This is Hidden Brain. I'm Shankar Vedanta. My guest today is Keith Chen. He's a behavioral economist at UCLA.
He's also the head of Economic Research at Uber, the right sharing company. Keith's going to talk about some of the behavioral anomalies that Uber has observed, and we're then going to talk about some of Keith's earlier work, which explored the evolutionary and cultural origins of certain biases and heuristics. Keith Chen, welcome to Hidden Brain. Thank you so much for having me on the big fan of the podcast.
I want to start by talking about search pricing. Uber charges more when demand is high based on the idea that there's going to draw more drivers into the pool and increase the supply of rights. Now, this makes perfect sense from the point of view for traditional economists, but you're a behavioral economist and you must know there's something about charging different prices for the same products that rubs customers the wrong way. They say, Hey, five minutes ago this ride was $10, now it's 20.
As you can imagine, I hear over and over again both at my dinner table and at family gatherings that search pricing can feel very unfair to customers, but it's been a really integral part of Uber success, precisely because the whole goal of the company was to replace a really frustrating experience with taxi, with a service that's just ultimately reliable. The only way to do that, the only way to be able to get basically everyone who lives in a dense part of a city car within five minutes was to do that through dynamic pricing, through giving drivers a very, very strong incentive to want to get to the places where they're needed the most, and also to get riders who could afford to wait a little longer, say they're in a bar, to say, well, you know, if it's more expensive to take a ride right now, go ahead and relax and sit back. If you can wait 15 minutes, the drink's on the person who's got to go now. Right.
The interesting thing, Keith, is that when I think about my own behavior, when Uber tells me that the price is now 1.8 times the regular price, I notice that and I factor that in and there's a part of me that feels it's a little unfair. When I'm waiting for just a taxi cab and now the taxi cab doesn't show up, I don't actually think of someone whom I can blame or someone whom I hold responsible, even though it actually has a bigger effect on me, that now I actually have to wait two hours or I can get a cab at all. Yeah, that's absolutely right. And you know, this basic question of how psychologically painful kind of the experience of paying a price is, is something that I worry about every day, especially because I actually think that it's one of the reasons that we've grown so fast.
It's one of the reasons that we've been able to displace taxi so quickly is because in a taxi, you sit in the car and when you're trapped in traffic, you literally watch your money taking away, like in front of you. You're just kind of forced to watch it. There's nothing else to see. It's hypnotic, absolutely.
It's hypnotic and it's the worst possible psychological experience. You know, taxis and gas pumps, right? Or the two places where you just watch your money take away. And the typical Uber experience, you know, you just hit a button, get in the car, and if it's not surging, you don't even need to know what you paid until tomorrow morning if you want to open the email, and if you kind of trust the Uber's the kind of cheapest possible option, you don't even need to look.
So you found that psychology plays a role in how search pricing works because there are places where it works in the way that a traditional economist would predict it would work and there are places where it breaks down. Definitely. So just like traditional economics would predict, as you raise the price, you know, search pricing starts to dampen demand. You know, when you go from kind of a surge of 1x, meaning no surge to 1.2, you actually see a very, very large drop in demand, okay?
And that initial drop in demand, actually early on, when we first started search pricing at Uber, going from 1x to 1.2x, you would see a 27 percentage point dropping people who would request. After some time though, like both after a surge has been in the city for a while and after people have gotten a little used to it, that drops to 7%. So people start getting used to it. It's not such an alien experience anymore.
They may not love you at the company because of it, but they're not quite as kind of put off by it as normal. And then as you take up the price further and further, you see further and further drops in demand. So 1.2, 1.3, 1.4, you know, people like surge less and less because they understand that they're paying more. The surprising thing is there is a very, very strong round number effect which we detect.
So when you go from 1.9 to 2.0, you see six times larger of a drop in demand than you saw from going from 1.8 to 1.9. So the amount more that you're paying for the trip is the same between those two steps, but 2.0 just feels viscerally larger to people. It just seems a lot like them. It's very easy to understand.
