SPEAKER 1: I'm pleased to welcome you all to this session. We were very excited. For the last two years, we've been planning it.
The biggest growth industry in economics for the last 30 years has been behavioral economics. It was born here. It's very unusual when a new field is born for people to be in such widespread agreement on who founded it.
We are lucky today to have our former colleague Richard Thaler here to talk to us. He has a new book coming out presently called Misbehaving. It's a history of the behavioral economics movement.
You will be reading a lot about this book in the coming weeks. You'll get a preview of it today. The bookstore was meant to bring copies of the book for him to sign and for you to purchase. Whether they will succeed in doing that, we don't yet know. Nobody has seen clear evidence that they've made it here with them yet.
We have a panel. The panel is pressed for time. So I'm saying very little up here.
The first speaker will be Dick. He'll take about 15 minutes to rehearse some of the history that he's chronicled at great length in his book. Then will come Ted O'Donoghue, our very own professor of economics, the senior behavioral economist at the arts college, a very distinguished scholar. After him comes Sendhil Mullainathan, whom Dick and I taught when he was an undergraduate here at Cornell. He was the student that we both agreed had more capacity to absorb information than either of us had ever met before or since.
Then comes Tom Gilovich, my longtime friend and collaborator. He had saved my life, literally on one occasion. He is a very distinguished psychologist.
And then I will be the last person on the panel. And we will see whether they leave me any time or not. If there's time, I will speak to you too So let me, without any further delay, invite Dick up.
RICHARD THALER: OK. That looks like it. OK, good.
Well, thanks everybody for coming. Thanks, Bob. I must say, one thing about the field is that everybody in it is nice, and we're all friends, which is surprisingly unusual in academia.
So let me start with a reasonable question, which is, why do we need a field called behavioral economics? And this is a question that was asked and answered by one of our forerunners, a guy called Herb Simon, who was a pioneer of about eight different fields, Nobel laureate. And he called the term actually a pleonasm, which, well, bright people like you might know what that means-- it's a redundant phrase-- and asked, well, what other kind of economics could there be?
You would think that economics is about behavior. And then he answers the question. The reason why we need a field of behavioral economics lies in the assumptions of unbehavioral economics.
So what are those assumptions? The core assumption of economics is that people choose by optimizing. So of all the things you could be doing this afternoon, you picked the very best one. Well, that's perfect evidence that the theory is absolutely totally right.
But we could ask whether that's a good description of human behavior. So really, economics is a theory about this guy. And I guess the hand signal is not necessary.
And if humans were all like Spock, then economic theory would be completely accurate, and there would be no need for behavioral economics. Unfortunately, we're a little bit more like this guy. That's actually the best thing in the talk. So I'm going to let that linger.
I guess I have to move on. So I call the character you read about in economics textbooks econs and the people that we interact with humans. How do they differ?
Well, econs are perfect calculators. They have rational expectations. So if they say they're going to get something done by next week, they do.
They have perfect willpower, and are complete jerks. Actually, Ted is, except for the last one, almost a perfect econ. The rest of us, not so much.
So humans are dumber, weaker, weaker-willed, and nicer. Nicer how? Well, here's a little early bit of Ithaca-based research.
Actually, my wife and I were driving around town yesterday, and we found one of these out by Treman State Park. This is an institution that economic theory says can't exist. So it's an honor stand.
They're selling rhubarb. I can tell you there's been some inflation since I took this photo. Because the rhubarb they were advertising at $5. This is $1.50.
But notice the key things about this. The picture isn't great. But you can see, there's some rhubarb here.
There's a box where you're supposed to put your $1.50. And there's a slot on the top. And then there's a lock.
Now what can we learn about human behavior from the existence of this market? Well, one is, enough people will put money in the box to make it worthwhile for the farmer to put some rhubarb out. That's the first thing.
The second is, there are enough jerks that it's necessary to have the Lock I think those farmers have just about the right model of human nature, that most people actually don't want to rip off the farmer, but if there was a big box of cash, somebody, maybe who took too many economics classes, would take it.
So here's another Cornell experience. So shortly after I arrived here, I was asked to teach introductory microeconomics class at Johnson School of Business. And I wrote an exam. And I wanted an exam that would kind of sort people into three piles-- the stars, the people who mostly got it, and the clueless.
And to do that, you need the exam to be hard, to screen out the really good ones, and some easy questions that everybody gets. So you need a wide variance and a tough exam. So I succeeded. I wrote an exam, and the average score was 70.
Now, I should say that we graded on the curve. The average grade was about a B plus. This score made exactly no difference. I mean the mean made no difference to them.
But they were furious with me. And I was a young assistant professor, believe it or not. And I was worried about getting tenure. And I had 200 MBA students furious with me.
So I'm thinking, OK, what can I do? And I didn't want to give up on my hard exam. So my next exam, I made it out of 137 points.
Now the average score on this was a 96. The students were enthralled. Now, the average percentage was-- this was a little harder. They only got 70%. It says 72.
But they were delighted with me. They were ready to vote me Professor of the Year. And the ones who got above 100, this was something approaching ecstasy for them.
Now this is a good example of what, in the book, I refer to as a supposedly irrelevant factor, or a SIF. So economists disagree about all kinds of things. Should we treat our current economic situation by spending more or spending less? You can get economists with great assurance to tell you both answers to that question and all kinds of other things.
But there's one thing economists-- at least traditional economists-- all agree about. There's a long list of things that absolutely do not matter. So one would be the number of total points on an exam.
Making all the five-point questions worth seven points can't possibly make a difference. And there's a long list of other things. So economists will tell you the best possible gift is cash.
It's probably an exam question on Bob's economics class. Explain why cash is the best possible gift. Now most of us don't give cash to our spouses, especially on Valentine's Day, or on our anniversaries. It sends kind of the wrong message.
Framing-- the term that my friends and colleagues-- Amos Tversky we saw the picture of, and Danny Conoman. The way of a question is posed can't matter. So if we tell people there's some operation, and there's a 2% chance you're going to die, that will have exactly the same effect as saying there's a 98% chance you'll live. That turns out to be false.
What we set up as the default shouldn't matter. But it does. Whether the bookstore shows up with the books or whether you would have to walk down to the bookstore to buy a book shouldn't matter much. But my prediction is that, if they don't show up here, exactly no one will buy a book.
And another example is what I call the endowment effect. And so one of the first experiments I ran upon arriving at Cornell was done with this, which is a Cornell insignia mug. Now there's something in economics that's called the Coase Theorem
How it got called a theorem, I'm not sure. But here's what it says, is that initial allocation of resources doesn't matter, as long as people can costlessly trade. Because there will be a market, and then things will get allocated to the people who value them most.
So Danny Conoman and I and a friend of ours, Jack [INAUDIBLE], decided to test this using Cornell mugs. And here's what we did. We went into a class-- actually a law and economics class, in which virtually every lecture includes a statement of the Coase theorem. And we gave every other student one of these mugs.
And then we set up a market in mugs. We asked people, if you have a mug, what's the least you would be willing to sell it? And if you don't have a mug, what's the most you would be willing to pay to buy it?
