

BEZOS: That's very pessimistic.

HILLIS: But with the commitment account you've created a new job for the moneylenders, which is to borrow the extra money to put in to put your commitment amount above threshold so that you can get it all out.

MULLAINATHAN: That's a good point. Let me come back to that in one second.

PARKER: Do these moneylenders s offer any other types of loans besides the daily loans to the food market type people? Do they offer weekly loans or monthly loans to other types of customers?

MULLAINATHAN: They do. You see much bigger vendors will be taking out weekly loans. You do definitely see it.

I want to come back to this point, because this is ultimate challenge. Let's just take this hypothesis. And it is true that there is some intra-personal difference or inter-contextual difference that's generating the ability of 40 percent and 60 percent. What's interesting about this perspective is because we're not just building it in and saying these guys are purely impulsive it raises the possibility that there are interventions that can make the 60 percent look like the 40 percent. It may not be the commitment savings account.

BEZOS: But that's the brass ring. That's what you'd love to find.

MULLAINATHAN: That's the brass ring.

And so the challenge is if we think we know the exact nature of temptation, the challenge is to create a financial architecture that effectively immunizes you, or at least partly immunizes you. I wouldn't say all is lost, but the goal would be to create such a product, but that is immune to other market participants undoing it. And this is exactly one of the concerns we have.

We think we're immunizing against temptation, but markets are very good at coming back and saying, hey, you've got this much money here. Why don't I give you a little bit to get the rest out or ... It's exactly what we would worry about and I can understand that.

One other thing before moving on is that the transaction cost point is absolutely essential. And if you think about the finance space and financing this population but people often focus on the product, but oftentimes it's transaction costs that drives the product. In the end it's the other direction because this is a retail business and customer based. It's not a finance business. It's all about getting little bits of cash and moving them back and forth.

BEZOS: Well, the other way to reduce the transaction costs is instead of these fruit stands that are basically sole proprietorships which are very inefficient they should be employees of a company which pays them money instead of financing their own thing and having their own working capital and all that which is very inefficient.

MULLAINATHAN: Exactly. And it's not just the transaction costs that are inefficient. If you think about it—why do I not suffer from the problem that these guys suffer from? It's because I have a job. Somebody has solved it for me. They give me cash every day. We don't realize how much of our social system embodies solutions to our self-control problems that we don't need to fix. It's great. I get cash every month. There's not much I can do that would lead me to not get that cash.

ROMER: Well, let me just try to link this to what Danny was talking about yesterday. If our brain uses analogies and metaphors to deal with things, there is this one very basic issue about food consumption that seems to have this aura of temptation. That when people temporarily go below their usual calorie consumption, it's like your body senses that it's in famine mode and suddenly consuming now becomes incredibly important. And it's the thing that just kills everybody's effort to try and lose weight by using will power to reduce consumption of food.

If that's the analogy, that your brain is using with these other temptation things, I'm not sure there's the threshold that you described in the beginning where it's flat, the initial segment there. It could be that you're in this consume a lot now, fritter away now kind of mechanism right from the beginning.

But then the other thing it made me wonder is with this population it might be interesting to collect data from them about things, how ... what word associations do they use about are they hungry or not? Do they feel like they're in famine mode? And even something weird like an intervention, like suppose you just took these people and you made sure they were fully fed, filled to the brim on food for six months. Make them feel full for six months and see if this changes anything about their financial market behavior.

MULLAINATHAN: That's interesting.... that will matter a lot in a minute.

Let me move to the other piece of the puzzle, or what is another piece. The second piece is that the lack of slack means giving into temptations is more costly for the poor than the rich. But in this other piece, we want to show that in fact the poor find it harder to resist temptations. And this is where I want to talk a little bit about something Anne raised. Here is a thought experiment.

Abe and Ben pass by a pastry shop and see a beautiful dessert. Both of them have been trying to eat healthier. Abe finds the mental reserves and resists temptation. Ben whose mind is occupied with the results of an upcoming critical medical exam succumbs and buys the pastry. This is a very intuitive. If our minds are occupied…

KAHNEMAN: It's experimentally demonstrated to be true.

MULLAINATHAN: Exactly, exactly. Charlie who is stressed about whether he'll be able to find enough money next month for his child's daycare, also succumbs. So far I've been focusing on a particular problem, but now I'm trying to articulate that there can be chronic consequences of being under a situation of scarcity. And one chronic consequence can be the depletion of willpower. Part of it can be because I'm using willpower, but part of it is because if your mind has just spent a lot of time making trade-offs you're tired.