Everyone understands that paying twice as much for this trip as I would have. I saw a paper on the came out on National Bureau of Economic Research a couple of months ago. This is by Matthew Bacchus, Tom Blake and Stephen Tedellis, where they looked at pricing on eBay. And they found something that sounds similar, which is they found that when you price things with round numbers, those things tend to sell faster partly because people believe that sellers who price things at round numbers aren't really wedded to those prices.
They basically put those prices on because they want to move something quickly and therefore they're actually willing to negotiate with you because they actually don't care so much about what the actual price is. Whereas when you pick a very specific price point, people say, this couch costs $74.26, some thought must have gone into this. And I don't have much room to negotiate. And there's no back and forth where somebody prices their couch at $100.
You think, well, why don't I just counter with $50? Then the conversation gets starting from there. We see exactly the same thing at Uber. And I think that's the main explanation for something really, really puzzling.
And that is, more people will take a ride out of Surge multiplier of 2.1 than would take a ride at 2. And it describes you how between 1.9 and 2, a lot of people stop taking rides. If anything, people take more rides at 2.1 than they did at 2. When it's more expensive.
When it's more expensive. It's as if they're telling you, I would rather pay you 2.1 times the normal price than I would to. And I think just like your intuition on negotiations on eBay, that's exactly the same intuition. I think the drives of the behavior here, when you tell someone, your trip is going to be two times more than it normally costs, they think, wow, that's capricious and unfair.
Someone just made that up. They must have seen it was raining and just decided to mess with me. Whereas if you say, oh, your trip is going to be 2.1 times more than it normally does, wow, there must be some smart algorithm in the background here that doesn't work. It doesn't seem quite as unfair.
We had the behavioral economist Richard Thaler on the podcast some time ago. And one of the things he was talking about is why it's often hard to find a cab on a rainy evening. And his theory is that this has something to do with what he calls mental accounting, which is the cab driver has a number in his or her head about how much money he wants to make on a given day. So the cab driver says, my expenses are going to be 100 bucks a day, I want to make $100 over that.
And so when I hit 200, I go home. And on a rainy day demand is higher, so you hit 200 faster. And so the cab driver goes home. And so you end up with fewer cabs with more demand.
Now you've done some work looking at dynamic pricing and search pricing and you're finding that actually that isn't the case with Uber driver. Uber drivers do not necessarily go home when it's actually a smart time to be driving. Yeah, exactly right. I find very different results than Dick does with New York City cab drivers.
And I think I understand a little bit of the psychology as to why we're finding different things. So exactly as you said, if you're a New York City cab driver, you're going to get the same amount for every trip no matter what. It's a regulated fare. So when it's raining, the only thing that's really changing is you're just picking up more people.
So in that world, it does feel very salient. Many of them are collecting cash. You can just see once there's a pile of $200 in the front seat, I'm just going to call it a day. Whereas on Uber, the main way that we incentivize drivers to move to the places that riders need them and to stay out a little longer if they can afford to is through search pricing.
And so I think that's very, very salient. So if you're an Uber driver and say you were planning to do about a two-hour shift this afternoon, you're driving around, but all of a sudden, unexpectedly it starts to surge 2.1 times. Like every trip you're going to make, I can't believe my mind just went to 2.1. But you're an Uber driver and it's surging.
You're going to get twice as much for every additional trip you do. What I actually see in our Uber data is that even compared to yourself a week ago in exactly this situation, Uber drivers will double triple the length of the shift that they were planning to do. If it's surging 2x, if it's surging 3x, they're just going to stay out because they can make a lot of money right now. They can take the whole weekend off if they put in another few hours today.
When you yourself use Uber as a rider, what is it you pay attention to? I mean, do you use your own trips as sort of research into how the company is working? You must. Oh my god, I do all of the time.
