EconTalk |
John List on Scale, Uber, and the Voltage Effect
Jul 25 2022

41huFv9BcSL._SY346_-198x300.jpg Economist John List of the University of Chicago talks about his book, The Voltage Effect, with EconTalk host Russ Roberts. He discusses what determines scalability and argues that the only good ideas that count are those that scale. Along the way, he draws on his experiences as chief economist of Uber and Lyft to peer inside the black box of ride sharing.

RELATED EPISODE
Michael Munger on Traffic
Does rush-hour traffic drive you crazy? Is a congestion tax on car travel a good idea? Michael Munger of Duke University talks with EconTalk host Russ Roberts about the economics of traffic and congestion taxes. It takes a while to...
EXPLORE MORE
Related EPISODE
Nassim Nicholas Taleb on the Nations, States, and Scale
A language, a flag, a national anthem and shared history—like a heart that has to pump harder to support a heavier body, the bigger a nation gets, the harder to curate an identity. Nassim Nicholas Taleb talks about scale and...
EXPLORE MORE
Explore audio transcript, further reading that will help you delve deeper into this week’s episode, and vigorous conversations in the form of our comments section below.

READER COMMENTS

Floccina
Jul 26 2022 at 2:13pm

There has been some econ 101 bashing lately but I have always felt that simple supply and demand are the most reliable thing in econ. I think your guest showed how powerful supply and demand really are as the prices moved back to prior equilibrium after tipping and the test of an increase in payout. Yay for econ 101.

AtlasShrugged69
Jul 27 2022 at 11:02am

Recently I was listening to the Millenium Villages Project episodes with Nina Munk and Jeffrey Sachs. I asked myself; given that those projects did not scale, which of the 5 failures discussed was responsible? It occurred to me that those projects did not even succeed at the Petri dish/micro level.. So any sort of scaling would only magnify that failure. Sachs’ assurances that his methods were being implemented by other African Governments now seems even more abhorrent…

The discussion of the relatively low response of Labor Supply to high-demand events was interesting; I wonder, in the days before algorithms, predictions, etc.. Could Labor Supply (particularly the 3rd category of those sitting at home on the couch) have been more strongly affected by such events? (IE – Chariots in Rome at the Colosseum, Carriages in Harbors in 1700s America) I would guess the predictability of the event is directly correlated to whether or not an adequate supply of taxi-ing services show up. And if too few show up, no doubt the one’s there catch on fast and are quick to clear their schedules for the rest of the day. It surprised me as well hear John say that the number of drivers willing to get off the couch to head to an event were insignificant, but it makes sense – we know any particular event has a FINITE duration, by the time you get dressed, get in the car, and drive there, what if the event/demand has ended? Even if you get a 3x ride, you drove to the location with an empty car, and now you’ve interrupted your previous plans… Understandable that the bulk of labor supply responses to unpredictable events are John’s first 2 categories if you think about it.

On tipping – My favorite local restaurant uses mobile kiosks to pay. The server comes to your table with the card reader, inserts your credit card, and turns the machine around so the touch screen faces you and gives you 4 choices: 18%, 20%, 25%, or Custom Tip. Having worked as a server for a few years, I always tip 25% unless the service was absolutely terrible. It is basically the other end of the tipping spectrum compared to how Uber handles it; the social pressure to leave a decent tip, with the server standing right in front of you, is HUGE. I guarantee if you compare the tip rates at THIS restaurant vs a restaurant where the server runs your card, brings the receipt, walks away, and THEN you write the tip – The former will dwarf the latter.

It really does feel like this episode was cut short, consider this a vote to continue the discussion. Great guest who is both extremely intelligent and entertaining. I would love to hear him talk about Uber expanding into transporting goods/services in addition to people – reminds me of the Munger episode on the Sharing Economy, here’s to it happening! I have a LOT of power tools I am not using and would love to rent out!

P.S. – I refuse to believe that a monopoly with close to 90% of market share could have lost that monopoly power by ANYTHING other than the federal government breaking them apart via the Sherman Act! How can something as insignificant as #DeleteUber take down a company with so much market share? Oh well, this situation must be an anomaly… We should DEFINITELY use the Sherman Act to break up Facebook/Amazon/Google… *end sarcasm*

Luke J
Jul 30 2022 at 11:39pm

I toss no stone at Russ’s generosity, and I’ve enjoyed reading John’s other work, especially with Chicago Heights.

Perhaps unintentionally, however, this conversation reaffirmed to me that tipping is evil and needs to end.  I stopped using Uber when the app started encouraging both tipping and giving drivers 5 stars. Ride-share is a wonderfully disruptive innovation but I was disappointed with Uber and Lyft. Maybe someone will do it better.

 

 

AtlasShrugged69
Aug 2 2022 at 12:43pm

Now THIS is a hot take! Why is tipping evil? Are either of the below examples evil?

Last weekend I sold a Full size Box Spring mattress on FB Marketplace – ‘Ol boy shows up in a Prius… Of course the mattress doesn’t fit. He is like, let me call my friend to bring his truck. I say “Look, my wife and I are headed into town to do some errands, I can just throw it in the back of the truck and take it wherever you’re going” – he agrees. I get there, we unload it, and he hands me my asking price plus $20. I protest, but he said it was important to have it quickly and he was happy to pay extra for me helping get it to him
About a month ago, I order a case of a particular beer from my local grocery store for a Poker Night I was hosting. The manager I ordered it with said it would be in that Wednesday (the night of the event) and I could pick up anytime after 3. I get to the store, inform a different manager I was there to pick up the case of beer I had ordered. She went in the back to look for it, came back, and said she couldn’t find it.. I insisted the manager I had spoken with said it would be in that day, that I used to work in a liquor store, and I knew very well what the box looked like; perhaps I could take a look in the back with her? Her response was that customers weren’t allowed in the back because of ‘liability’. I told her it was for an event that night, and she replied that I would just have to wait until the next day when the original manager I spoke with came in for his shift… She returned to the front of the store. An employee nearby who had overheard our conversation came up to me, and offered to let me look around in the back. We went in and immediately spotted the case of beer sitting next to the doors… After paying for the beer, I went back in the store, found the employee, and shook his hand; transferring a crisp $5 bill from mine to his.

Having worked as a server I loved the tipping system. It rewards people who provide a superior service, and punishes those who do not (in the form of higher or lower wages). I’ve dined at a few restaurants in Europe which mandate a minimum wage for servers with no expectation of tipping, the service there was HORRIBLE. I had to chase the server down to get a refill on water.. In most restaurants in my town, the servers never let the glass get less than half full! Some people never leave a tip, some people always leave 20%, but I think the majority of consumers tailor their tip to the level of service. If you provide an inferior haircut, or dining experience, etc.. you can and should be compensated less for that lacklustre work. I can’t think of a better or fairer compensation system than tipping for the low-skill service industry.