Now, the Coase theorem says that who got a mug will have no effect on who walks out of the room with a mug. False. Basically, if you've got a mug, you wanted to keep it. And if you didn't get a mug you didn't care whether you had one.
I call that the endowment effect. Now one question you could ask is-- everything I've said so far seems just so obvious. My grandmother would say, you can make a living from this?
Basically everything in behavioral economics seems perfectly obvious to psychologists and to anyone with any sense, but not necessarily to economists. So here's a question that I've been answering for nearly 40 years. They say, yeah, in your experiments, people misbehave. And the truth is that my spouse and my students and my dean and the president and members of Congress, yes, they're all idiots. But that same guy who admits all of that writes down a model of econs. So there's some disconnect there, right?
So why do they do that? And the answer you will get is an answer that, in this book, I call the invisible hand wave. And here's the way that answer goes.
Yeah, yeah, I understand in your experiments, and my wife, and so forth and so on. But in markets-- and then it's my contention that no one's ever finished that argument with both hands in their pockets. It's just not possible to do.
You have to start waving your hands. Because if you think about it, OK, there are markets. So what? So suppose I choose the wrong career or the wrong spouse or I fail to save for retirement or I take out a mortgage that I won't be able to repay if interest rates tick up and house prices tick down. What happens? Nothing.
If I don't save enough for retirement, I won't have much when I get old. But I don't die. I don't disappear. The market doesn't intervene to force me to be rational.
Now, there are market forces, surely. But think of the number of people trying to sell you stuff so you won't save versus the number of people trying to sell you not to spend. Clearly, the sellers of stuff are winning against the people that are trying to sell you not to save.
So how do real markets work? We'd only end up with a world that looks like a world full of Spocks. The technical term for this is, there are no limits to arbitrage if you could always intervene. And the truth is, you can't do that.
If you marry the wrong person, I can't short your marriage. If your favorite football team does something stupid in the draft next week-- I've written a paper about this. It's discussed in the book-- you can't sell that team short. You can't sell that pick short.
The only thing you could do is try to buy that team. Personally, I'm a billion or so short in my ability to do that. And even if you tried to buy the team, you might lose to somebody dumber than you with more money, which describes most NFL owners, actually.
So where are we headed? Fortunately, the field is full of brilliant young scholars, two of which we're about to hear from. And I end the book calling for something that I rudely call evidence-based economics.
Now, again, you might ask Simon's question, what other kinds of economics is there? Well, actually, economic theory is not evidence-based. Economic theory is a theory of econs. And econs don't exist.
Economic theory is based on axioms about how econs would behave. So we have a theory of fictional creatures. We might as well have a theory of Martians.
So it's time for theories based on the behavior of humans. And it's time for policies that are based on the assumption that the world is populated by humans, and not econs. So my book ends with this, which I will let you read while whoever is up next comes to the stage.
TED O'DONOGHUE: OK. So as Bob said at the start, I'm Ted O'Donoghue. I've been a professor in the economics department at Cornell since fall 1997, two years after Dick left. I sometimes think he ran away before I got here.
RICHARD THALER: Against all expectations.
TED O'DONOGHUE: Yes, exactly. So my talk's going to continue the theme of the evolution of behavioral economics, some of the resistance that was met, and how it was overcome. So let me begin with just a little background. I think Dick already hit this this pretty well.
I mean, in general academics, we're all trained to be skeptics. I sometimes think economists are trained to be sort of super skeptics, to be skeptical of everything. Even standard econ results, In addition to behavioral economic results, it really took a lot of convincing. And indeed, it's worth saying that there's still a sizable minority of economists that sort of just don't believe behavioral economics is relevant. So they still need to be convinced.
But because it sets the stage for what I'm going to talk about-- my experience in what I call the second and third wave of behavioral economics-- let me just give a little bit of perspective to this skepticism. So it's worth sort of just taking a moment to think about the way most economists think about the standard economic model, which is, it's a variable model of individuals that will help us make progress in understanding market outcomes and macroeconomic outcomes.
And in particular, it's not meant to be accurate. So on some level, I feel that non-economists who criticize economists for using such an obviously wrong model, in part, are misplaced in that criticism. Because the goal of economics has never been, let's develop the best model of the individual.
The goal of economics is, we start with a model of the individual that's simple and tractable, so we can put it in our macro models and market models, but, at the same time, not too far off. So in some sense, the big question is not, is the standard model of individual wrong? It is.
The big question is, when we use it for our economic analysis, does it give misleading conclusions? Do those simplifications matter? So in some sense, the way I think about all behavioral economics is arguing that, yes, it really does yield misleading conclusions, and, moreover, that we can actually make further progress, that by incorporating ideas from psychology into our analyses, coming up with a better, but still relatively simple, model of individuals, we can reach better economic conclusions.
So in some sense, the mantra that I always talk about with all my students is, the big goal here is to do better economics. So with that background, let me sort of jump in to talk a bit about what I mean by the second and third wave of behavioral economics. So just very quickly, how I think of the first wave-- this is sort of the "economists are skeptics, round one--" is, I think a lot of the initial reaction to some of these behavioral ideas, when Dick started going around, telling everyone about this, is, a lot of this stuff is just noise.
We build these models. Yeah, people might deviate. But it's noise. In some sense, it doesn't matter that much.
And so I feel like a lot of what I think of as the first wave was going around. And a lot of stuff Dick did, and others, was sort of showing, no, it really isn't just noise. This stuff is systematic. It tends to go in the same direction.
It's widespread. And it can be modeled the way economists like to model things. In fact, you can build simple models of the way people behave different from what the standard economic model says.
So I've done a few examples up here. The one I'll highlight a lot moving forward in my examples will be this idea of present bias. It sometimes goes under the label of hyperbolic discounting, a notion that, from a prior perspective, you like to behave in a relatively patient, even-handed manner, but, in the moment, when you're called to act, you tend to indulge immediate gratification and show a bias for now relative to the future. And I'll keep coming back to that.
So when I came onto the scene in the mid-1990s, economists were starting to sort of be OK with, maybe there's some systematic errors-- at least some economists. Not a lot. But a new sort of defense emerged which was that this behavioral stuff-- yes, it might be systematic. It might be widespread. But it's unlikely to be important in economic context.
And I think, often, what they mean by that is, its effects are going to be second order, relative to all the stuff we're currently studying in economics. So the second wave of behavioral economics, which really started in the mid-1990s, was essentially trying to answer this [INAUDIBLE], to say, no, it's not. It can be really important in economic environments.
So specifically, what the second wave is going to consist of is going to be a series of theoretical analyses-- that is, modeling analyses to try to demonstrate that, in fact, when you take people with slightly different models of individual behavior and put them in standard economic models, you can get very different conclusions, in some sense, showing standard models are misleading.
So let me give you two examples. And I'm not going to write down any math and make you work through the model. I just want to tap through what came out of these analyses.
So the first is actually work from David Laibson, an economist at Harvard. So in the mid-1990s, he started looking at present bias in savings consumption decisions, which was a great place to start, because this is a context that economists have studied for years and years and years. How do people make decisions between savings and consumption?