This goes back to what Anne had mentioned earlier that there are notions of ego depletion. Just more broadly that executive control requires energy, executive control of all types. Notice this is not just about self-control. Self-control is one manifestation of executive control. Making a trade-off is another manifestation of executive control. Making computations is another manifestation. But when it all comes together then the point is that these are now intra-linked. If someone is now put into a situation where they have to engage in executive control like making complicated trade offs in another dimension or resisting temptation, they will be worse at this task. Having engaged in one mentally depleting activity, they are now “depleted” and they will find it harder to do a variety of other tasks.

This is speculative, but I want to tell you an interesting way we could get at this and how we can go about modifying it. This is the Stroop test.

You get subjects to read as fast as they can and to name the color of the font. The first color is red, the second color is white, and the third color is green. It's very easy. Go to line three. It's much harder. Okay? This is a task that requires executive control because your mind says white on line three. you should say red, but your mind wants to say white. It's because you have two competing influences. You're reading white and you're seeing red

THALER: And the fact that you've had the first two lines in which word and color match makes it much harder. If you started on line three, you'd have some chance.

MULLAINATHAN: It's very hard. But what's interesting is not just that this is a difficult task. That would just be a curiosity. Suppose we gave them things that look like the first two lines and have them do that task for a minute—and measured how fast they did it. Then gave them things that looked like the next two lines and had them do it. And find the difference in speed. You slow down. It turns out that your slow down is a good measure of how depleted you are. That is a measure of how much executive control you have in the moment. It would be interesting to ask questions like are the poor more depleted at any moment in time?

A set of questions we could start asking—if you could start using measures such as this—is just to even measure the extent of ego depletion in the population. Does that make sense to everybody?

THALER: Are you making a literal comment that people who are depleted would be worse at the Stroop test?

MULLAINATHAN: Yes. I'm making a literal comment.

KAHNEMAN: Well, it sounds like a good hypothesis to me.

MULLAINATHAN: I knew you would say things like this so I'll show you a few examples ... I'll show you one depletion thing.

This is a general phenomenon. Racists get depleted with encounters with blacks. This is a great study. Here subjects were told they were going to do a task and they were led into a different room. And part of the task of the second experimenter that was waiting for them ... and they were told that they'd be delayed with the second task and they were asked to do some work with the other experiment. And the other experiment was a white person. OK? And then we do a Stroop test. Here is the measure of discrimination. We line people up here as a function of how much ... how implicitly anti-Black the population is. And what you find is on the Y axis is the Stroop test performance right out. This has nothing to do with poverty. This is just making an argument that this is a real measure of depletion.

THALER: Yes, yes, yes.

MULLAINATHAN: Even on dimensions like ... as simple as ... Here this population ...

KAHNEMAN: They are running this not with the Black experiment. Are they doing this ... it's the Black experiment they're doing the Stroop with?

MULLAINATHAN: No, the Black experimenter is giving ... is in the middle of a task.

THALER: They exit in between. Yes, that's fine.

MULLAINATHAN: You have this ... you have this ...What you notice is the non-racist —whether they had a Black experimenter or a White experimenter it doesn't matter—they're showing equivalents effects on the Stroop test. The racist they're showing to be much more depleted. They're much slower on the Stroop test after having interacted with the Black experimenter.

BEZOS: And how are they measuring racism?

MULLAINATHAN: This is an interesting effect. Remember Danny Kahneman had talked about priming yesterday. One of the most interesting ways you can measure it directly... You guys should all go to this, it's a lot of fun. It's the implicit attitude test. One of the interesting ways of measuring discrimination is this task that Danny mentioned yesterday about you see a Black face, good word, Black face, bad word, White face, good word, White face, bad word.

KAHNEMAN: And you're not supposed to respond to the faces, just to the words.

MULLAINATHAN: Some people find it easy to just respond to the good words and click the right hand. The way this task works is if you see a good word hit your right knee. If you see a bad word hit your left knee. Then we measure how quickly you can do that task when whenever you see a bad word there's a Black face versus how quickly you do that task whenever you see a bad word where there's a White face. And what you find is that the difference in that speed is pretty interestingly diagnostic of many things. For example, like Dick's example, his diagnostic of your false positive shooting rate... of shooting blacks in that quick responses.

But this is one example. And this is just an interesting example of depletion.

THALER: So it's the same as the Stroop test thing.

MULLAINATHAN: Right.

THALER: Except that you're made race one of the attributes.