And it frustrates me because I kind of see so many psychological biases in my own life. So for example, partially because I think it's just one of our most exciting products, I take Uber Pool a lot. So Uber Pool is the service by which now up to three passengers can share the same Uber car. And when you're in the car, somebody just two blocks ahead, it just happens to be going along the exact same route that you are.
And then just suddenly tell your Uber driver to pull over and take it this new person. And then because we can get people to share rides and get more people into seats, we can reduce the price. And on average, Uber Pool can save up to 50% for most people's rides. Not to mention saving on gas and emissions.
And all the environmental issues. And increases driver pay because literally now drivers can be constantly utilized. They can be constantly making money and they don't have any of that downtime of driving to someone or sitting idle somewhere. But a big part of making Uber Pool work is minimizing the amount of psychological frustration that people have with that experience.
So for example, I find for myself that initially when we were writing those Uber Pool algorithms, we took a very rational view. We want to pass people the most amount of savings that we can and inconvenience them as little as possible. We think convenience usually being just really focused on how much time is it going to get you from your point A to your point B. One of the things that we started to discover is that I feel very viscerally when I'm in an Uber Pool is that there's something really, really psychologically painful about going backwards.
So even if this is an amazing match for you, forcing your car to make three right turns and circle around the block and for a short period drive backwards, drive away from your destination is just three times more painful than you would have expected just from the added time. So we try and avoid that as much as possible. I feel there will be times when I miss an exit on the freeway and you know that you have to go to the next exit and then turn it on and come back. And every inch of the way as you're going to the next exit, you're reminding yourself, I'm going to have to retrace this inch.
I'm going to have to retrace this inch. I made a mistake. It's true you're really beat up on yourself when you feel like you're doing something that's taking you in the wrong direction. Yeah, I'll do crazy things.
Like I realize the other day that you know there's two airports that I can fly out of. I have to fly south and there's two airports I can fly out of. One of them is north of me and one of them is south of me. I can't bring myself to book a flight out of the airport that's north of me because I feel like I'll spend the whole drive like going in the wrong direction.
One of the things I want to talk to you about is that Uber is collecting massive amounts of information on what people are doing. And I think a lot of people don't quite realize how powerful this information is. I remember doing a story some months ago looking at some social science research analyzing how people were using cell phones in a poor country and how the way people use their cell phones and whether they kept their cell phones topped off. So this is a country where you sort of prepay your cell phone.
Whether you keep your cell phone prepaid and whether you have a lot of incoming calls or you have a lot of outgoing calls, how wide your network is, all these things can predict with a remarkable degree of accuracy, your creditworthiness. And so if a bank sort of just looks at your cell phone usage data, it can make remarkably accurate predictions about whether you're likely to repay a loan or not repay a loan. And I mean who would think that just the way that you're using your cell phone could tell a bank that they should give you a loan to start a new business. And I feel like Uber has sort of similar access to vast amounts of data.
I mean they've been reports for example of, you know, so Uber broadly knows where I live and where I work because I'm often taking caps back and forth between those two locations. But let's say one day I leave work and instead of going home I go to another location which is known to be the address of a bar. And then a few hours later I go to another location which is not my home address. And then I come to work the next morning.
You know, you can draw the conclusion of what I was doing that evening. It feels like you actually know more about people's lives than perhaps they realize they're letting you know. We do have access to a tremendous amount of data. And because of that we have kind of a privacy officer within the firm.
Because of that kind of even as an employee of the firm I have to be very, very careful about what kinds of queries and what I look at in people's data. Yeah precisely because this is people's lives and you're right. But we have to take very seriously this responsibility that we're becoming a big part of how people move around the world. And we just want to be very careful with that.
Have you ever been concerned about the way you're using the services that might reveal things about you? I mean as someone who sees things from the point of view of the institution of the company and knows how powerful this information is. Has it changed your own behavior and how you interact with any number of these sites? Well now that you're talking about it I'm getting even more worried.