So his big focus was on illiquid assets-- so assets like a house, a retirement plan-- assets where it's difficult to turn it into cash, turn it into something you can use to buy stuff. Now, standard economics had talked about illiquid assets, but in a very simple way. We think, illiquid assets, well, they tend to pay a higher return, usually because of the illiquidity. But they're costly, because sometimes you want to be able to get access to your cash. And that just feeds together to determine your demand for illiquid assets.
What Laibson pointed out was, in fact, if you move to a world where people have present bias, unlike in the standard economic model, now, the illiquidity itself can become valuable to these agents. They'll actually want the illiquidity, in a way where standard agents don't want the illiquidity. They're willing to accept it for an excess return. But these agents want it.
And the intuition is that, if you're prone to indulge immediate gratification, it's a bad thing to have cash in your pocket. It's a bad thing, if you're sort of in the mall and you see a cool looking TV, to be able to buy it on the spot. It would be helpful if you couldn't do that.
So what Laibson went through was, look, if you can tie up your wealth into various forms of illiquid assets, put it in your house, put it in a retirement plan, put it somewhere where you can't get access to it in the mall, then that can help overcome those urges for immediate gratification. He had an important follow-up result as well, again highlighting how introducing present bias can yield two completely different conclusions. He went through the introduction of quick and easy credit into this world.
So imagine proliferation of credit cards, which started in the 1990s, or payday loans-- ways in which people can quickly get credit for additional liquidity. In fact, for people's present bias who have tied up their wealth in illiquid assets, this can have a big impact on consumption and overcome that illiquidity. So to really highlight how this sort of contradicts what standard economics would say, standard economics would say anytime you give people access to new credit, that can only help. Because maybe I have credit constraints, and that helped me overcome them.
But if people have present bias, giving people access to new forms of credit could, in fact, hurt them. It could overcome the ways they're dealing with their self-control problems. So that's one example, where Laibson went through and showed how introducing this idea could really change some of the conclusions that we reach in economics.
A second example I want to talk about comes from my own research with Matthew Raben on procrastination. So he and I, back in the mid to late 1990s, were really interested in the idea that people might complete tasks later than they ought to, and, in particular, later than they themselves think they ought to. Prior perspective, you think you ought to do something, say, on Saturday, but on Saturday, you change your mind.
Standard economic model has almost no way to even ask that question. What does it mean to do something different from when you ought to? But the instant you introduce present bias, now we can talk about this. Because now, we can have a notion that, from a prior perspective, you want to perhaps carry out a task next Saturday, but when next Saturday arrives, immediate gratification and present bias kicks in, and you change your mind and delay.
So in fact, in our work, we sort of explored this. And we reached further conclusions than just that. So one of the points we pointed out was that, look, if people are even a little bit overoptimistic about the future, this idea of procrastination can suddenly become very severe procrastination, in the sense that it's one thing to say, well, I don't want to do it now. I want to do it tomorrow. It's another thing to say that day after day after day after day after day.
So I've thrown a sort of stylized example up here, which I won't read for you. But you can easily imagine a case where, each time you say, I'm going to do it tomorrow rather than today, it has a very small cost. But if you do that 100 times, that adds up to sort of very big costs.
One further thing we showed-- and this really gets at the sense that this stuff can be important-- is, we actually showed that there's a natural logic, at least in some circumstances, where more important tasks might be the ones you're more likely to procrastinate. And the logic for this goes as follows. Look, the more important the task that you face, the more that you actually want to sit down and do a really good job in that task. But the more that you want to sit down and do a really good job on that task, the more likely it is that you'd rather do it tomorrow than today.
So what we had in mind for this was the idea of retirement planning, sort of one of life's most important tasks-- sitting down and planning how to invest your assets for retirement. Well, that's something people really want to do a good job on. As a result, that may be something they never do a good job on.
So the bottom line for these two examples from the second wave-- it's examples like these that started to pretty clearly demonstrate that, as you introduce some of these ideas of the ways in which people deviate from the standard model, you actually get very different conclusions from the standard economic model. The results really are misleading.
So then, a new defense emerged. Everything I just described was very much a theoretical analysis-- kind of "if then." If people had present bias, then how do conclusions change?
So now, economists started to react with, well, yeah, you've convinced us, in principle, this can matter. But will it really show up in economic context? We don't really believe it'll show up. So in some sense, perhaps this goes back to Dick's invisible hand wave. And this is where the third wave came along, which really started 10 to 15 years ago, where, here, the answer is, yes, it really does.
So what this wave was really trying to do was to go to empirical data and try to show that, when you look at economic field context, the stuff really does show up. So I'm getting towards the end of my time. Let me just talk through one example of this, again with regard to present bias.
So this comes from work by Stefano Dellavigna and [INAUDIBLE], two behavioral economists at Berkeley. So the stage here-- imagine that you're a person that wants to go exercise at the gym, and you have two options for how to pay. You can pay $10 per visit to the gym, or you could buy a monthly contract and pay $85.
So they got data from three health clubs that had roughly this structure. There was some nuanced differences, but roughly this structure. And they looked at people in that second group, and they discovered that, in the initial months of their gym membership, these people were paying, on average, $17 per visit.
From a standard economic model, that makes utterly no sense. If you're going to go to the gym five times, why would you want to pay $85 for those five times you go in a month as opposed to paying $10 per visit for those five times? But there's a natural interpretation, in terms of present bias, which is, when you show up at the gym and you're thinking about buying a monthly contract, you think, I'm going to exercise a lot-- 10 times, 15 times, 20 times this month. So that $85 monthly membership makes a lot of sense.
But then, in the moment, when you're sort of sitting on your couch, deciding whether you should go exercise today, present bias kicks IN. You decide not to go. So you end up only going to the gym on those days you're super, super excited about exercise, or maybe a friend calls you.
Moreover, they had a follow-up-- a further result that's sort of consistent with this present bias interpretation. This group that had this monthly contract-- in fact, the monthly contract had an automatic renewal feature, where you put your credit card down, and it was automatically charged every month, and you had to exert a little effort to cancel it. And they see procrastination in canceling.
So in particular, they see, in this group, on average, these people went 2.31 months from their last use of the gym to when they canceled, paying $187 for nothing. So again, sounds a lot like present bias. So let me skip over my last example so there's time for others.
But just very quickly, on what's next, my sense is that the behavioral agenda is really starting to win over at least the majority of economists. There are still pockets that are resisting. So where to go from here? Just two things very quick.
One-- I think Dick handled this a little bit-- is, I think there's a natural sort of fourth wave that really started, I think-- and it's starting to pick up steam now-- of getting serious about policy analysis. This stuff really helps us understand the world. Now, can we use this stuff to help us improve the world?
The other thing that's just sort of interesting to think about where we might be in 10 years is, will we have a disappearing act? So that is to say, will the label of behavioral economics disappear? Because once all economists get convinced, this stuff may just become regular economics, and not something that's sort of new and different. So let me end on that.
SENDHIL MULLAINATHAN: All right. While I'm changing sides, they wanted me to make an announcement that there will be books for sale. Where will they be?
SPEAKER 3: There will be books for sale, which I'm guessing will be signed. Yes? Yes?
RICHARD THALER: Sure.