Luke J
Aug 6 2022 at 1:59pm

Hi AS69:

I admit that evil is unnecessarily hyperbolic. Poor word choice on  my part. I’ll say tipping is an undesirable norm. Freakonomics Radio has covered tipping many times and my general takeaways are that tipping:

is discriminatory
is not strongly correlated with service
is largely a factor of the guests’ preferences & values

John List himself has been on a few of those episodes. In this conversation, he said tipping is largely the effect of the rider, not the driver. He also said that the distribution of stars is “pretty tight.” That sounds right to me, but they should be “pretty tight” around 3 stars, not 4.5-5 (Uber encourages riders to over-rate drivers and Russ says he feels bad giving few stars).

My take on the examples you provided:  If this is tipping, then we we need a different word to describe what happens in restaurants & bars, like:

automatically adding 20% on takeout orders.
placing tip jars at drive-through windows and registers bars
requiring guests to self-serve water and bus their own tables
turning a 45 minute meal into a 90 minute commitment (who is waiting on whom)?

I wish good things to those who find opportunities to improve the experiences of those around them, but tipping (to me) is hindering.

AtlasShrugged69
Aug 16 2022 at 12:35pm

Russ tweeted his thoughts on serving differences in Jerusalem vs the US yesterday (HERE). I’d like to think our humble discussion inspired it… (and even if it didn’t, no need to dissuade me of my notion).

Those examples all sound like exceptions rather than rules, but I agree in those situations there is no ‘service’ being provided beyond what you are paying for already, and I wouldn’t feel obligated (and likely would not) leave a tip in any of those scenarios.

The experiment I would LOVE to see would be: Take a random sample of restaurants around the US. Have 3 control groups of servers: Group A gets customer’s orders right, and checks on the table at least twice after bringing out the order (refilling drinks, etc.) Group B deliberately gets something wrong on the order, doesn’t ever correct it, and never refills drinks or checks on the table until bringing out the check. Group C is control, just servers doing their job as they normally would.

If there are not remarkable differences in tipping levels from that experiment, I will eat my hat. I know when I worked as a server, yes some people always tipped, some people never tipped, but the best predictor of getting a low/no tip was providing terrible service. I’d be curious to know what compensation system would provide better feedback loops for servers to gauge their effectiveness

Ajit Kirpekar
Aug 1 2022 at 12:15pm

The last anecdote from John about using the last 25 users as the basis for where to invest their advertising dollars makes sense but I wish John could have articulated the dangers of going with a thin, but more recent sample.

As in life, the answer is never one or the other, but it depends and small sample sizes are dangerous for a reason. What if the last 25 drivers paid off reasons related to randomness? How do we know 25 drivers is a meaningful small sample?

In the extreme case, one could take this approach anywhere and get totally misled.

Shripad Agashe
Aug 3 2022 at 12:56pm

What an amazing episode.

The lessons of efficacy test experiment vs experiments that can scale is a very useful lesson to learn. In many settings including a lot of corporate ideas fail because they work in a small setting with select audience and not general audience. Also, I really enjoyed his narration of failure of unannounced price surge. This where theory meets practice.

Comments are closed.


DELVE DEEPER

Watch this podcast episode on YouTube:

This week's guest:

This week's focus:

Additional ideas and people mentioned in this podcast episode:

A few more readings and background resources:

A few more EconTalk podcast episodes:

More related EconTalk podcast episodes, by Category:


* As an Amazon Associate, Econlib earns from qualifying purchases.


AUDIO TRANSCRIPT
TimePodcast Episode Highlights
0:37

Intro. [Recording date: June 23, 2022.]

Russ Roberts: Today is June 23rd, 2022, and my guest is economist John List at the University of Chicago. He is the former Chief Economist of, first, Uber, then Lyft. His latest book is The Voltage Effect: How to Make Good Ideas Great and Great Ideas Scale. John, welcome to EconTalk.

John List: Hey, thanks for having me, Russ.

00:55

Russ Roberts: What is the voltage effect?

John List: The voltage effect is a description that tells us what happens to a program's effects when we go from the small to the large. And, in particular, the voltage effect is usually a voltage drop. And what I mean by that is it looked great in the Petri dish; and then we scaled it up, and it doesn't look so great.

1:28

Russ Roberts: And, you talk about the different reasons that might be true. Why don't you tell us some of those?

John List: So, in the book, I talk about five reasons why there's a voltage effect.

And, I begin with the simple one of false positives. What that means is--of course, a lot of your readers will know--that it looked like it had voltage to start, but in the end, it really didn't. The data were lying. And what I mean by lying, they're literally not lying. You just got a bad sample. It's not representative of the underlying population. So, false positives is Vital Sign Number One.

Vital Sign Number Two is, in many cases, we overestimate the number of people who can be helped by our program. So, that's what I call: Know your audience.

Vital Sign Number Three is: Know your situation. In a lot of cases, we generate data that is really an efficacy test, and then we don't tell anyone it was an efficacy test; and then policymakers go and try to scale it, and it ends up that there's a big voltage drop. So, Vital Sign Number Three is about a representative situation.

Vital Sign Number Four is about spillovers. In many cases, our ideas have spillovers not only in terms of spillovers on others--you might go to a new market-wide equilibrium--ut there are also behavioral spillovers. Think about a rule or regulation that causes people to change their behavior. In economics, a lot of times we talk about moral hazard or behavioral economics has a discussion here as well. So, that's Vital Sign Number Four.

Now, as a Chicago economist, you might be thinking, 'Well, Vital Sign One, Two, Three, and Four, that's all on the demand side. What happened to the supply side?'

That's Vital Side Number Five. Vital Side Number Five is: What are the supply side considerations? Or, does your idea have economies of scale, constant returns to scale, or decreasing returns to scale? And that's the fifth vital sign. That's essentially the first part of the book, which is about trying to figure out: Does your idea or does the proposed policy have the DNA [Deoxyribonucleic Acid] of an idea that has a chance to scale?

And then the back half of the book talks about execution. It is basically a handbook that uses standard economics, behavioral economics to help people make better decisions not only in their lives, but in the workplace.

4:07

Russ Roberts: The book is called The Voltage Effect, but in many ways--you know, I think of this book, which is a wonderful mix of applied economics and behavioral economics and empirical work--as illustrating what I think of as how the truth is elusive. It's hard to know what's true. It's hard to know when a result is reliable. And you spend a lot of time in the second half of the book trying to give examples of what might be more likely to scale.