MULLAINATHAN: That's exactly right. It's an interference task much like that. What's interesting is not that those two correlate, but the act of having sat with the Black guy

HILLIS: Do we know what depletion is? It's intuitive that you do get mentally depleted, but is there some physical correlate to it like do you have lower glucose levels in your brain?

MULLAINATHAN: In fact, there is interesting evidence that if you give glucose right after one of these tests you can undo the depletion effects. There does seem to be some physiological correlates as well. There is a physical depletion. The glucose thing is interesting. You can give people sugar water or water with Splenda and see a difference in undoing depletion. It's pretty interesting.

THALER: And we could deplete you by putting you on an exercise bike or having you solve crossword puzzles. Are they all the same?

MULLAINATHAN: It's not obvious to me that they are, but in this view they are treated similarly. But I don't know if they are ...

THALER: So, right now we're just calling it depletion.

MULLAINATHAN: That's what most of the literature seems to do. It's not obvious that's accurate. I'm not saying it's inaccurate.

TREISMAN: You do get transfer effects, don't you?I believe there is evidence that the different forms of depletion all have similar effects across different tasks, suggesting that they all use the same internal resource.

KAHNEMAN: It's just work. Consistent to ...

MULLAINATHAN: And this example kind of shows ...

THALER: Yes, but then the exercise wouldn't be.

TREISMAN: Yes, it does.

MULLAINATHAN: No, exercise can deplete.

TREISMAN: Squeezing a handgrip.

THALER: ... Well, but that is not system two being depleted.

TREISMAN: But it's keeping on squeezing when you want to stop

THALER: Oh, okay.

MULLAINATHAN: But it's tiring to exercise because you want to stop. Exercise is the ultimate executive control because you're forcing you body to do something that it doesn't want to do.

Let me show another interesting example of transfer that is more germane to the presentation. Here participants were given a lot of text from a boring book on the biographies of scientists. (I know that it's hard to imagine that there are boring biographies of scientists. But they do exist.)

The no control participants just have to read it out. The behavioral control participants have to smile and be happy and enthusiastic while reading this incredibly mundane thing. What's interesting is at the end of this study the subjects were told, here's ten dollars in cash, it's yours for participating, and you can take it home. We also have products you can buy. These are impulse purchases. You're getting ten dollars and now the guy is offering you a Snickers bar to buy. The question is who buys more products?

What you find interesting is in the behavioral control condition, the subjects who had to smile and be enthusiastic while reading this thing, were more likely to buy something.

BEZOS: If you smile while you're reading this thing you're more impulsive?

MULLAINATHAN: You're less able to ...

TREISMAN: They didn't want to smile so ...

BEZOS: He didn't want to smile so you're depleting stuff so smiling depletes you?

MULLAINATHAN: No, you're depleted by the act of forcing yourself to smile.

BEZOS: the acting depletes you.

THALER: Right, right.

PARKER: But that's not necessarily depletion though. It could be that you're more suggestible for some other reason.

MULLAINATHAN: Let's talk it through. What would be another reason? And I'm not saying this proves the case, but I'm ...

PARKER: It's back to your point about if you're smiling while you watch something, you enjoy it more. You're acting more enthusiastic so maybe that makes you more awake or more socially connected to the person who is selling to you or ...

MULLAINATHAN: Totally right. There could be a priming effect. I picked this particular example, but like Danny was saying earlier, you can do the depletion studies many, many ways. For example, as Anne was saying—I give you a task saying grip this bar and hold it for a minute and you don't want to. That takes a lot of extra control. That also could lead to things like more impulsive purchases and things like that.

THALER: But I thought forcing yourself to smile might make you happy.

If you're reading something boring and somebody forces you to smile shouldn't you be happy afterwards?

MULLAINATHAN: That's a good point. No, no, that's a great point. It's unfortunate I don't have another one to show you, but ...

BEZOS: I'll take your word for it.

MULLAINATHAN: Well, take my word.

KAHNEMAN: If I make you smile by putting a pencil in your mouth that's not effort. You should redo that experiment with a squeeze and the psychologists in the room are predicting you'll get the same results.

MULLAINATHAN: I'll go further. It's not a prediction. it's a fact.

In experiment one—even in the same study the intervention was not reading, but it was something mundane like having to hold this thing for a while and exert control to not. The exact intervention in this case is not important. People do a variety of them. What is interesting is that whatever the act of the volition you have to do at the beginning appears to lead to more purchasing.

KAHNEMAN: I run experiments like that every day in Berkeley. I walk four miles. fast walk every day and there is a speed at which I can no longer think.