SENDHIL MULLAINATHAN: I can sign if you'd like, yeah.
RICHARD THALER: Any of us will sign.
SPEAKER 3: Right outside the door.
SENDHIL MULLAINATHAN: Well, let me just start by saying how terrific it is to be back to Cornell. I was an undergrad here, as Bob alluded to. And I think the time here made such an important difference in my life, not just personally, but even professionally, even within economics. I took an amazing class with Bob Frank. I found myself many times asking, what would Bob think here? And it really was helpful.
I also took a class with Dick. I found myself many times asking, what would Dick think here, just so I could do the opposite. So really, both of you have been just super influential.
So let me start. I'm going to be talking about something totally different. I don't know what wave this fits into, Ted, but let's go with the fifth wave. So far, I love the quote from Amos Tversky about human stupidity, or natural stupidity, I guess, versus artificial intelligence. Today, I want to talk a little about human intelligence.
And I'm going to take a completely different look at this. I'm going to be talking about how machine intelligence, I think, is actually teaching us a lot about human intelligence. And that's what I'm going to start with. So ignore everything you heard about people for a minute. Just think about machines. We'll come back to people.
So thinking about machines, I want you to just introspect about one very simple problem-- not about complex decisions, but amongst the things we do most naturally-- language. Think about how amazing it is that you can pull out your phone and you can say to Siri-- I assume you have an iPhone. I think everybody does.
You can say to the Siri, or the equivalent of Android-- I don't know what that's called. I think that's called Dick. You say, tell me the weather, or even just weather.
And Siri understands that what you want to know is the weather in the city you're based on at the time right now. That's a lot of semantic understanding. It's actually pretty impressive that Siri understands that.
So let's just try and understand what's going on there a little bit. And let's do that by starting with a very, very trivial natural language processing task-- the most trivial thing of all. Imagine you're the head of a company, and you're like, you know, my product gets discussed a lot out there. I want to know what people are saying about me.
That's what Dick is always asking me. What are people saying about me? But you want to know that about your product. What are people saying about me?
So you might go on the web. And say this is your product, the Hutzler 571 banana slicer-- a very useful product. Probably makes a good Valentine's Day gift. Certainly better than cash.
And you decided that what you'd like to do is, you'd like to read the reviews of this. Now, you as people can read reviews of the Hutzler 571 and very naturally understand whether this is a good thing people are saying about your product or a bad thing about your product. That's one review. Didn't like it very much. Here's another review-- again, a negative review. Here's another sadly negative review, probably from a European. And this is another negative review. Sorry I'm feeling so negative today.
I actually told Dick that his book should be called "All My Bananas Are Bent the Other Way." But apparently, that's not good enough for him. That'll be my autobiography.
So what's amazing about this task is, we read this. We immediately understand whether it's negative or positive reviews. And we understand them even without the stars, just from the text of it.
That's actually amazing, that, as people, we can do that. So let's try and figure out how to get machines to do that. When Amos Tversky was referring to artificial intelligence in the 1960s, people were like, oh, we can crack this problem. How are we going to crack it?
Well, we do it so naturally. Let's just copy how people do it. After all, what do I do?
I look for certain words, and I say these words are indicative of good reviews-- dazzling, brilliant, cool, gripping, moving. These are, I think, Ted's teaching reviews. I think that's where I got these words.
Or you could say, well, what makes for bad things? These are my teaching reviews-- suck, cliched, slow. I just don't understand how a class can be cliched. You know what I'm saying? I understand awful, but cliched? That felt wrong to me.
You can make these type of lists and say, I introspect, I program it up, and now I look through reviews and screen for words like this. It's a pretty straightforward idea. So you give reviews like this, and you find that this approach really sucks. It says about 60%, 65%-- whatever it is.
Lots of times, it doesn't even have anything to say. In fact, this is a big metaphor for what totally failed in this whole field . If we simply try and introspect and copy what we seem to do so naturally, we don't produce machines that do anything even remotely well. So how is it that Siri developed?
Well, things like Siri develop not by trying to copy humans, but by just turning the whole problem around. And there's actually something profound related to what Dick said. Economics has a very deductive feel. Let's "deduct." Let's deduce from top down.
As scientists, it almost feels like we're fooled by physics that are like, we should deduce from top down. That's what's going on here. Let's deduce what the right thing to do is.
Forget that. Let's not deduce it. Let's not program it.
Let's just induce it. Let's just learn it. Let's turn it into a purely empirical problem. Let's get thousands and thousands and thousands of reviews and just let the algorithm learn what makes a good review what makes a bad review.
Some of what comes out of this procedure is pretty straightforward. Now, these are words that I've made up as examples. But take love, superb, great. These are the type of words that might come up.
Bad. There's no surprise here. So some of what learning happens is just learning. It's just, the obvious stuff happens.
But the difference between deduction and induction is that we can surprise ourselves. I'll give an example. Here's a word that's highly diagnostic of a good review-- still.
This is in the context of movies. The acting was really shitty. Still, I somehow found myself gripped.
Still wouldn't have made your top list if you were deducing. But yet the data tells you. Here's another one.
Question mark-- what you guys think, positive or negative review? Pretty good. Yeah, negative. Who would like this movie?
Here's another one-- exclamation mark. What do you think, positive or negative? Positive, negative?
I think that the story of the internet is that we get really excited when we hate something. And I think that that's a negative review. That's what it is.
Through techniques like this, we can actually get up to pretty high levels, even in the simple task of sentiment analysis-- 95%. Humans are at 100%. As a point, too, keep in mind, 99.9%. And that's been the story of a lot of what you hear about machine intelligence.
It's the story of automation. Let's take tasks that humans do pretty well-- for example, driving a car-- and then let's try and get a machine to drive a car as well as humans-- in some cases, better. Not me, I'm a very good driver. But maybe some of you out there.
So this I think of as automation. The delta is slightly better than what we're going to do. In some cases, it can't be better. We define ground proof. We define what a good review is.
But I think that's a perspective which says, oh, the human brain is really good. Let's try and mimic it. But as Dick pointed out-- now, see, Dick, there's been some movement since you created your slides. Now psychology has gotten into the neuro world. So if I were going to talk about prediction, then I was going to talk about prediction as our ability to look into those concepts where we think humans do badly and we think we can do better.
So then it's almost like, I would compare these two creatures and ask, can we beat humans? No longer are we mimicking human intelligence. If you took everything that you've heard before, if we think humans can fail, well, maybe we can do much better. It's not driving cars or analyzing sentiment.
So what's a task where we can do much better? Well, first of all, I should say one thing. I think when I say the word prediction, I think most people think of something like this. And this is not what I mean.
I think most people think of a macro-prediction. And I don't mean macro-prediction. I mean prediction of the type that we saw with sentiment, where there are millions of individual instances.
The thing about macro is that there's only one draw every quarter. That's not many instances for anyone to learn from, much less a machine to learn from. You want things like billions of movie reviews, or millions of movie reviews.
So let's think of a prediction instance. So here's a prediction instance. Suppose that someone is arrested-- I don't know, for stealing rhubarb from one of those things. I didn't even know that was possible. I'm going and getting some rhubarb for myself.