I want to start with this philosophical discussion of what it means to scale. It's a technical term, kind of. It means to get bigger, to have a wide impact across a large group rather than just maybe the small group that you worked on.

In my reading of the Development literature, there was an enormous focus in the early decades of the Development literature--the attempt to help nations escape poverty--on programs that would scale. Programs that would change the economy, create opportunity, create millions of jobs. And I feel like in recent years the Development literature has gone to smaller, micro questions: Is buying textbooks helpful? Does deworming help students stay in school and, therefore, do better later on in their life?

So, in a way, there's kind of two senses of scale.

One is: Does this small policy, even though it's relatively small, it could help a lot of people--in which case it would scale.

But, the other sense I think people use in the Development literature is more: Is it something like a quantum leap rather than just a small marginal impact? React to that. How do you mean the term, at least in that setting?

John List: Yeah; no, no, absolutely. So, when I think about high voltage, if you want an analogy, high voltage, to me, gives you a chance to reach new people in new situations with your idea. And, what I say in the book is, from the very beginning, if we're about making change in the world, we should understand what are the constraints or regulations or changes that we will be facing when we scale an idea and bring that back to the Petri dish and add that to our A/B test. [Note: Also called AB tests or A-B tests--Econlib Ed.]

So, I work with a lot of firms. I work with a lot of governments. Everyone today does A/B tests. They say A/B, they talk about--

Russ Roberts: Explain what that is.

John List: A/B, think about a medical trial. You want to figure out if the cholesterol drug works? So, what you do is you give the control group a placebo. That's called group A. Group B, you give the cholesterol drug. And then you look at some outcome that you care about. That's an A/B test.

In Development Economics, in the broader field of economics, we also do a lot of A/B testing. Firms and organizations always do A/B testing. So do governments.

Now, what that typically revolves around is: I bring forward the best inputs I can marshal. And, in employing those best inputs, I'm effectively doing an efficacy test. So, that efficacy test is important in an A/B comparison, but it's not very important when it comes to scaling. Why? Well, when you scale, you're not going to be able to employ the best possible inputs.

Well, let me give you an example. So, Steve Levitt was one of your guests. Roland Fryer and I started a pre-K [pre-Kindergarten] in Chicago Heights, which was a school of three-, four-, and five-year-olds. And, this was probably the moment where Steve and I had the most contentious debate in our long friendship because we were trying to decide which types of teachers to hire.

Steve said, 'You need to hire the best possible teachers and you need to leverage that this is a Chicago and a Harvard program and get the very best ones. Because,' he said, 'you can't go back to Ken Griffin with a null result and you also can't get null results published in academic journals. So, John, you're an idiot if you don't just hire the 30 very, very top teachers you can.'

And I said, 'No. What I want to do is I want to hire teachers exactly like Chicago Heights would hire those teachers, because I want a test of the policy that's in place.'

Now, Steve was right for his purposes. I was only half-right for mine. Let me tell you why.

So, Chicago Heights worked, and it worked in the sense that we helped kids in Chicago Heights and we got some academic publications. Good and good. Okay.

Now, what about scaling? So, my idea was: Hire the teachers exactly like Chicago Heights would hire them. So, we hired 30 teachers. That's really good for horizontal scaling. So, we could replicate that in Detroit, in Denver, in Washington, D.C., in Vegas, etc. That's horizontal scaling.

But, if I wanted to vertically scale that idea--and what I mean by that is have a bunch of pre-Ks in and around Chicago--I might have to hire 30,000 teachers from that input market. No way am I going to be able to hire 30,000 really good teachers like I hired those 30 teachers in Chicago Heights.

So, I was creating a program that was a test of horizontal scaling, but not vertical scaling.

Now, that's an issue when it comes to scaling because from the very beginning, if you're interested in scaling, I should have sampled some marginal teachers. That's what I would call Option C, or: Put in critical scale features into your A/B test and figure out. A/B is your efficacy test and C is your scale test: this is a kind of constraint I'm going to face at scale.

So, that's why I want to sample those types of teachers. That's part of the situation. Because then I can figure out, 'Does my idea scale?'

That's where we fail as social scientists. We tend to do what Steve wanted to do, and then we get our academic publication. We forget to tell everyone else it was an efficacy test. So, then when people scale it, it doesn't work.

And then when we do it right--like, I thought I was doing it right--I'm only half right.

And, at the time--this is 2015--I didn't really understand until I started to dig into the science of scaling.

11:13

Russ Roberts: Well, when you say 'half-right,' you--Steve [Steven Levitt], I'd like to hear his view--

John List: Absolutely--

Russ Roberts: I wonder if he's totally convinced by now. Steve was arguing, 'We're going to get the biggest impact of this work on these kids with the 30 best-possible teachers we can find.' You were arguing, 'We should just get 30 pretty good teachers like the school might be able to hire.' So, I don't understand the half-right part for you. What was half-right about your part?

John List: If I was primarily interested in horizontal scaling, that works because I think every community can get 30 pretty good teachers. But, if I'm interested in vertically scaling within the Chicago market, I think it's going to be very hard for me to get 30,000 really good teachers like those 30. That's what I mean 'half-right': is--

Russ Roberts: 30,000 is who are even pretty good because--

John List: Exactly--

Russ Roberts: you'd be digging down into the bottom, the left-hand tail. That's what I thought[?]--

John List: Oh, gosh! Yeah. I'm going to get killed on the supply side if I keep the quality the same. Right? Because I'm going to have to be getting day-traders and Citadel-traders, etc.

Russ Roberts: Well, they might be phenomenal.

But the point--I understand; I just want to make sure what you were were saying.

12:28

Russ Roberts: And, the reason I love this is that it really ties in with one of the other examples in the book, which is the challenge of Jamie Oliver's restaurant chain. And, I've always loved the insight that the reason McDonald's is successful is that they figured out a set of procedures that someone could execute who was pretty good at cooking, but not a great chef. And that's--if you need a great chef to run your restaurant, by definition you can't have a big chain. You can't have a chain with tens of thousands of restaurants and franchises. But, the incredible thing about this is that you were trying to test a curriculum.

So, if it was going to be a real test of the curriculum, having the 30 best teachers teach your curriculum is not measuring the marginal impact of the curriculum. It's a combination of curriculum with the best teachers.

And so, having them be the standard kind of teachers--and you couldn't use the existing teachers in the school exactly--because you were starting your own school. So, you did have to hire from scratch and you had to decide how good, special, whatever they should be.

What kind of impact did you have on those students?

John List: Yeah. I mean, in Chicago Heights, we had a pretty big impact. If you look at--

Russ Roberts: It's a tough school district--to make that clear.