BEZOS: Right.

KAHNEMAN: I monitor. If I do a 15-minute mile I can think about what I'm writing. I try to do a 14-minute mile and it's gone. The only thing I can do is walking fast.

BEZOS: Right.

KAHNEMAN: And any ability to think is deliberately gone. It's very much the same thing. That you're training, because you're making yourself do something that your body doesn't want to do, the rest goes.

ROMER: I noticed this same thing about exercise with ... in a more titrated way where I'm on a Stairmaster and the difficulty level of the material that I can read is a function of my heart rate. At a low pace I can read a journal article, but as it goes up, I have to go back to the newspaper.

KAHNEMAN: I was doing that more than 50 years ago in the Army. Try to do fractions while you're on forced march. Beyond a certain speed I could not add fractions.

ROMER: And there's a heart rate at which you can't even read the newspaper.

KAMANGAR: In the gym I found myself watching "The View" while I was exercising. I could never make myself watch that if I wasn't exercising.

BEZOS: Yes, I find I start reading the same paragraph over, and over, and over. And then when I finally get to the point where I've read it like five times and I still don't understand it ...

LECLERC: Can one be chronically depleted? How does it work.

TREISMAN: Well, people who are sick or who are in pain.

ROMER: This is where I want to go back to this thing about the famine response. And this is partly available glucose in your brain, but also you've got glycogen stores in your liver. Once you get into this famine mode your body tries to husband all that glycogen and thus doesn't produce very much glucose for you. It's conceivable you can be chronically in this state because your body is trying to husband the fuel that you turn into it like ...

BEZOS: But one experiment you could do with your subjects in the fruit market is measure their depletion with one of these techniques in their everyday life and then send them on vacation somewhere for a week.

MULLAINATHAN: I'm going to show you an interesting experiment without even having to do that, which will show you that experiment in just one minute.

ROMER: You think it's vacation, I think it's feeding.

BEZOS: I would take them on vacation I would make them relax and not let them do any work.

MULLAINATHAN: Let me argue ... for large parts of the poor what we see is that famine type things are not that important. Just to give you an example of that. If you look at the elasticity of calories consumed with respect to income, what you find is as this type of population gets more income they do spend more on food, but they don't buy that many more calories. They're just buying more sugar, fat.

ROMER: But this famine response, it's a subtle thing. Because you can be very overweight and consume quite a bit of calories and still have yourself in this kind of famine mode. And this is the overweight person who is trying to lose weight that is still eating lots of stuff and is still overcome by hunger.

MULLAINATHAN: That is what I'm trying to explain. I'm trying to explain this subjective sense of scarcity. It's not a physiological famine response. Phenomena are what I'm trying to get at. Which is the notion that it's hard to constantly have to resist ...

ROMER: I'm claiming that there could be a physiological famine sense even in people who are overweight.

KAHNEMAN: But just to complete something I was saying yesterday. The way that John Barge conceptualizes having a goal, that having to goal is like being primed to recognize the things in the environment that are relevant to that goal. And so you can be chronically ready ... that's your response. Which means that it would take more to suppress it. You notice more things that are relevant to your goal. This is what it means. That has a lot to do with being in that state. That there is a state, it could be physiological or it could be psychological, but what mediates it is an extreme sensibility to cues to possible satisfaction.

ROMER: When people go on diets they're thinking about food all the time. And everything they see makes them think of food.

KAHNEMAN: And you don't have to be dieting. Consider those experiments on starvation during World War II. They were thinking of nothing else.

MULLAINATHAN: One of the elements that are not clear, and I was asking Anne about this, is how you add in this notion that attention as a resource is not free to allocate. That is, one of the things that is difficult and we notice in this population is that preoccupation is very important. You are not able to say "Now I'm going to focus on this." The back of your mind is thinking about the fact that, "Gosh, where will I get the money to pay for this or that." It doesn't help that right now you can't do anything about it. It's still coming back to the surface. We haven't made much progress on that for today. Okay. Just to close the loop on that and basically pointing out that it's not that they are more tempted or they may be finding it harder to resist those temptations.

There's a bunch of things to follow up. The third piece of the puzzle—repeated trade-offs and temptation resisting depletes. As you point out, if it's like a muscle it might be getting strong because you're facing more depletions, but it's like carrying a heavy bag all day. Sure, you're more able to carry it on the third day, but you're still carrying a heavier bag.