And now they go before the judge, and the judge has to make a decision. What's going to happen to this person? It's also a situation that's not like a macro-prediction problem. It's actually a problem where we have tons and tons of instances.
12 million people are arrested every year in the US. I know what you're thinking-- a lot of data. That's great. You shouldn't think that.
That's not good. And that's wrong. I just want to put that out there. That's wrong. I certainly never thought that at any moment.
But all that data, all those people, give you data that you can then learn from. You can say, well, what makes a person who's going to? Now what are we going to predict?
Well, the judge's decision about release or not is actually not a decision. It's a prediction. It's a prediction of, will you show up back in trial? Will you commit a crime while out on jail?
That's it. Not only is that what judges tell you they're doing, it's actually what the law says you have to do. Literally, those are the only two things you can look at. So what do we have here? 12 million prediction problem being made by people.
So let's create an algorithm. Let's create an algorithm. Much like we taught it to train sentiment, let's train it to predict not sentiment, but predict whether a person is going to come back.
So you can do that, and this is what you get. On the left-hand side, this is what the algorithm produces if we release nobody. It turns out, if you don't release anybody, there's no crime. I don't know if you guys know this.
But this is a big discovery. I think I found a way to solve the crime problem. They're going to lock us all up.
Then what happens if you release everybody? You get up to here. So how do we compare this to the judge?
Well, what we know is, the judge is at 73% release. So let's do an apples to apples comparison, or rhubarb to rhubarb, as I'm going to call it from now on. A 73% comparison, and let's say that, at 73%, the machine gets it 7.25% crime rate. So about 7.25% will fail to appear or whatever.
Where's the judge? The judge is at 17.45%. That's crazy.
You're not going to do an intervention on these people to reduce the crime rate by this much. This just comes through predicting better. This type of magnitude, our ability to just transform society, I want to call this retail prediction, because it happens all the time.
So many of our most important problems have, at their heart, prediction. I'll give an example. Somebody gets unemployed.
Well, how much should they cut back on consumption? How hard should they start by looking for a job outside of the sector they're in? Should they enter a new training program? Should they sell their house?
You know what all those decisions are predicated on? A prediction about the length of unemployment. And you know what we have tons and tons of data on?
Just like Amazon can tell you you will probably like the Hutzler, we have even more data to tell you our best guess of your unemployment spell is this length. The transformative ability to just move us away as individuals from having to predict the thing that we're not good at to actually having machine prediction.
This shows up when you're going to go buy a house. What are you doing when you're deciding on the size of your house? You're making a prediction about how much you'll earn, whether you're going to get unemployed, how long you'll stay in that job.
You know who is, a little bit, trying to make this prediction, oddly enough, with all the data in the world? They're not trying to help you. But there is a group out there that tries to make a prediction about whether you're buying too big of a house.
It's the bank. Because it's their own money. So they're predicting a narrow question. Will I get my money back?
They could predict the broader question. We have the data. Here's another one.
Health-- you go to your doctor. You're complaining of chest pain. The doctor is trying to decide while you're-- yeah, this one's pretty good.
The doctor is trying to decide, is this chest pain indicative of heart attack? We should do a stress test to find out. Is it indicative of a blood clot, pulmonary embolism? We should do a CT scan to find out?
Or is it just indigestion, in which case, here's some heartburn medication. Go home and come back later. Big decisions at every side.
A CT scan is about $15,000 to $20,000. If we give a CT scan to everybody who had some chest pain, we'd be in a lot of trouble. That would be the health system of the United States. Yeah, that's actually true.
So how do we make that decision? Well, notice, that decision is entirely a prediction. In fact, most of the decisions in health care are predictions.
It's shocking how much we have of this. So this sort of says that maybe prediction and behavioral science go perfectly together. Because we know people make these kind of mistakes.
We know people do very badly with randomness. We know this is exactly the kind of thing people get wrong. And we just learn that. And maybe this is where machine intelligence can really help us.
I think the only problem with this story, which is what I'll conclude on, is that the story's a bit one-sided. It's a bit one-sided because it has the feeling, which is I think the ethos, the meme out there right now, that, oh, machines are just smarter than humans. So let me give you a sentence.
Let's take this sentence. It was the principal of the matter. I hope you all realize-- Dick, this is an error. I'm just helping you out. That's the wrong word. It should be principle-- P-L-E-- of the matter.
I want to end by going back to language. As humans, you all picked up on that immediately, that in that sentence, that word-- you didn't pick up on that? Oh that, is-- wow. Oh, my god. I'm really glad I got a Cornell education while it still is working.
Wow. Sorry, Ted. Hey, I'm sensing an effect here. Look, Dick leaves, Ted comes, then you guys don't know what principle is. All right.
You take a simple thing like this. This is a simple task-- word sense disambiguation. You can imagine there are many words like this-- principal versus principle. And you can say, wow, this is another training thing. Let's train an algorithm to decide whether principal goes here or principle, P-R-I-N-C-I-P-L-E.
There's a paper that's done this, not just with a little bit of data, but with a ton of data. And here, they use four different algorithms. And here, because they use the corpus of sentences from Google Books, they went from having 0.1, 100,000, million, 10 million, 100 million, billion.
Now, this is a sense in which data is really important for machine intelligence. Look how the performance accuracy keeps going up and up as we're moving from 100,000 to, still, a billion instances. Now why am I showing this graph? The dream for this would be if, with very little data, we could do very well.
Because lots of things in the world don't have 100,000 trading instances. But guess what. We have a machine that can do that. It's called a child.
That is the miraculous power of human intelligence. We've talked about what human intelligence fails, and we've talked about how machine intelligence can do much better than it. Sure, when sample sizes get very big, machines seem to do very well.
But we have an astonishing ability to actually work with tiny samples and still produce signal-- brand new things, things we've never seen before. And if you're going to make effective predictions, if you go back to this prediction problem, you have to somehow combine the huge amount of data you have with the fact that there's new things all the time.
For example, what if there's an auto theft game? What if the local economy worsens? What if x happens?
In some sense, there is something that is fundamental that offsets the machine intelligence, which is not human failures, which is what I think we've been talking about, or we've had to talk about, because it's in reference to a truly pointless model, like econs. But it's not about human bias. It's about saying human intelligence is so exceptional.
And if I were to describe what I think the next wave is, it's going to be, I think, a deeper understanding, a movement away from the view of, well, we have this ideal model of people. People fall short. That's super useful. We have to understand our failures, because that's where I think prediction and other things can help us.
But what about our successes? How do we manage to do all of these things in such an amazing and intuitive way? How does a child know principle versus principal?
How is it that, if I said to you, is a biker more likely to like Crystal Pepsi or a soda called Red-- now, you probably have never tasted Crystal Pepsi. You probably didn't even know there was a soda called Red. But you probably all have an opinion on this matter, even though none of you know a biker.
I don't know. You guys might. Because you don't know the words P-A-L and P-L-E. But certainly, most of you don't know a biker. That's the amazing part of human intelligence.
And so what I guess I want to think about, what I've started moving towards thinking about, is actually not just marveling at our failures as people, but starting to marvel at our successes, and how we're going to understand that, and how we're going to understand where we succeed and where we fail miserably. Thank you.