John List: It's a tough school district. So, yeah, let's look at Chicago Heights. This is a school district in a city that the modern economy has left behind. This was at once a very proud community that had manufacturing jobs and a thriving local economy. This is a city that's about 25 miles straight south of Chicago. And, the manufacturing jobs left, you have a community that is proud but having its difficulties.

We started in the high school in Chicago Heights. And we started in the high schools because when you look at their data, at least back then, every thousand kids that started high school, about 480 of them got high school diplomas and 520 dropped out.

And, these are families that are very fractured. But we realized that if you start in the high school, it's very difficult to make big changes to a 15-year-old who is reading at a first-grade level or doing math at a second-grade level.

So, we helped some. Our best estimates there were we got 42 more people per class to get high school diploma. So, we're proud of that. But, we realized very quickly that if you want to make big changes and game-changing, or at least potentially game-changing types of situational changes for folks, you need to go back to the earlier years. And that's why we started the pre-K.

Now, that first class, they're sophomores and juniors in high school. So, we've been tracking--

Russ Roberts: Now--

John List: Right now--they're sophomores and juniors in high school. We taught them and their parents when they were three, four and five. So, we're continuing to track them. I think the results are relatively positive.

There was some depreciation. When you look right away, we do a pretty good job of pushing both cognitive test scores and executive function skills, and some of those stick.

So, I think we still have a real result, but the result is certainly not as impressive as when they first went to kindergarten. They looked like we were going to really knock it out of the park. But I still think there's a nice, positive effect within the checked data.

Russ Roberts: Well, you might have knocked it out of the park if you started in elementary school. It's an amazing result that you had any impact on just three years and then they went on to the regular, existing schools.

It's a really interesting example of the scale thing. You said you 'only'--quote, "only"--got 42 more people to graduate. So, it went from 480 to, what'd you say how many more? 42. So, 522--

John List: 520--

Russ Roberts: I don't know if that's a statistically significant result to spend a good class.

But, if you really did get 42 more people to finish high school, that's really extraordinary.

But then of course, as you would've pointed out, if you'd written about this in the book, you want to look at the cost. There are many things you can do with a lot--it was an expensive enterprise. Forty two is maybe disappointing, but it is 42. It's pretty good. Assuming that it's real.

John List: Every little bit helps, right?

Russ Roberts: Yeah. So, that's my counterpoint to scale. It's, like: I'm happy to help one person, but if it costs a lot of money, it might not be the best way to help people who might be able to help, too, with a different strategy.

John List: And that's my point about scale, right?

Russ Roberts: Exactly.

John List: I mean, and, that's why I talk about why scale is important--and you know this. You're a Chicago economist. I don't need to convince you. Look, I love decentralization. I love evolution, but there's also something unique about scale that helps us to be a lot more efficient, and that's not debatable.

17:39

Russ Roberts: Yeah. So, let's talk about Uber, because Uber has come up many times on the show. I've written about it many times in essays and in blogs, and you actually know something about Uber. The rest of us had no access to the black box; but you've been in the black box and you helped create the black box, and you did so also with their competitor, Lyft. So, to start with, talk about January 27th, was it 2016?

John List: Yeah, 2017. 2017.

Russ Roberts: 2017, yeah, sorry. Of course. Of course, it was. Tell us what happened that night. I mean, it's really quite interesting.

John List: Yeah. So, President Trump, January 27th, 2017, issues an executive order on immigration. A lot of your listeners might remember that executive order like it was yesterday. People around America went nuts. Many people did. It caused the taxi cab drivers around JFK [John F. Kennedy Airport] to go on strike--

Russ Roberts: The airport, the airport in New York--

John List: The airport, JFK, to go on strike. They were going on strike because they were so angry about the executive order. Okay.

So, Uber at that moment had a policy in place that when something like this happened--a market disruption--they did not want to be viewed as price gougers. So, they turned off surge. Makes sense, right?

Russ Roberts: Explain what surge is. Explain what surge is for people who might not know.

John List: For those of you who don't use rideshare or don't use Uber, surge is essentially a change in price--an increase in price--that is trying to balance the market. It's trying to balance the market because there's much more demand than supply. So, the idea is raise price, hopefully bring supply onboard or move cars to that location. And it also, of course, because of the law of demand, lowers demand, lowers the quantity demanded. Okay?

So, the idea is to try to bring the market back into equilibrium but still have prices clear the market rather than waiting time like traditional cabs had done.

So, Uber, in its mind, is doing the right thing here, but a particular taxi cab driver took offense to that lack of surge and thought that Uber was trying to break up the strike.

So, that taxi cab driver goes to Twitter, lets the world know--at least the Twitter world know--what he thought of Uber, and at the end wrote, '#DeleteUber'. That is a game changer now, because when you put that tweet along with everything else that was happening at Uber--and I won't even re-live it. You can re-live some of it through my book. I talk about some of it through my book. Right, Russ?

Russ Roberts: Yeah, yeah.

John List: In the intro, I talk about my wrestling with Travis, and then I come back to it in the culture chapter. I'll leave that to the sidelines, but what happened here: this particular moment is, I'm the chief economist at Uber; we're actually killing Lyft at this time. Lyft has maybe 5% to 10% market share back then. They're literally waving the white flag: 'We're done here. There's nothing to see. Uber's going to take over.'

Well, all of the, let's say, events that happened around this time--and it happened that #DeleteUber ended up being the proverbial straw that broke the camel's back. What happens is a lot of consumers and drivers start to delete Uber and they move to Lyft.

Fast forward to today: I just got done, as Russ mentioned, as the chief economist at Lyft. We had 30% to 35% market share across North America, and the inflection point was right there at that moment, that Saturday night. So, Travis came to my team, back--

Russ Roberts: The CEO [Chief Executive Officer], Travis Kalanick.

John List: Yeah, Travis Kalanick, the CEO and founder, came to my team and said, 'John, you know we have problems. You can see this.' My team, by the way, was called Ubernomics. So, if you type in 'John List Ubernomics', you'll see some stories about my team--and type that in Google.

So, Travis dispatched my team to start to think about: How do we get drivers back online? For months, I had been arguing that we should have tipping in the Uber app. And,I wasn't getting much traction around the company; but voila, January 27th, 2017 happens. I'm dispatched to get the drivers back. What is an economist going to say when somebody says, 'John, what's the best way to get the drivers back?' 'Add tipping,' right? Think about a way to raise wages.

So, I ended up partnering with a few people around the company. One was Aaron Schildkrout and one was Daniel Graf. And, we went door-to-door, basically, to the execs and we ended up winning.