This has a kind of interesting counterintuitive prediction, which I would love to test which is if you take someone from this population and someone from a middle-class population and put both of them in an equivalently tempting situation, this population may be better able to resist.

BEZOS: Which population?

MULLAINATHAN: The poor vegetable vendors. I'm making the argument here is that they find it hard to resist because they're depleted all the time. But there is this long run view... the muscle is getting stronger. Not stronger enough that they're offset, but they're at least getting stronger. But it's possible if you bring them into the lab, relax them, and bring the middle-class person into the lab, relax them, you'll find that this population has greater self-control in their task.

THALER: If you give them something like Walter Mischel's experiments for adults. These are classic experiments. He had kids age five. And the experiment was they could have one cookie now or two cookies if they wait. And the kids tortured themselves.

BEZOS: You have to do this with my kids.

MULLAINATHAN: He was just a sadist.

HILLIS: I'm curious if there's a test like that correlates with whether people in the 40 percent or the 60 percent.

MULLAINATHAN: The interpersonal difference part ... this would be very interesting. Just not thought about that. It would be great to know are there things that predict who goes where.

THALER: Walter started doing these experiments 40 years ago at Stanford and he's kept track of these kids. And the amount of time you can wait predicts SATs scores. It predicts drug use.

BEZOS: How long ago did they start the experiments?

THALER: Forty years ago.

BEZOS: He's got lots of data.

KAHNEMAN: It also predicts parental income, so it predicts where you're from. It's a big deal. That single question, the lollipop ... it tells a lot about where the children come from.

MULLAINATHAN: I want to close a loop, which I'm calling "The Irony of Poverty." On the one hand, lack of slack tells us the poor must make higher quality decisions because they don't have slack to help buffer them with things. But even though they have to supply higher quality decisions, they're in a worse position to supply them because they're depleted. That is the ultimate irony of poverty. You're getting cut twice. You are in an environment where the decisions have to be better, but you're in an environment that by the very nature of that makes it harder for you apply better decisions.

This is what I think of as a scarcity trap in which under conditions of scarcity your quality of decisions need to go up, but under decisions of scarcity, exactly because you're in a condition of scarcity, you're unable to provide that higher quality decision making.

I want to make two other observations. Let me tell you a few studies as to where we can go with this.

We had talked about let's look at the poor when they're better off seeing the Stroop test changes. It turns out that the agricultural economy provides us with a remarkably interesting test every year. If you look at sugarcane farmers, they get their entire income once a year when their harvest comes in. It's all of a thousand dollars. They're like the food stamp guys, but over 12 months. One consequence is in the last two months of the harvest year, they're struggling to make ends meet. They're poor. The first two months they're bountiful. You have this interesting experiment where the same person is in a condition of scarcity for two months and a condition of bountifulness for two months.

It's very easy for us to go and survey them right before and right after and ask is the Stroop test changing? Are the mental accounting phenomena that I showed you at the beginning changing? Are they engaging in narrow bracketing right after harvest because they've got lots of cash. Is this a good thing to buy? Is this not a good thing to buy. But engaging in broad, global bracketing right after. Did they show higher levels of cortisol, which is a physiological measure of stress. Do things like glucose have a bigger effect on them? Without even having to physically move people we have a natural experiment afforded to us by this context. This would be a very easy and clean test of this hypothesis. It doesn't help us understand fully the interpersonal difference because it's the same person, but at least for the part of it that we care about, the common part, this is a great deal.

I'll tell you one interesting fact about this. (And now you guys will learn something useful for your lives—how to plant and harvest sugarcane.)

It turns out that there are fertilizer applications that you have to put three times, six times during the year. The last application of fertilizer has very high returns. And the last watering of the crop has very high returns. The modal thing we see farmers doing is putting the first few, but not managing to put the last one. They all understand the last one has the highest return, but think of the timing of this production—the highest return is coming when you're at the condition of greatest scarcity.

It's the worst timed thing that you can imagine. Whereas the lowest return thing is coming early when you're in a condition of bountifulness. Lots of guys use ... buy the first rounds of fertilizer and do the first watering, but they don't ... so there's something telling ... and think about this hypothesis already.

Let me tell you the second thing we're doing. This is more ambitious, but it has a potential for higher returns. Again, holding ... trying to create the notion that the same person, but under different conditions of scarcity, we realized there's a set of people. People who are graduating from Ivy League schools who want to go to Teach America. They have two or three months before Teach America starts ... free.