And this might be naive but I've often thought to myself that my experience inside of large companies that have access to these huge kind of treacherous data is that you almost always just look at these broad, broad aggregates. And I guess I've always just taken comfort in the fact that I'm boring enough of a person that no one would ever, no one would ever. I mean it's almost like, you know, so for example, like some people I know like my sister kind of shreds all of her credit card statements and does everything before she tosses paper in the trash. And I always just throw everything in the trash because I figure like I just don't feel important enough for someone to run it through my trash.
We do though, you know, in the Uber data see a lot of really, really interesting patterns. So for example, a data scientist named Peter at Uber discovered somewhat accidentally this really, really kind of interesting fact. And that is one of the strongest predictors of whether or not you're going to be sensitive to surge. In other words, whether or not you're going to kind of say, oh, 2.2, 2.3.
I'll give it a 10, 10 to 15 minutes to see if surge goes away. Is how much battery you have left on your cell phone. Oh, that's fascinating. When your cell phone is like down to like below 5% battery and that little icon on the iPhone turns red, you know, then people start saying, well, I better get home.
Because I don't know quite know how I'm going to get home otherwise. And I absolutely don't use that to kind of like push you a higher surge price. But it's an interesting kind of psychological fact of human behavior. I'm talking with Keitchen, he's a behavioral economist at UCLA and he heads up economic research at Uber.
When we come back, I'm going to ask Keith about some of his earlier work which explores the origin of some very interesting biases. Stay with us. We'd like to say a quick thank you and share a message from one of our sponsors, History. This Memorial Day, the groundbreaking series, Returns with a New Vision, a story of survival, hope and courage.
Starring forst Whitaker, Lawrence Fishburne, Jonathan Riesmeyer, Annika Noni Rose, Anna Pacoen, Tip T.I. Harris, and introducing Regge John Page and Malachi Kirby as Kunta Kinte. The journey of an American family begins with a name. Roots premieres Memorial Day, only on history.
Support for Hidden Brain also comes from Wells Fargo. Wells Fargo understands that you work hard to get the most out of life. From starting a business to saving for retirement, there are a lot of things you want to accomplish. Working together, Wells Fargo will take the time to understand what you're trying to achieve to help you reach your financial goals.
That way you can care for the people and things that matter most. Learn more at Wells Fargo.com slash Together. Wells Fargo. Together we'll go far.
Keith, in your life before Uber, you conducted different experiments into the origins of various human biases, including some issues that often come up in behavioral economics. Some of your most fascinating early work was with monkeys, and I understand to conduct these experiments, you first train the monkeys to become economic actors. How did you go about doing that? Oh my gosh, this feels like a lifetime ago, but with a bunch of colleagues at Yale University, we set out to answer this kind of somewhat ill-defined question, but one that we were kind of obsessed with, which was if monkeys were taught to use money, one, could good monkeys be taught to use money, and two, if monkeys could be taught to use money, would they behave in the same way that humans do, in all the kind of psychologically rich ways that we interact with kind of prices and think about wealth and think about how we spend our own money?
So basically what we did was we had a small colony of kapuchin monkeys that were living with us at Yale. I also did some work with a colony of tampered monkeys back at Harvard, and my Yale students were thrilled to hear that the Yale monkeys were much smarter than the Harvard monkeys. I didn't tell them it was because it was a different species. They attributed it to the school.
But basically what you do to teach a monkey to use money is you hire a bunch of Yale undergrads to basically just live with the monkeys for a long time. So these Yale undergraduates were typically psychology majors, they were studying the monkeys for various other things, the monkeys live in a big, spacious, comfortable habitat. They typically ignore the kind of humans that are moving around, but you have a Yale undergraduate every now and then just drop a coin on the floor. And we had chose these large metal washers to stand in for a coin.