So, when you win something like that at a company like Uber, then your group gets to help ship the product. So, that's when I had some fun that summer of 2017 in testing a little bit about the Uber app. You know, testing presets: Should you put in $0, $1, $2 in the preset for tip values, or should you put 0%, 5%, 10%? So, all that kind of stuff, and I've written a few academic papers exploring different aspects of trying to induce people to tip more through the app. As you're going to find out in the book in Chapter Four, we learned a fair amount about scaling through those exercises as well.

24:03

Russ Roberts: Well, there's so many fascinating things about this. First of all, I'm a big fan of price gouging and have defended surge pricing many, many times in all kinds of settings. And I want to try to defend it here. In particular though, I want to concede, based on an off-the-air conversation we had before we started this recording, that some of my arguments may be not as persuasive as I had expected in terms of the facts. So, let's back up a little bit.

John List: Yeah. Yeah. Yeah.

Russ Roberts: So, when you say, 'We've got to get drivers back online,' I assume what happened was because so many customers deleted their app or decided not to use Uber--they probably didn't literally delete--some of them didn't literally delete it, but they stopped using it. There was less opportunity for drivers and they started going over to Lyft, which is a different kind of equilibrium thing that's going on: As ride-share customers went over to Lyft, so did drivers.

So, the tip--the tipping idea--was a really interesting idea, and it's good to know that it's your fault. I hated it when it was added. Because, one of the great things about Uber as a customer was: once it shows up, I'm done. I don't have to have the emotional issue of handing over the money. I don't have to have cash. So, it's still: you don't have to have cash. I like that. I was happy to tip. After all, I got used to that extra annoyance of going back on my phone and adding the tip--which is not annoying at all. Just different. Takes a while. Just a little cultural change. So, I really--

John List: It's one more click, Russ. It's one more click. And you can click zero. Please do if you want to.

Russ Roberts: Yeah. I don't like doing that. I'm really not mad at you. And I like to tip, and I tip even when--here's the part about this. We're going to come back to the surge in a minute and my possible error, but I want you to clarify a few other things that listeners wouldn't know who were on the outside like me, which is a couple things. Talk about the longer-term--at first, the role of tipping did bring some drivers back to Uber, but the long-term effect was surprisingly small. That's the first thing.

The second thing was how many people tipped, which shocked me. Normally, I would say, 'You know that's ridiculous.' But, I think you know. So, I'm going to trust you on this one. So, tell us what happened. How many people tipped, and what happened to driver wages?

John List: Yeah. No, absolutely.

Let's start with the wages. So, what's interesting about Uber and tipping is that all the drivers wanted tipping because they thought that their wages would go up. It makes sense, right? I'm going to be paid the same amount of money in time and distance. That's how Uber pays its drivers. If you have somebody in your backseat, you're paid by how far you drive them and how long it takes. That's a typical formula. And, the formula varies by city, but it's the same general formula for everyone across the United States or across the world.

So, we add tipping, and this is now the Fall of 2017. We start to roll tipping out--in an experimental way because, of course, I do field experiments. So, I randomly have some cities receive the chance--the drivers--the chance to receive tips, and other cities don't. And, we do a nationwide tipping experiment. Jeff Wooldridge, and Ian Muir, and Bharat Chandar, and Uri Gneezy are all co-authors on these two papers that we've written.

But, the important element here is what happens is a very Econ-101 result. So, you add tipping to the app. Drivers supply more labor hours. Some drivers come off the couch because they say, 'Wow, I can get tips now.' Some drivers work more hours.

And, what you find is those drivers drive around with an empty car more often after tipping than they did before tipping.

And in fact, the amount of driving around with an empty car and its effect on wages exactly offsets the effect of the tips on wages.

So, what we have now is a new market equilibrium after we scaled tipping up to all the drivers: It looked good in the Petri dish. When only 5% of drivers got it, it looked really good; but when all of the drivers received it, guess what? The hourly wages of the drivers, hourly wages are identical, pre-tip and post-tip.

And that's a great Econ 101 story about how we've come to a new equilibrium in market-entry matters. The perfectly competitive model does a pretty good job in the long run predicting that wages would be similar.

Russ Roberts: I'm a big fan of that model. I used to teach it relentlessly to my students because I think it's a great way to organize your thinking even though, quote, "Model--markets aren't perfectly competitive." No kidding. Of course, they're not. But it's a great starting place. I get it.

The puzzle here that you didn't talk about in the book is: You were there. You and Uber executives saw this. You could have raised the take, right? Now, the idea that in theory it's not true, but in theory, the tip was paid by the customer. And, that's not exactly true, because that also changed. Once you had tipping as a norm--if it's an actual norm--you change the amount of rides people want to take because they realize they're going to be a bad person if they don't tip. We'll get to the frequency in a minute.

But, Uber could have overcome that if they wanted to, at least in theory, by raising the proportion of the ride that went to the driver. But they decided not to do that evidently. Correct?

John List: Well, look. At Uber, we've tried several different approaches to raising driver wages. They did try to change the rate card. There's a nice paper by John Horton, John Hall, and others that shows when you change the rate card--what I mean by changing the rate card is giving a higher pay for time and distance--the exact same thing happened that I'm talking about here--

Russ Roberts: Well, of course, it does. Of course, it does--

John List: So, it's unfair to say that Uber hasn't tried.

And the same thing happens with commission rates. When you say 'change the take,' you see the same thing across 20% and 25% commission rates.

Look, battling economics is not a battle you're going to win very often, and that's what you're trying to do here, and that's what Uber and Lyft have tried to do. And that, so far, as long as you have enough drivers waiting by the curb looking to make a little bit more money, these things are going to happen. Hall and Horton have that really nice paper looking at that same effect with rate card changes.

Russ Roberts: Now, you'd have to--to make it effective, of course--you'd have to restrict the number of--

John List: That's the thing--

Russ Roberts: of drivers. In which case, you've ruined the whole efficacy of the system, the whole idea of it. But, because then you'd start to have customers waiting. Which offsets the whole value of the product, which is that it's better than a taxi because the way it's going to be eliminated effectively, more or less, by the role of pricing, etc.

John List: And that's against the independent contractor notion. But by the way, that's happening right now in New York because of the legislation there. There are supply controls around New York City.

Russ Roberts: Explain that.

John List: I guess there's not much to explain. There was a debate about the number of Lyft and Uber and rideshare drivers around New York City some years ago, and we ended up placing supply controls in and around New York City because they argued that--well, several arguments. But, one of them was they argued about congestion and about the effects of Ubers and Lyfts around the city. So, we have gone through periods where there are supply controls.

Russ Roberts: So, that explains why the last time I was in New York in April or March--it was March--that I was shocked at how expensive Uber was. Is that the reason?

John List: I suspect that might be a contributing factor.

32:33

Russ Roberts: Okay. Well, let's go on to: How many people tip?