What we're trying to figure out now, we're working on the logistics of this now is can we get this population, move them to a different city, a city that they're not used to. And half of them we'll put in one dorm and we'll give them an allowance—30 days, 30 dollars a day. You get paid weekly. We'll give you 1,500 dollars for the student, whatever. You have to pay us rent, etcetera, etcetera, but kind of bountiful—enough that you can live pretty well. And we tell both populations we are taking away their ATM cards, just live on what we give you.

The other half—we give them a very tight budget. You only get eight dollars. Okay? We're taking the same population and putting them under conditions of scarcity or under conditions of bountifulness. We can recreate things like pay day loans. We can offer all of them the opportunity to get their paycheck three days early, but they'll get ten percent less. We can offer all of them a chance to engage in planning exercises, like your dorm doesn't come with a television.

If you'd like you can buy a television and, you have to make these payments or we'll repossess your television. We can recreate all the elements pretty easily of financial planning in this kind context. Everybody knows ... the poor know they're only getting this much, the rich know they're only getting this much.

We've picked the population, which presumably is selected on high self-control so under that view the scarce guys consume less, the bountiful guys consume more, but there won't be anything. Under this view that there's something inherent to scarcity we expect the people who were given less to start showing behaviors like taking out the pay day loans or buying a television because they're just so bored.

All right, I'm going to get a television because it's silly sitting around this dorm. And then five days later, well, I can't make ends meet, the TV is going to be repossessed. The challenge is—and this is a bit more ambitious, but it seems worth trying—is will we find when we recreate a high IQ, high self-control population, will we start to see the traces of what we see in other populations.

THALER: It's an experimental curiosity question, but would you get them to agree not to supplement their incomes?

MULLAINATHAN: Yes. We would ask them, we would, "Look, we're taking away your ATM cards and it's part of this experiment. We're asking that you live on what we give you." And some people may not do it.

ROMER: You could finance it as a reality TV show.

MULLAINATHAN: Yes, We do have to find a way to finance, but I don't think a reality TV show is the way. I'm excited about this. There's a few logistical issues such as can we convince them to do it. But focusing on these teachers solves the biggest logistical issue. Because it's hard to find a high end population that has three months of free time that would be willing to do this.

We also think that an organization like Teach America would be interested in sending their people to this because if you're going to send a population of Princeton undergrads to go teach inner city kids it's kind of interesting that they've lived through scarcity for three months.

PARKER: Some of this doesn't work because it's only a three-month period. If you want the TV why don't you just take the payday loans —because you know you're getting your ATM card back in three months anyway

MULLAINATHAN: Agreed. Some of it is, with two weeks it wouldn't work. But those three months you'll be spending ... when you're at that boundary it means you won't be able to eat out as much or go drinking with you friends. You are making a true sacrifice for those three months.

I'll tell you an anecdote and then I'll move on from this. When I was in India about a week ago I went to the ATM. It was Sunday, all my employees were gone and I was going to be spending the day by myself. There was no food in the fridge so I went to buy some food and I realized I only have 100 rupees in my wallet. I needed to buy a bottle of water, you can't drink the water, and there was a bunch of crap I had to buy.

And I said, but I only have a hundred—let me go and take some money from the ATM machine. I go and the card doesn't work. Holy shit, I had 100 rupees. I've got to live today on 100 rupees. It was very surprising how quickly my mind entered into a scarcity mode. It was very powerful because in the sense that when I went to store I looked at the items totally differently. If I buy this I won't be able to buy this. It's a very different logic than I'm normally used to. Right?

How quickly I entered the scarcity mode was one surprise. The second surprise was—and this shows the quality of my IQ—it was only after I went through all of this stress that I realized maybe the problem is not with my card—the second question about the computer—maybe it's the ATM machine. Wow, so let me try this other ATM machine. Low and behold, 5,000 rupees pop out. You know what my first instinct was when I got the 5,000 rupees?

Go buy stuff. It was a very interesting response. I felt it in my gut. .

THALER: It was a windfall.

MULLAINATHAN: But it was more. It wasn't that I wanted to buy the things that I resisted. I wanted to engage in the act of buying because I'd spent so much time not buying.

THALER: Meaning an hour.

MULLAINATHAN: No, but it is remarkable. I am very ... I'm fairly frugal. I don't spend on a lot of things and the fact that I end up having ... It's stressful to think today I'm going to live on 100 rupees. I may not be able to have two bottles of water if I wanted it. That's an odd statement. It's more like I can only have one bottle of water today. If I get thirsty ... right?

KAMANGAR: It's the hoarding instinct.