TOM GILOVICH: Mine's configured differently. That's interesting. We've got to go back. Yes. Right.
At the same time that the behavioral revolution was happening in economics, there were parallel changes underway in psychology. One of them, led by some of the very same people responsible for the changes in economics, was the first serious attempt to study happiness scientifically. And it shouldn't come as a surprise that those two changes happened at the same time. Because after all, why do we strive to make better judgments and decisions if not to increase our well-being, including that part of well-being that we call happiness?
So the question I want to raise today is, what have psychologists found out in the study of happiness? A lot of the lessons that have been learned from the scientific study of what makes us happy have been put forward in just a huge number of popular books on the subject. And you probably don't want to read all of them. So I thought I would summarize some of the main findings.
And a lot of the main findings wouldn't surprise you. You have a much better chance of being happy if you have gratifying social relationships. You have a much better chance of being happy if you have a meaningful connection to work. Those are important things to strive for, surely.
But notice you can't just decide to have more gratifying relationships. And so I want to talk today in very practical ways about things you can do-- just decide to do-- from today on out to increase your chances of being happier. I'll talk about those three things that we can each do now, and then turn it over to Bob in the one or two minutes we might have left him to talk about what we can all do together to try to create a happier society.
So the first thing we can all do is shift our expenditures a little bit away from materialist endeavors to more experiential endeavors. This idea, I think, is perfectly well captured in this New Yorker cartoon. It's a cartoon, of course, because we know that no one would say that.
And the idea behind that cartoon is reinforced by all sorts of empirical research. So for example, if you ask people to think about the most recent or most gratifying material or experiential purchase they made over the last year or the last five years, how happy does it make you, as you can see, they think their experience has made them happier than their possessions. Spending money on their experiences was money better spent.
You might not want to put too much credence in this, because you might think, well, people are faking it. They know that it's bad to be materialistic. If I said that you were materialistic, you know that I have not complimented you. And so you might think that these subjects are just saying that.
Well, you can test it in other ways as well-- get people to write about or think about their most gratifying experiential purchase over the last year or material purchase, and then you just assess their mood. It turns out thinking about what you've done rather than what you have makes you happier.
This leads to the question of, why do our experiences make us happier than our material goods? Our material goods are designed to make us happy, and they do, just not as much as our experiences. Our experiences connect us more to other people.
If you and I find out that we have the same smartphone, we're now a little closer than beforehand, but not as close as if we found out that we both went to the same concert, both enjoyed the same movie, both vacationed in the same spot. So social relations are very important to happiness. Experiences advance them more.
They also are a bigger part of our identity. A lot of marketing is trying to get you to connect to your material goods. And we do announce a lot of our identity.
You're announcing lots of your identity as Cornellians with Cornell sweatshirts. So our material goods do provide that, but, again, not as much as our experiences. However much you like your stuff, however much you're identified with it, it's still out there. It's not part of you.
Whereas the experiences really are in here. And arguably, we are the sum total of our experiences. And our experiences contribute more to who we are.
And they're also less likely to prompt deflating social comparisons. Suppose you just bought a new Mac Air and you're really happy with it, and I show up with a new Mac Air, and we start comparing what we have, and mine has a faster processor than yours, a brighter screen. I got it for less money than you did. How much would that bother you? A lot.
Now suppose that you just vacationed in Italy. And I said, oh, where'd you go? Oh, I went to the same place. We start to exchange details.
Where I stayed was better. I had better weather than you did. My rudimentary Italian allowed me to connect with Italians more than you did.
How much would that bother you? Well, that wouldn't sit that well either. But it wouldn't bother you nearly as much.
After all, you have your memories and your experiences. You're not going to trade them for mine. And it's important to understand and it's important that psychologists pursue why our experiences make us happier.
Because-- and you're probably concerned about this already-- this is not a hard and fast dichotomy. There are a lot of things like a bicycle. Is it a material good? Well, of course. On the other hand, it's a material good you use to have experiences.
And so there's no magical distinction between these two. However, it turns out to be the case that the experimental things that you buy tend to produce more of these things than the material things you buy. And if you get a material thing that does that work for you, that's going to make you happy as well.
So lesson number one, just shift your spending a little bit more away from material goods to experiences, and you'll be happier. This is not by any means an anti-materialist rant. We get a lot of enjoyment out of our material goods. I don't want to take that away from anyone.
This is not a call to spend, spend, spend on experiences. A lot of times, as Ted pointed out, we need to save. Saving to have a down payment on a house is certainly not anything I want to undermine. It's just that we often face these dilemmas, all of us with limited income.
I only have so much to spend, and I want to buy this thing or take this trip. And it's very common for people to say, ah, I'd love to go on the trip, but it'll be gone really quickly. At least I'll always have the thing. And although that's materially true, psychologically, it's exactly the opposite.
You'll get used to the thing. You'll adapt to it. You won't even notice that you have it.
Whereas you will always have your experience, as depicted here in what the American Film Institute calls the third most famous line in American film history. And I'll leave you to figure out what the first two are.
- I'm saying it because it's true. Inside of us, we both know you belong with Victor. You're part of his work-- the thing that keeps him going. If that plane leaves the ground and you're not with him, you'll regret it. Maybe not today, and maybe not tomorrow, but soon, and for the rest of your life.
- What about us?
- We'll always have Paris.
TOM GILOVICH: OK, well, that was worth it for those in the first couple of rows. So invest in experiences. You'll always have your experiences.
OK, point number two, this weird word pair, which comes from Teddy Roosevelt, who, as many of you know, suffered from some degree of bipolar disorder. And in order to make himself happier and urge his kids to live a more fulfilling life, he said, "Get action. Get out and do things. Don't fritter away your time. Create, act, take a place wherever you are, and be somebody. Get action."
And what he was urging his kids to do and himself to do has been reinforced in spades by psychological research. There are all sorts of findings supporting the idea that we evolve to be goal-striving creatures. We're happier when we're striving for things, when we're about to get somewhere rather than when we're there.
Just a few representative findings that reinforce this idea-- if you beat people, if a smartphone interrupts them and says, how happy are you right now, if you happen to get them while they're watching television, it's pretty fun. Watching television is pretty Fun but there there's a strong negative correlation between the total amount of time you watch television and how happy you are.
Watching television is like eating donuts. You have one donut, it's great. You have a few, eh, you start to feel terrible.
Same thing is true with television. So leave the couch. Get out in the world. Get action, and you'll be happier.
As many of you know, adolescence is sort of a tough time, in terms of self-esteem. That's particularly true for girls in the modern world. One buffer against that dip in self-esteem is, again, the Roosevelt injunction to get action. Participation in sports is a buffer against that.
If you look at the nature of people's most common long-term regrets, that implies, also, that, in the difficult choice-- should I act or not-- we might want to get out of our comfort zone a little more and take action. So if you ask people, when you look back on your experiences in life and think of those things that you regret, what do you regret more-- those things that you did that you wish you hadn't done, or those things that you didn't do that you wish you had?
And by a large margin, people's biggest regrets in their lives are the things they didn't do, rather than the things that they did. And you can just decide to be more active. It doesn't have to be a physical kind of thing.