Now a monkey thinks that's fascinating, runs over, grabs the coin, kind of chews on it, bangs it on the floor, every now and then would throw it around, almost dangerously, but would eventually lose interest in this metal disk. Then you'd have the Yale undergraduate stand there and stare at the monkey with an open hand outstretched. And just stare at that monkey uncomfortably, just setting this uncomfortable situation. And every now and then the monkey would actually pick up the coin and just put it in the undergraduate's hand, give it back to the students.
And then what we did was we trained the students to say, why thank you in really exaggerated tones and then hand the monkey a piece of food. So like a little apple slice, one undergraduate came to be known to the monkeys as like the apple undergraduate. If you gave that person a coin, they would always hand you an apple slice. Another person was the pineapple undergraduate, another person was the orange slice undergraduate.
Now what's amazing is you do this for about six months, nonstop for about six months. And eventually you start to realize that the monkeys understand that this is fiat money. Now what does that mean? Well what that means is, the monkeys had been very familiar with basic ideas.
Like there's a lever on the wall, if I pull this lever and apple slice falls from the ceiling. But money is something fundamentally different. When I find this coin, it's not just an apple lever. It's actually a choice between an apple lever and orange lever and a pineapple lever.
Because I can take this coin, I can carry it around with me, and I can wait until the, if I feel like having an apple, I've got to run over to the apple person and I can spend it there. Money is kind of fungible across different kinds of foods that I can purchase. So we started to see that. We saw that monkeys started to use these with each other.
And to save them, to kind of hide them from other monkeys, and then to make very, very rational decisions, rational looking decisions. When the price of apple doubled, you know, when that apple undergraduate started only giving one piece of apple, not two when you hand in a coin. You know, demand for that for apples went down and demand for oranges and for pineapples went up. How did the monkeys protect their money?
A fascinating component of the of the monkeys starting to understand money was that they displayed signs that they realized not only that they understood the value of this disc, but that they understood other monkeys recognized the value of this disc. So for example, you know, early on when the money that we were using was these physical discs, it's a little bit hard to shield from other monkeys, right? And you don't want to be kind of carrying all of these things around. So you would see monkeys like hide the discs, like you know, over in the corner under a pile of like wood shavings, right?
And then you would see the monkey like a little bit of hide the discs. Later actually, we taught these monkeys to use touch pads. And so like then kind of in some sense, like monkeys learned, you know, to use currency as if it's just kind of in ATM and if it just gets like Venmo to transfer a dozen other players to the apple guy to the... So this is like an apple wallet basically?
Yeah, basically. They're ahead of us on this mention. You know, they're a completely cashless economy at this point. You eventually got to the point where you would also introduce to the monkeys the idea that sometimes when they would give a certain amount of money to an experimenter, the experimenter might give the monkey three things, and sometimes experimenter might give the monkey one thing.
So in other words, it was unpredictable what the reward was going to be. You taught the monkeys to gamble. Yeah, that's basically exactly what we did. So we introduced them to two new undergraduates.
So one undergraduate would always approach a monkey with three pieces of apple in their outstretched hand. Let's call this undergraduate atom, okay? Would show three and either give over all three or would take two back and would only give one. Then we introduced them to another undergraduate, Ben, who always showed one, okay?
But if you gave Ben a coin, Ben would half the time hand over just that one. Half the time would add two apple pieces to his hand and hand over three, right? So now the fascinating thing is both Adam and Ben are presenting you exactly the same deal. There are 50-50 gambles between three apple pieces and one apple piece, right?
And you can give the monkeys a lot of experience trading only with Adam and a lot of experience trading only with Ben, so they kind of get that this is a 50-50 gamble. The interesting thing is what the monkey showed us through their preferences between Adam and Ben is that they experienced something just like humans do, and that's this very powerful psychological force called loss aversion, right? And that's the idea that it's more than twice as painful to experience a loss than it is to experience a similar size gain, right? Now what does that have to do with Adam and Ben?