MULLAINATHAN: Oh, that's an interesting thing. Hoarding is interesting. You hear about this from people who have been through the concentration camps, periods of starving. They engage in very small, irrational hoarding.

And the other thing we're exploring right now is whether animal studies can teach us about conditions of scarcity. Some of these responses are very core guttural responses to scarcity.

Okay, I'll close on the mental model.

[From "Breaking Out of the Pocket" by Peter Boone & Simon Johnson, Brookings Global Economy and Development Conference]  

This is just one graph, which summarizes the idea of mental models. Say we'll have a great new technology, a seed that increases yield. And we'll say, wow, we just need to get these farmers to adopt it. And so the adoption problem is basic. And so we spend lots of resources in kind of going and showing them the seed works, lowering costs, subsidizing all of this stuff.

This is a sample of Indians in various states who were asked the question: When your child has diarrhea should you increase, maintain or reduce fluids for that child? Here's the percentage of people saying they should reduce fluids.

I don't know if you guys know that is the worst possible thing you can do for a child is now ... This is a way to kill your child. A good way to kill your child if that's what you want. And you'll notice in states like West Bengal, Ragistan—the number is 40, 50, and 60 percent.

Once you move to the top most populated states you're getting 30, 40, and 50 percent of the population saying I know for a fact that I should reduce. These are not even the ones saying don't know. I excluded them. These are the ones saying I know for a fact. If you look for the percentage that say I should increase it's much smaller. Because a bunch of these say I'm not sure. And why am I making a deal of this? Diarrhea is not a new problem. It's an old problem that's been there for generations. The idea that whether you should give more or less fluids is a problem that's been there for thousands of years.

It's interesting—and India is no exception, I can show you video from other countries—but it's interesting, the mental model the world has settled on is a mental model that suggests that when your child has diarrhea give them less fluids—even though that mental model is exactly wrong and you've had strong feedback. Diarrhea happens at very high instances. It's not a situation where you say, fine, it's a one in a year type thing.

If you look at just the children you get to observe you'll see this often. And infant mortality is not so low that it's a one in a thousand thing. They will have seen children die. The point here is that people can have mental models that are, in one sense, very, very wrong. And this is example is very, very wrong. Do you guys have a sense of why this is a model that's right?

HILLIS: Sympathetic magic. In sympathetic magic if you want an effect you apply something that doesn’t have that property. And so since diarrhea is watery, you don't want to apply something watery, you want to apply something dry.

BEZOS: The water is dirty.

MULLAINATHAN: It turns out even if the water is dirty you're still better off giving the child water.

BEZOS: Right, but a mental model would say I don't want to give dirty water to my sick child.

ROMER: But there's a more direct feedback which is the more water you give the kid, the larger the volume of diarrhea.

MULLAINATHAN: Almost every woman says I don't want to give water, the child is not able to hold it down. This is very intuitive.  Here's the thing, it's a leaky bucket, what's the point of adding it in?

ROMER: Well, and the feedback is you do produce more diarrhea.

MULLAINATHAN: Exactly, exactly.

ROMER: It isn't just a model.

MULLAINATHAN: Yes, it makes sense. The idea behind mental models is that it looks like across this class of issues—whether it's health or new technology adoption—that people have mental models that make a whole lot of sense. They just make sense. Very intuitive, but ...

BEZOS: But they're wrong.

MULLAINATHAN: But they're wrong in very specific instances.

BEZOS: Counter-intuitive.

KAHNEMAN: What explains the differences between the states? Kerala is educated.

MULLAINATHAN: Yes. What's interesting about these populations is they tell us something else kind of interesting which is you can stamp out very specific wrongnesses. Everyone in these states is taught when your child has diarrhea give them more water. We haven't changed the mental model. You've just given them and told them, beat this heuristic into their head. But the mental model that they have of the body as a whole across all health categories hasn't changed. I can put different things up here and you could find different examples.

Let me give you a second example of this and then I'll stop.

More crop technology. Fertilizer has three components —nitrogen, phosphorous, and potassium. N, P, and K. And crops go through phases. There's two important phases. There's this phase where they have vegetative growth. IF you have something like rice, that's the phase where it's sprouting the leaves.

Then there is this phase where it's starting to form the seeds, which is what we care about in this. Let's take that simple model.

Here's a mental model that every farmer has. And the model's not wrong. It's right. The model is how will I look at a field and determine if it's healthy. How green is it? It's an excellent model. It's a summary statistic of crop health in so many ways.