John List: Yeah. Let's talk about tipping. So, when we designed the tipping within the app, Travis Kalanick said, 'I don't want this to be simply a price increase. I don't want this to be restaurant tipping where everybody has to do it.' The strong social norm, of course, is you do it, in America at least. You do 20%, 30%, whatever.

Russ Roberts: Fifteen.

John List: Fifteen. Whatever.

Russ Roberts: Whatever.

John List: He didn't want that. He said, 'I want 10% to 15% of people to tip.' Where did he get that from? I'm not sure.

So, we did a few things. One, we said: 'The tipping decision then has to be divorced from the ride itself.' So, what I mean by that is the ride is completed, you leave the vehicle, you cannot give a tip until after the driver gives you your star rating. Then when you're long gone, but before your next trip, you are asked, 'Stars?' and 'Do you want to tip?' Because--that's point number one, and that turns out to be very important. That's what I think effectively moves the environment from a very strong social norm to one of: Now it's kind of a self-image concern because I'm in my house or in my office and I open up the app and I say, 'Well, should I tip Jane or not?' Sure, there's still a social norm around that, but that's internal. Now, that's very different than the social norm that we tend to think of around tipping. Okay.

So, we then, of course, experimented on the presets. We experimented on how to present the information; and all of that. And we ended up meeting the goal. When we did the tipping in the app, it's roughly 10% to 15% of trips are tipped.

Now, what's interesting is the composition of who gives those tips. And, I think some of the stark results are along the lines of only 1% of people tip on every trip.

So, Russ, you seem to be very confident, and, 'I tip this, I tip that.' Well, Russ you're in the 1% then. So, when people ask me, 'How do I get to the 1%?' I say, 'Tip your Uber driver every time. You've just made the 1%.'

35:22

Russ Roberts: But, the problem I have with that, John--the puzzle that that raises--is that: How did that then have such a big effect on drivers? Right? It couldn't have brought too many drivers into the pool. Well, I guess it took 15 weeks--in the book, you talk about. I guess they were hoping they were going to get some tips. After a while they found out, 'Well, actually, most people don't tip,' and it ended up back being similar.

John List: No, I think that's right. I mean, this is expectations, right? And beliefs. That's one of our primitives, right? You have preferences, beliefs, and constraints. And there are[?] in tipping, I'm going to kill it.

Now, remember, though, there's a lot of heterogeneity in who receives tips. So, the marginal driver might be very different than the average driver. And we talk about that in the papers.

But, the other side of that coin is that three out of five people never ever tip.

And that's very different than when you look at the data on tipping in Yellow Cabs, for example, where we finish the trip and face-to-face I decide to give a tip.

Russ Roberts: Sure.

John List: Right? There you have 90%-95% of people giving a tip. That's very different than the restaurants, where you have a large fraction giving a tip. Now, there you have social pressure, social norms, 'I want to do the right thing.' I don't want to be viewed as a frugal economist--and you can put a lot of different words in there--a donkey, etc.

So, that's why I think it's a little bit of a different calculus, decision calculus, and that's why you get such different results across rideshare tipping in, say, Yellow Cab tipping.

Russ Roberts: You should add--because, again, I mean, I spend a lot of my time in Ubers talking to the drivers because I'm interested in their world, but I don't know if every listener knows that we, as riders, get rated. You said that you're asked to be tipped after the driver has given you the number of stars. Everybody might not know that.

And also, I heard a rumor that if I have a high rating from drivers--and you can look at it if you look for it--that I will get good drivers: that Uber will match me with good drivers who are rated highly by their customers in turn. Is that true?

John List: I was not aware of preferential dispatch based on rider ratings when I worked at Uber or Lyft. But, we can pretend it's true so you continue to be a nice person. If that's what people need to be nice to their drivers, we can pretend it's true, but as far as I know, it's not true. Now, when you look at the distribution--

Russ Roberts: But we do get rated. But we do get rated--

John List: You absolutely get rated--

Russ Roberts: And if we're drunk and abusive, after a while, you won't send me a car, I assume, because my rating is so low.

John List: There are different activities and actions that you can do as a customer that will get you kicked off the platform. Just like if drivers misbehave, they get kicked off the platform. So, there are certain restrictions in your behavior. But whether you're a 4.8- or a 4.95- or a 4.3-rated rider--now for all the listeners, the stars go one to five, so 5 is the highest-rated driver and the highest-rated customer--as far as I know, there is no preferential dispatch around that.

And, just to be clear, though, the distribution is pretty tight. A lot of customers are receiving a lot of stars. And that's kind of by design, right? Because, if you receive a bunch of 1s, you might stop using Uber and Lyft, and that's not good for Uber and Lyft. So, let's talk about what the incentives are for the driver, first of all.

Now, secondly, I think what's kind of interesting about the star system is: when you look at the driver ratings themselves, those only have a small influence on the amount of tips that they receive. So, when we look at what determines how much the driver receives in tips, you can kind of break it down into three buckets. You can say rider characteristics, driver characteristics, and trip characteristics. It turns out that the first of those--the rider characteristics--those are more important than the other two combined.

Russ Roberts: Huge. Yeah.

John List: So, it's really the rider fixed effect that is causing drivers to get a higher tip. That, in terms of explaining the variation, much less the trip characteristics--and when I say trip characteristics, we have GPS [Global Positioning System] so we know if you're speeding, if you have a fast stop, a quick start. Those are all things that matter, but they don't really matter that much.

Russ Roberts: Yeah. That whole thing kind of ruins the romance I have about tips as an emergent phenomenon to incentivize good behavior. But, okay.

40:31

Russ Roberts: Early on in this conversation about Uber and Lyft you said, 'We were dominating Lyft' when you were at Uber, and they had a 5% to 10% market share and they were waving the white flag. There is some debate about whether Uber has any chance of ever being a viable, profitable entity--even with 100% market share, if they drove Lyft out of business. And that much of what Uber's play is, is an attempt to build ultimately a driverless car network; but that the current model is inherently unprofitable. And it's just a--right now, all they're trying to do is get market share and be ready for the world of driverless cars. Do you think that's true? Do you think Uber has a viable profit opportunity in its current mode?

John List: Yeah. Yes. So, let's back up and first of all talk about scaling. So--

Russ Roberts: That's what it's about--

John List: Yeah. That's what it's about.

So, Uber was able to scale this far because they don't need Danica Patrick or Michael Schumacher or Al Unser Jr. as drivers, right? Let's get back to the uniqueness of what's actually providing the service. It's one piece of labor, one piece of capital; and that labor can be people like me driving--the normal people. So, okay. So, we can now scale up--

Russ Roberts: It's a genius idea, but go ahead.