It could be intellectual activity, artistic activity, getting out of your comfort zone and striving more. These are the kinds of things that produce happiness. Or as my colleague in the psychology department, Shimon Edelman, put it in his aptly titled book The Happiness of Pursuit, "There are good reasons why the accumulation of experience feels good and why it promotes happiness. Feeling good about mastering novelty through learning appears to be the currency with which the brain is bribed into leaving the couch and venturing outside." So step number two, take action.
And then the third thing I want to leave you with is to mind your peaks and ends when it comes to arranging your hedonic experiences. To put this in concrete detail, imagine you're going to vacation in Fiji, and your finances are such that you can stay there for two weeks, but in order to stay there for two weeks, you've got to rent a hotel that's fine, but not so great-- about a quarter of a mile away from the beach. Or you can spend on a nicer hotel right on the beach, but you can only stay for a week. What do you do? It's a quality-quantity trade-off.
Well, the research evidence makes that a no-brainer. You won't remember how long that you stayed there, but you will remember the intensity of it. So spend on the more intense experience. This was demonstrated in experiments where, let's say, you're a subject. You show up and your job is just to watch a bunch of videos and then rate how happy you are at the time.
And if you're lucky, you're in the positive condition, you watch videos of puppies. Everybody likes that. Soothing waves, waterfalls, Ithaca's gorges and all that, penguins-- penguins are hilarious. So you're watching all of this and indicating how happy you are from moment to moment.
And that's going to vary depending upon how much you like penguins and waterfalls, et cetera. If you're less fortunate, you're in the negative condition, and you see industrial slaughter of pigs-- that's not pleasant-- child drowning, gruesome mutilation, the liberation of Dachau-- just one relentlessly horrible thing after another. And you watch this for varying amounts of time.
As you're doing it, you rate on this little meter of how awful or pleasant it is. And then after it's done, you just ask, how did you like that experience, or how much did you hate that experience? And interesting things emerge here, which is how long it took, how much you had to suffer through that bad stuff, it matters, but not very much. Those are correlation coefficients.
You'll notice they're much closer to zero-- no relationship-- than to one. But what was it like at the moment when you said, boy, those penguins were really hilarious now? However most positively you rated it, that correlates with your retrospective evaluation pretty highly, as does what it was like at the end. So if you're planning that vacation, spend more on an intense vacation rather than to have it longer. You won't remember that.
And by all means, don't waste that last day rushing around, trying to get presents for people back home and dashing off to the airport. Do something special On that last day. Have a good peak as well as a good end.
OK. So that's what you can do to make yourself individually happier. Simple steps we can all just do. And now to think of some things that we can all do to make us happier, I turn it over to Bob.
SPEAKER 1: Well, as any behavioral economist could have predicted, if you have five people on a panel, the last one will end up getting less time than all the others. We have the room, I understand, until 4:!5 or 4:20. I will try to cut my presentation a little bit short. But we'll leave time for questions. And if anybody has to be anywhere, we'll give you a moment to duck out as well.
As Tom said, I'm going to talk about income inequality. It's a subject we've talked about at great length, but largely unproductively. Mostly, the debate has been framed in terms of justice and fairness. And those are terms people have great difficulty finding agreement on. Instead, I'm going to ask what behavioral economics can teach us about the practical consequences of inequality.
And I'm going to try to defend a "man bites dog" conclusion, namely that inequality actually is a bad thing, in purely practical terms. And it's bad not just for poor people and middle class people, but also for rich people, the ostensible beneficiaries of the fact that incomes have been growing mostly at the top for the last 35 years.
One of the things behavioral economists bring to the table that is not part of the standard toolkit in economics is the idea that, when we evaluate things, the frame of reference matters enormously. Strangely, economists make no allowance for that fact. Is your car OK? Well, it just depends on the absolute qualities of the car.
Here's a quick visual experiment. Which of these two vertical lines is longer? You can stare at them and try to make a judgment.
You're mostly strategic thinkers. So you're probably saying to yourself, he wouldn't ask unless they were the same length. And indeed, they are the same length. I made the one on the right after having first made the one on the left, and then hit the Duplicate button.
But if you really think they look the same, then you should schedule a checkup with your neurologist. Something's gone wrong in your brain. Your brain isn't functioning normally.
You should think that the line on the right looks longer, whether you reason that they must be the same. It's that phenomenon at work in every evaluation that we do. Is it cold out?
I grew up in Florida. If it was 60 degrees in November, I knew the answer. It was freezing cold out. We would dress in the warmest clothes we had.
On a March day, 60 degrees out, is it cold in Montreal? They know the answer too. But it's a different answer.
Are you kidding? It's a beautiful day out. I'm out in a t-shirt celebrating that fact.
So is my house OK? This is the kind of question that econs ask themselves. And the answer supposedly depends only on, how big is it? How many rooms does it have? What's the other features of the house?
That was a house-- the one shown here-- very much like the house I lived in when I was a Peace Corps volunteer in rural Nepal long ago. It had two rooms. It had no electricity and running water.
I experienced that house for every day that I lived in it as a perfectly satisfactory house. It never once seemed in any way below what a house should be. I invited people over. I was proud to have them come to my place.
If I lived in that house in Ithaca, I would be ashamed to invite people over. My kids wouldn't want their friends to know where we lived. They would be ashamed.
I don't think I'm a bad person to feel that way. It's just that, in each context, there's a standard. And a house like this in Ithaca would be so far below standard as to be embarrassing. This is, in fact, my house in Ithaca, New York.
If my friends in Nepal could see it, they would think I had taken complete leave of my senses. Why would anybody need such a grand house as that, they would wonder to themselves? Why so many bathrooms? Is there a problem?
But you wouldn't think that if you've lived in an American city and traveled amongst middle income people. It's a totally expected dwelling for people at that time and place. What's happened as a result of the income inequality growth, which has taken the form of almost all the significant income gains going to people at the top of the income ladder, and the higher up you go, the more pronounced the gains, the people at the top have been spending more on everything, including their houses.
Many people wag their fingers at them. Shame on you for building so much bigger house than you need. But that's the wrong way to think about it, I believe.
That's a failure to recognize that frames of reference matter. Rich people travel in a different circle. That's the kind of house that a rich person lives in. It's a normal house, just the same as my house is a normal house for somebody of my station in life.
The middle class doesn't get angry when they see pictures of these houses. On the contrary, they want to see more pictures of them. A yacht? Let's see the yachts too.
We'll be rich someday, they think. Or maybe our kids will be. They're wrong about that.
The level of mobility now is lower in this country than in any other industrialized country. So the expectation that you'll be rich is almost surely mistaken, if you hold it. But people aren't angry.
The people just below the top, though, are influenced by the houses that the rich buy. Maybe it's now the custom that you have to have your daughter's wedding reception at home. So you need a ballroom.
There are people just below the top. They travel in the same social circles. They go to the same weddings. Now they need a ballroom.
And then there's a group just below them. They've got now the obligation to host dinner parties for 24, not 18. They need a bigger dining room. So they build bigger, and so on, all the way down the income ladder, so that the median new house now is about 50% bigger than its counterpart from 1980.