Pests eat away at the leaves so you would see less greenness. Less water means less growth, less green. But greenness is an excellent, excellent measure of yield. But just like good mental models can go wrong, it turns out that nitrogen has one funny feature. As a fertilizer it works by extending the vegetative growth period. That's good up to a point. But after a point, you're cutting into the seeding period so you're reducing the number of seeds. There's an optimal amount of nitrogen use.

And what you find is farmers surely for local yield reasons use way too much nitrogen. Why? Because they say, look, my crop is getting greener. Great. I'll put in some more nitrogen. Of course, you can stamp this out in areas where you tell farmers: stop experimenting and figuring out on your own. You put two bags of nitrogen. You can stamp it out by fiat.

But why am I making a big deal of this? Both of these examples illustrate that if you want people to adopt the new technology or to get a behavioral change, it's important to understand the way in which that behavioral change cuts against the grain of their mental model. Notice it's not about the data you receive. These farmers are receiving annual data on the fact their yield is lower. But they're not looking at yield, they're looking at greenness—as well they should. What's interesting is not what the intervention produces from a real returns point of view, but how that intervention works within that mental model.

LECLERC: It seems like it's also the feedback that comes sooner. It's a short-term feedback as opposed to the long-term feedback.

MULLAINATHAN: That's a good ... that's right.

THALER: Right. It's just like Paul's point about the diarrhea. Giving the kid more water gives you more diarrhea. The immediate feedback is wrong.

MULLAINATHAN: I don't even think it's a myopia issue. It's an issue of when you're engaged in any learning process it would be silly to say I'm going to do this, wait six months, and then I'll see what happens. You don't have that. Most mental models are constructed by creating proxy variables that are good indicators, local proxy variables.

BEZOS: Are you saying if you are trying to teach somebody something to change their behavior you have to understand their existing mental model so that you can explain the flaw in it?

KAHNEMAN: You have to understand the feedback they are getting.

MULLAINATHAN: Right. Exactly.

You have two choices. Let's say I'm doing an experiment that would get farmers to increase nitrogen. One choice is to change their mental model. Okay? That's very difficult because it's so intuitive. In fact, it's correct in some sense. The second choice ...

ROMER: No, but you ... created a brand new mental model for everybody in the room in about one minute with what you put on the sheet there. I don't think it's necessarily as hard to change mental models. And maybe the answer is you just have to give people different models.

MULLAINATHAN: That's totally one feasible approach.

HILLIS: I'd like you to explain to my family about setting the thermostat. How it doesn't cause the temperature to change any faster.

ROMER: But think about getting physicians to wash their hands. There was clear evidence that, if you didn't wash your hands you were going to kill people, but it only got adopted once people had a germ model and then as soon as you had the model ...

MULLAINATHAN: One last comment and then we'll stop.

Which is that one approach is to change the mental model, one is to teach the rule. But the third possibility, which is largely in this core, is to reframe the intervention so that it makes sense within the existing mental model.

We're struggling in this one, but another example about something else called drip irrigation. Normally you flood your field with water and people think that's great. Drip irrigation puts water right at the root. And most people think that's a great water saving device. But it's not a water saving device. It's a yield-enhancing device. And so one way you can frame drip irrigation as to why it's yield enhancing is you tell people you're no longer feeding the leaves, you're just feeding the crop. I'm not changing their mental model, because in their mental model they understand that all crops need water, more water is better.

BEZOS: But they understand that weeds compete with crops.

MULLAINATHAN: Exactly. And so that's the thing where I'm not giving them new data. I'm not going in and saying let me show you drip irrigation works. I'm giving them a way to understand this technology within their mental model. And the main point I'm trying to make here is that these types of reframing exercises, putting them back into the mental model ...

BEZOS: Is that the reason drip irrigation works?

MULLAINATHAN: Is it a good frame, or is that the truth?

BEZOS: Is it a true answer or a fake answer that happens to make sense?

MULLAINATHAN: That is partly true. You are not fighting the weeds. But partly it's the fact of getting it to the root you're getting it more to where it needs than as sitting on the side and has to seep in. It's both.

PARKER: But if you flood the whole field isn't it getting to the roots as well?

LECLERC: But you see, a lot of it's evaporating.

DYSON: You're watering the plant rather than the seeds.

MULLAINATHAN: You're targeting from above... if you flood it all the way up, you can get the same but targeting from the bottom is better.

HILLIS: But would you use a fake, but satisfying answer?

MULLAINATHAN: These are the philosophical questions.

PARKER: Then what is the example there where you're changing the mental model?

THALER: These are the perfect questions for the wrap up.