John List: No. Look, the genius idea is that we're using prices to clear markets rather than wait times or rather than 'might makes right' back in the day, long time ago.

Russ Roberts: And to be able to find a trustworthy person--who is a stranger--you're willing to get in the car with. It's an incredible change.

John List: It's an incredible change.

Russ Roberts: We underappreciate it.

John List: Oh, absolutely. I can remember when I was raised in--I'm a bucolic. I was raised in Sun Prairie, Wisconsin. My mom and dad--and this is a trucker and a secretary--they would've killed me had I gotten into the backseat of a car, some stranger's car. They still don't understand it today. So, if mom and dad are listening to this, they still wonder, 'What in the world are you doing, Johnny?' That's basically what they would tell me. My grandpa's from Germany. He'd have that thick German accent saying, 'Johan is a fool.'

Okay. So then the model is set up where you're right to an extent--I mean, drivers are taking 80% of the take. That's changed a little bit now that we've disconnected what people pay and what drivers receive. When I first started at Uber, those two were connected. So, if a rider paid $10, the driver would get $8, and then the other $2 would come to Uber. Some would be used for insurance and overhead, etc. But now that's been delinked, and it was delinked between 2017 and 2019 in both firms.

So, now, you have a little bit more of a chance because you can price discriminate by routes and you can charge a little bit more, and then the driver doesn't always get 80% or 85%--75%, 80%, 85%. Okay. So, that helps a little bit.

But, I think you're right in the sense that the winner of the big tent--the winner of the big scale--is either going to have to scale beyond moving people--so, it might be moving freight, moving food, running a helicopter service, so I was in the flying car division at Uber called Uber Elevate--or, even move people in a different way than what we're imagining today and also move freight and food at the same time, where you have economies of scale by being able to move a lot of different things or services or products or whatever. You have a shot that way.

But I think the biggest tent comes, and the biggest reward will come, to the player who develops autonomous in a safe way, in a way that we can all trust and we can all allow to scale. And then it's going to be the owner of the capital who is going to receive those awards or rewards. And Uber and Lyft were both in that game.

When I was working at both of their shops, they're no longer in the game because they realized that they weren't good enough at it, and they ended up turning it over to others. I think Waymo, as far as I know, is probably still the leader in autonomous. I'm sure Tesla and Jaguar and Ford and others are close competitors.

But in the end, it's going to be the owner of that capital, and then we'll see if the capital can be replicated or if the IP [Intellectual Property] is so good and owned by one player. That's where I'm going to bet it ends up. A lot of the rents will end up flowing to that particular party. So, I think there are kind of two outs for Uber and Lyft, and the autonomous one is sort of the most obvious one.

45:48

Russ Roberts: Yeah. So, my favorite Uber story is that a driver told me once he took a guy from Washington, D.C., I think, to New York City--no, it was Atlantic City. I think it was a thousand-dollar fare, something like that. And the driver asked the guy--I may have told this story; I apologize to listeners--but the driver said to the customer, '$1,000? I mean, you can fly, you can take a train, you could buy a car, you could buy a used car and it'd be cheaper. Why would you spend this amount of money on an Uber?'

And, the guy said something extraordinary. He said, 'Well, I'm going to a jewelry conference convention in Atlantic City, and I have an enormous amount of jewelry on my body right now, and if I drive myself with a rental car or I get on a train, I might get mugged. I'm just afraid. So, I'm here.'

Of course, the driver being self-aware said, 'Well, aren't you worried about me? Why would you tell me?'

John List: So, that's what I was going to say. What if that Uber driver takes the wrong turn?

Russ Roberts: Right. And the answer, of course, is that Uber knows where he is. And this guy was really smart.

Just to illuminate the economics of it, which I find very beautiful, we step into the car of a taxi driver because we trust the brand--the Yellow or whatever it is--to have filtered and screened the driver to not be a maniac and to keep tabs on them in some way that is reliable. I get into the Uber, which is much weirder because the brand Uber is much different; and the reason I get into them is I trust that GPS [Global Positioning System] system and I trust the rating system to do the filtering.

So, it's a really extraordinary--and of course, it's working with Airbnb also. It only really works if people give honest ratings, which of course they don't exactly. I have trouble giving a bad rating even to a bad driver who is reckless. It's very, very--it's emotionally difficult.

So, it's a very interesting giant social experiment, these sharing and these different kinds of apps and market interventions that have come along.

But the part I want to clarify where I made the mistake or at least unexpectedly was surprised by what you told me off the air, is that I've always assumed that surge pricing had both an effect on quantity supplied and quantity demanded. So, quantity demanded, obviously, when it says to get to the restaurant you were headed to, it's now 3x [priced 3 times higher--Econlib Ed.]. You go, like, 'Oh, I'll eat at home.'

Very important when there's something that lets people who have either a more urgent desire to travel or have more money--it's both--to get your car. You step aside--another amazing thing. You don't have to know what they're interested in or why they need to get away or there could be something dangerous going on, they need to get home from the airport. You just don't go because it's 3x.

But I just thought there was a similar effect on the supply side. A lot of drivers sitting--would-be drivers--on their couches get an alert from Uber that says, '3x Surge Pricing. Go get your car. Go drive over to JFK. There's a lot of opportunity,' if they had used surge pricing that January 27th night. But you say that supply-side effect is quite small. So, talk about that, because that's surprising to me.

John List: No. Absolutely. So, you're right. On the demand side, I can go catch a train, I can catch a cab, or I can just stay home. I just don't use the service, or I use Lyft--

Russ Roberts: or go later, go in a couple of hours.

John List: Yeah, go later. Intertemporal substitution is important.

On the supply side, there are sort of three ways in which you can bring on a little bit of extra supply in those situations where there's an unanticipated surge. So, you can have a person make an extra trip. So, instead of going home, I'm going to take one more trip because it's 3x. You can have a car relocate, and there's some relocation. Or you can have just totally new supply coming onboard. A person sitting on their couch and they have their app open and they see a heat map and 3x, 5x, 10x, whatever. That third one isn't as large as what most believe.

In my analysis of both Uber and Lyft data, if you really want to combat the type of problem like the New York Yankees baseball game just got let out or the Mets game got out or the Patriots just finished in Foxborough, you really need to set up longer-term incentives and set up things like: if you drive around RFK stadium [Robert F. Kennedy stadium] in Washington, D.C. around 3:00 in the afternoon when the Redskins have an early game--the game will end around 3:00 or 3:30 PM, or 4:00 or 4:30 PM on East Coast time--you'll receive guaranteed surge.

And companies do that because a lot of surge is predictable. A lot of surges, we know when events happen and we can make sure to move around supply, and we let people know that on Sunday night: that on Thursday night, there's an NBA [National Basketball Association] game happening at the Garden in New York City and we're going to make sure to have extra cars around there. That works. So, people will shift around their effort during the work week when they can schedule it.