How do people afford to buy the median house then? And why are they buying so much bigger house than they used to buy? It's not because they have more money.
There's been very little income growth in the middle. The median earner now has actually a lower real wage than in 1980. So why are they spending so much more?
The answer is that people like them are spending more. But that's just pushing the question back one level. Why are people like them spending more?
It's because people above them are spending more, and they travel in the same circles. So ultimately, it's because people at the top are spending more. And how do the people in the middle do it? With great difficulty.
They're working every conceivable margin. Counselors report that couples having difficultly always mention financial trouble. In the counties where inequality grew the most, the divorce rates went up the most in those counties, between two adjacent census groups.
You move further from the center. That's another way you can try to make ends meet when you're having trouble. In those cases, people have experienced a greater increase in long commutes in the counties where income inequality grew the most. And the most transparent measure of financial distress-- filing for bankruptcy-- those increases went up the most in counties where the inequality had grown the most.
I'll skip over the toil index. Let me just mention some recent data I came across just last week. This 2010 figure was updated. Now, $31,000 is the average wedding in the US. In Manhattan, 76,000 is the average cost of a wedding.
Not very long ago, it was about a third that much. Does anyone believe that the people getting married today are happier because their weddings cost three times as much? I think that would be a very tall order to demonstrate. There is now a study that indicates that the more you spend on your wedding, the less time will elapse before you're expected to divorce.
And so really, what the simple insight that framing affects matter tells you is that, when all the income concentrates at the top, and people spend more, and then the cascade comes down, and, one step at a time, it influences the spending of people below, much of that is just pure churning.
If all the rich people live in 50,000 square foot mansions, they're no happier-- probably, they're less happy, because it's a pain to manage big properties-- than if they lived in 20,000 square foot mansions. If they all spend 10 million on their daughter's coming-of-age party, they're no happier than if they spend 1 million. Those extra dollars devoted to those things serve only one purpose-- to shift the frame of reference that defines adequate for the category.
So here's a little bit of fiscal magic that we can harvest from a simple insight from behavioral economics. The argument is completely simple-- a few moving parts. Nothing at all controversial about it.
Scrap the income tax. Get rid of it tomorrow. In its place, adopt a steeply progressive consumption tax. Here's how it would work.
You report your income to the IRS the same as you do now. Then you report how much you saved during the year, as you would for a 401K account or another tax-exempt retirement account. The difference between those two numbers-- how much you earned this year and how much you saved this year-- that's how much you spent this year.
And then we would have a big standard deduction to allow for the fact that poor people spend most of their incomes. The tax rate would start low. So here's the Jones family. They earn $50,000. They save 5,000.
Standard deduction, family of four-- 30,000. Their taxable consumption would be 15,000. The tax rate starts off low. So they'd pay a little less than they would under the current income tax.
But then the rate you pay on the next dollar you consume starts to go up. And it goes up and up and up. And there's no restraint here the same way there would be with the income tax.
If we tax income too heavily at the margin, we discourage savings and investment. Here, we're not taxing savings and the corresponding investment. We're actually encouraging saving and investment if we have the rates set higher.
So if you think about a family with a $5 million annual consumption bill, and it was considering putting a $2 million addition onto its mansion, that addition would now cost it $4 million. And even if they're rich, they'd say, well, let's rethink this. Let's have the architect show us a smaller addition.
If everybody scales back, they would all build $1 million additions. They'd pay a million in tax on the addition. So the outlay's the same.
And the additions-- and here's the magic part-- would serve them just as well as before-- maybe better-- because it's relative mansion size that matters beyond a certain point. There's no loss here. They've not been made any less happy as a result of that. But look at what happens when you get the money from that tax.
Here's a thought experiment I'll leave you with. In the left-hand panel, we have the current tax regime. The richest people buy the best car. If you don't buy cars, imagine some other version of this experiment.
The richest people buy the best car, the Ferrari, for $333,000. Here's the regime with a progressive consumption tax. The rich say, I'm not a chump. I'm not going to pay the government tax on a $300,000 car. I'm getting the Porsche 911 Turbo instead.
That's a pretty good car, mind you. Here's the law of diminishing returns with a real vengeance. You've already got most of the features that matter in this car.
You can build a car twice as expensive. Not much better. Many car enthusiasts would say it's not even any better. But if it's better, it's not much better.
And so here are the people in the progressive consumption tax world. They're driving a car that would be just as satisfying to them, because it's the best car there. That's what really is important to them. They're no less happy than those people who are spending twice as much on the car.
But where does the money go? We've now got all the revenue from the tax on this car, and we can repave the roads with that money. So who's happier? This is the thought experiment I'll leave you with-- somebody in this environment, who drives his Ferrari on roads riddled with foot-deep potholes-- that's roughly the current environment-- or somebody who drives a Porsche-- a lesser car, by assumption-- not much-- on smoothly maintained ones.
I've yet to meet the person who says that the Ferrari driver on pothole-ridden roads would be happier. And yet the cost of the package-- the Porsche on smooth roads-- would be lower, substantially, than the alternative package, which is the Ferrari, and shortchange the public sector. We don't realize how much better off we could be if we-- and Tom mentioned he was going to tell you what you could do.
I can't do anything about this. But if we vote for a change in tax policy, then we can change the frame of reference that defines what we need to make us feel like we've got good enough. And we can spend more on all those things that we're not spending money on, because we think we can't afford it. People think that if they pay another dollar in tax, oh, that'll make it harder for them to get what they want. It's natural to think that.
Every time you've experienced an event that left you with less money-- a divorce, a house fire, a loss of job, whatever-- you had less money, but everybody else have the same amount of money, and you were less able to get what you want. When we each pay a little more in tax, then we all have a little less money. And the things that we want that we have to bid for end up going to the same people as before. There are only so many apartments with views of Central Park.
And no matter how much tax the rich pay, those apartments are going to end up in the same hands as before. Because taxes don't affect relative purchasing power. And it's all about frames of reference.
So we have gone over our time by 10 minutes. We've got some time left that we still have access to the room. There's another event in Bailey Hall. Is that right? If people need to get to another event, by all means, feel free to leave. But thank you very much for your time and attention, all of you.
We've received your request
You will be notified by email when the transcript and captions are available. The process may take up to 5 business days. Please contact firstname.lastname@example.org if you have any questions about this request.
Departures from traditional economic models are the focus of behavioral economics, one of the most important new areas of research in economics to have emerged in the last three decades. Cornell was the birthplace of this field and remains a leading center for work in this area, which continues to change beliefs long thought to have been settled in the discipline.
Top behavioral economics pioneers discuss why traditional economic models do such a poor job of predicting behavior. Part of Cornell's sesquicentennial celebration, April 24-27, 2015.
Moderator: Robert Frank (Henrietta Johnson Louis Professor of Management and Professor of Economics).
Panelists: Richard Thaler (Ralph and Dorothy Keller Distinguished Service Professor of Behavioral Science and Economics, University of Chicago); Sendhil Mullainathan '93 (Professor of Economics, Harvard University); Tom Gilovich (Professor, Department of Psychology); and Ted O'Donoghue (Professor, Department of Economics).