Well Adam half the time gives you, in some sense, what he's shown you, right, the three apple pieces, but half the time he delivers you a loss. He takes away two and then hands over only one, right? So half the time he delivers you a loss of two. Ben half the time just hands you the one that he's showing you and half the time delivers you a gain.
He gives, he puts an extra two pieces of apple there and hands you an extra two apple pieces. What we find is that when given the choice between Adam and Ben, monkeys vastly prefer Ben, right? The guy who initially only shows one apple piece, but half the time gives you a gain, then they prefer Adam who initially shows you three apple pieces and then half the time delivers you a loss. They, you know, six to one prefer trading with Ben to Adam.
And that's fascinating because that cuts against their very, very core instinct, which is, well, if I have got a coin, why don't I go trade it with the guy who's showing me three instead of the guy who's showing me one, right? Like every kind of fiber in their body tells them that they should be going for more food, not less. And yet because of loss aversion and because they actually feel this very viscerally in the same way that people do, they actually prefer the guy who preaches it. They actually prefer the guy who promises less but delivers more.
I have to say that I'm thinking about what I would do in that situation and I have to say I'm clearly no smarter than a monkey because I would definitely prefer to trade with Ben because you're getting something that seems like a surprise. It's a gift, it's like, it's wonderful, it's unexpected, it's happy and at the worst, it's only going to be what he's promised you in the first place. But with Adam, he's taking away something that you thought was yours. Yeah, yeah, absolutely.
And interestingly, we see in our Uber data this loss aversion at work as well and you can, you can even just feel it psychologically. Like when you're a rider and you know, search pricing hits, like that's a loss, right? And that feels very, very bad. We've often been asked, well, why didn't you frame search pricing as discounts instead of search charges, right?
Why isn't Uber's standard price twice as high as it is? And then most of the time you're getting a discount, right? Why not kind of frame it that way? Because behavioral economics would predict that you would actually make people much happier if you did that way.
Well, behavioral economics would predict that you would make riders happier if you did that. I mean, the critical thing to notice is that we're two-sided marketing. That move which would make riders happier would also make drivers feel less well. So I mean, so we had thought for a while, why not just frame Uber's pricing as how much cheaper we are than taxi?
Because in many major cities in the United States, we're up to 60% cheaper than taxi fares when we're not searching. And you know what that means is that you can surge like 2.1 and still becoming basically out even if you would instead just take an attack. But even though because Uber drivers are in such an efficient system, they constantly have a paying rider in the back of the car, they're actually making more than they would have if they had been working as a taxi driver, framing this to them. Oh, you're making 60% less per mile than you would if you were driving a taxi.
That would be a cost of this pricing system. Keychain is a behavioral economist at UCLA. He heads up economic research at Uber. Keychain, I want to thank you for joining me today on Hidden Brain.
Shankar, it's been incredibly fun. Thank you for having me on. The Hidden Brain podcast is produced by Karamakar Callison, Maggie Pendwin and Max Nastrak. You can follow us on Facebook, Twitter and Instagram and follow my stories on your local podcast radio station.
If you like this episode, consider giving us a review on iTunes. It will help other people find the podcast. I'm Shankar Vedantu and this is NPR. Thanks for listening to Hidden Brain.
You should try out the NPR One app for your phone for conversations you won't hear anywhere else. This week, Guy Ra's interviews Ted curator Chris Anderson. They discuss the secrets of giving a great talk. You can find that conversation by searching for Ted Radio on the NPR One app where you can also find stories from your local station and more great podcasts.
NPR One is in the App Store now. I also wanted to let you know that Invisibilias Season Two returns on June 17. This season, the show goes to a prison and oil rig, a McDonald's in Russia and a beach in New Jersey to explore the worlds of work, family and government. Catch up on Season One and listen to the Season Two preview on May 20th at NPR.org.
I'm Jeff Potkas and on the NPR One app.