What doesn't work as well is an unanticipated 3x getting a person off the couch. That's in our imaginations if we think that that's real. That will not--in my time at Uber and Lyft, I never really saw that having a deep impact in terms of the supply curve affecting the price and then a great number more of trips occur at an affordable price. I think that's more dreaming than reality.

Russ Roberts: Which means that the price goes up more than it otherwise would, I assume--

John List: That's right--

Russ Roberts: because you need to choke off some of that demand. If the goal of the algorithm is to always ensure a relatively short wait, which is the bread and butter--and I just also want to add, all this is, again, enormously fabulous. To me, it's better to have the possibility of having a ride than no ride at all. There's a fixed number of taxi medallions in New York [medallions are licenses to drive a taxi--Econlib Ed.]. They, I'm sure, are used more often on a rainy--I assume they're maybe are used more often on a rainy night than on an exit of a game. But it's still fixed. And at least with the Uber, you do have this possibility for--there's more of a possibility of increased supply, and certainly in the long run if an area is growing. I think the number of medallions in New York City hasn't changed since 1936 or something.

John List: Wow.

Russ Roberts: That's what we used to say. Like, don't say, 'Wow.' It might not be true, but--listeners, help me out there. I suspect for those listeners who remember bootlegger and Baptist regulation, I suspect that the restrictions on supply that New York City has imposed--true, they could argue it was for congestion--but I'm sure some of it was a response to medallion holders trying to maintain the value of their asset.

John List: Nah. You wouldn't think rent-seeking is at work, would you, Russ? I can't believe you said those words.

Russ Roberts: I didn't. You said it.

53:08

Russ Roberts: Okay. So, I want to move to an example that--it's based from your experience at Lyft--but it's an incredibly valuable general principle, which is the difference between marginal and average. And, talk about what that meeting you were at Lyft and what jumped out, and how it came up.

John List: No, absolutely. So, let's start with a caveat that I could tell a story like this for my days in the White House, for my days at Uber, for my days at Lyft, for my days in any organization. So, my folks at Lyft, please don't kill me when you hear this and you read it in the book.

Okay. So, that said, there are various groups in Lyft that are tasked with doing certain chores. The one group I was talking to on a certain afternoon was responsible for bringing in more drivers to the Lyft platform. Okay. So, their typical mode of operation is to place advertisements, and they ended up placing ads on places like Facebook and Google; and they were in a mode of: Logan Green, the founder and CEO of Lyft, gave them a new tranche of dollars and said, 'Bring in as many drivers as you can.'

So, they came to me with some data. And, the first tranche of data that they gave me said, 'You know, for the last thousand drivers that we've brought in through Facebook ads, it's cost about $500--$300 to $500--per driver, on average.' Okay. And then they showed me the same data for the last thousand drivers from Google ads, and it was about $700 on average. So, they said, 'With this new tranche of dollars, we're going to do it and spend our money on Facebook ads.' Makes perfect sense, doesn't it, Russ?

Russ Roberts: Data talk[?]. Let the data speak.

John List: Evidence-based decision making, Russ. Evidence-based decision making.

So, I said, 'Look, a thousand drivers on each, that's a lot. Is there any way you can tell me, say, about the last 25 or 50 drivers?' They said, 'We don't have that, but we'll send you that tonight.'

So, that evening they sent me an email, and they said, 'Professor, on Facebook, the last 25 drivers were about $1,000 each. And on Google, the last thousand were, like, $750 each.'

And they said, 'We understand your point. We wish we could go back and move some of those Facebook dollars to Google, but now moving forward, we're going to actually invest in Google ads instead of Facebook because the last few is the most predictive of the next few.'

So, it's not exactly marginal thinking the way we teach it in our class, where we say, 'What's the benefit of the next dollar spent?' That's marginal benefit. And, 'What's the cost on the margin for the next unit we produce?' But it's closer than the average.

Russ Roberts: It's close, yeah.

John List: So, what the point of this chapter is, is: Any time you can take thinner slices of data, you should do so.

Now, the misnomer that a lot of us have is just add and add data, and because your estimates become more precise, they're going to be more informative. That's a typical way to think about it, right? Big data: We estimate a regression, the standard errors of the regression become very precise, is we add sample because we know that N is in the denominator of the variance term. Okay? So, that makes perfect sense.

But the point here is: If you add data from a different regime or a long time ago, that can be very misleading when you're making your decisions, and any time you can take a thinner slice of the data, that will lead to a much better decision. And it's what we call thinking on the margin.

And we teach this all the time. But look, I've taught Principles for 25 years, and I always get the feeling that when students walk out, they've memorized a concept and they can write it down mathematically, but when it comes time to take that concept and put it to use, fails.

And let's be honest: Logan Green [Lyft CEO], he is an Econ undergrad major--

Russ Roberts: Smart guy--

John List: Very smart guy, very clever guy, an Econ undergrad from UC Santa Barbara [University of California, Santa Barbara]. He's smart.

But it's just not first nature to think, 'I need to look at this on the margin.'

So, as I write about in the book, I wrote what I called the Adam Smith memo, and the Adam Smith memo was a simple memo on and about thinking on the margin. And it gained traction during COVID [CoronaVirus Disease] because we had to find a way to downsize in an efficient manner.

That morning, we cut 20%/22% of our workforce. And Logan said, 'We need to cut even more spending.' That's when ideas tend to have a chance--is when people are desperate. They say, 'Okay. We need a new way to think.'

And when change has to be made, that's when the Adam Smith memo--I published that all around, left and became a very famous memo, because it gave people a proper guide; and it makes sense after you hear it, right?

There are dollars thrown all over the ground as I write about in that memo, and here's a way that we can pick some of them up and save a little bit for not only customers, but also shareholders and the firm.

And, once you start to think that way, people readily see it works. So, that's literally what Lyft is going through right now is a bunch of Adam Smith thinking and making decisions on the margin.

Russ Roberts: Well, I think I would've called it the Menger memo, a name you do mention in the book for marginal thinking. But, Adam Smith is a little better known, and it probably got people's attention. So, that's probably marketing.

John List: It's a little bit better known, just a little bit better known.

Russ Roberts: Just a titch.

My guest today has been John List, and I'd love to talk to him more about what's in his book, but he has to go to class.

John List: I do.

Russ Roberts: So, the book has many, many other things, but I appreciate, John, you sharing your special insight from your Uber and Lyft experience and your general insight as a field experimenter, and thanks for being part of EconTalk.

John List: Russ, thanks so much for having me and have a great day, everyone.


More EconTalk Episodes