DUBNER: Explain, if you would, why these data are particularly useful in trying to answer this question?
LIST: So we have mounds and mounds of data. We have millions of drivers. We have millions of observations, and 25 million driver-weeks across 196 cities. So just the depth of the data and the understanding of both the compensation function and the production function of drivers gives us a chance to — once we observe if there is indeed a gap — gives us a chance to unpack what are the features that can explain that gap.
DUBNER: All right. So describe the data. I want to know both the overall universe of Uber driver data in the U.S. and then which subset your data comprises of that.
LIST: We look at driver-weeks for Uber drivers from January of 2015 to March of 2017.
DIAMOND: That is over 1.8 million drivers during this time, and over 740 million Uber trips. So we have really a lot of data to work with. And for part of the paper, we focus on one city.
HALL: So we picked one city to go deep on for very practical reasons. The work that we’re doing is very data-intensive.
LIST: Being from Chicago, I said, “Let’s do Chicago.”
DUBNER: Now, how much of that was because you’re from Chicago and how much was it was you’re lazy and it’s going to be easier for you to work with data from Chicago?
LIST: No, no, no. No, no. So the team is out in San Francisco, and we’ve since — done Detroit, Houston, and Boston, and we find similar results.
DUBNER: And give us gender breakdown in Chicago, and how representative that is of the rest of the U.S.?
DIAMOND: So in the nationwide sample, 27 percent of drivers are female. And in the Chicago data set, 30 percent of drivers are female. So it’s a slightly more, but it’s pretty similar.
DUBNER: Let me ask you this: how big — if you know — is what we call these days “the gig economy”?
LIST: Some estimates suggest that up to 15 percent of people are full-time employed in the gig economy. And other estimates tell you that up to 30 percent of people are employed at least part-time in the gig economy.
DUBNER: And does the gig economy tend to lean more male or female?
LIST: The gig economy looks a lot like what we have on Uber, which is about a third of female drivers and two-thirds of male drivers.
DUBNER: Is the Uber algorithm gender-blind?
HALL: The algorithm is gender-blind, both in the literal sense that it doesn’t — that gender is not fed into it.
DIAMOND: It does not incorporate gender into the calculation at all.
HALL: And in the sense that it doesn’t facilitate discrimination by the users, the human users, who are more clever than the algorithm.
HALL: But that does not guarantee that the platform couldn’t facilitate some kind of nefarious discrimination.
LIST: There are two kinds of discrimination that might actually occur on Uber’s platform. The first is from the dispatches or from setting the wages. And that’s what Uber’s job is. There’s no discrimination there. But on the other hand, there could be customer-side discrimination. It could be the case that we as riders prefer men or prefer women as our drivers.
DUBNER: So did you find discrimination on behalf of the users of Uber? Did riders tend to prefer male drivers to women?
LIST: No, we find no evidence of discrimination on the customer side, meaning that riders don’t prefer men to women or women to men. They view men and women the same when it comes to being their driver.
DIAMOND: That’s right. And we don’t see overall differences in rejection rates between male and female drivers. And if you were to put that in the regression, it doesn’t contribute to a gender gap.
DUBNER: Right. So let me just make sure I’m clear. You’re saying there’s no discrimination on the Uber side, on the supply side, because the algorithm is gender-blind and the price is the price. And you’re saying there’s no discrimination on the passenger side. So does that mean that discrimination accounts for zero percent of whatever pay gap you find or don’t find between male and female Uber drivers?
LIST: That’s correct.
DUBNER: All right. So you were telling us that your prediction was that there’d be either zero or a positive pay gap for women. What kind of pay gap did you actually find if any, between male and female Uber drivers ?
LIST: We found something very surprising. What you find is that men make about 7 percent more per hour on average …
DIAMOND: … which is pretty substantial.
LIST: For doing the exact same job in a setting where work assignments are made by a gender-blind algorithm and pay structure’s tied directly to output and not negotiated.
DUBNER: So a 7 percent gap, how does that compare to the best research in other occupations?
DIAMOND: So there’s been some previous work that has looked at within-firm gender pay gaps. And seven percent is not very different than the overall average we see across all firms, even in the traditional labor market.
LIST: Sadly so.
DUBNER: Were you despondent or depressed or a little sad when you saw the size of the effect here?
DIAMOND: I just wanted to know more. I wanted to know where it was coming from, and what were the causes.
DUBNER: Okay. So I want to get into what are the factors. In the paper, you write that there are three. Number one?
LIST: So after reaching the dead end of discrimination doesn’t seem to be a determinant, we then decided to ask, “Well, what about where and when?” So what I’m thinking about here is time of day, day of week, and where in Chicago they actually drive. And here, we had some success. So what we find is that after you explore the where’s and when’s, we find that we can explain roughly 20 percent of the gender pay gap by choices over where to drive and when to drive.
DIAMOND: And an important contributor to the gap is particularly where the rides started. So different neighborhoods are going to differ in the types of rides that you’re going to get, and also potentially the frequency of rides you’re going to get called for. So men and women tend to target different neighborhoods of where they’re driving, and men are targeting more lucrative pay areas than women.
DUBNER: And does that have to do with, like, at 3:00 in the morning on Saturday, and I want to go out to where all the bars are, and there might be a surge? Or is it more — I don’t know, early-morning airport trips? Can you characterize the nature of those most lucrative trips, that men seem to be doing a little better at?
LIST: So what is more important than when you drive, is exactly which trips or routes do you tend to focus on. So one particularly salient example here is that airport trips tend to be the most profitable trips on the Uber platform. So what you have is that men tend to complete more airport trips than women complete.
DIAMOND: So we find that where people pick up is more important for contributing to the gender gap than the when. There are differences between when men and women drive. Men are much more likely to drive the graveyard overnight shift, which could have more people coming home from bars or whatnot. But women are actually dramatically more likely to drive the Sunday afternoon shift, and that is also a very lucrative driving time. So it’s not so much that there aren’t differences about when men and women drive. It just doesn’t seem to be super-related to driving a pay gap, because they’re both driving at lucrative times, they’re just different times. I mean, Sunday afternoon, that’s when football is on. Maybe women are more willing to go drive for Uber then.
DUBNER: And why — why is Sunday afternoon a more lucrative time to drive? It it because so many male drivers are watching football, so they’re not flooding the market with supply of drivers and therefore the price goes up?
DIAMOND: I mean, that’s a theory. We haven’t unpacked what’s so magical about Sunday afternoon, but pay tends to be high then. And women work disproportionate hours then.
DUBNER: But for all those potential differences, the absolute amount is still relatively small. You’re saying 20 percent of a gap of 7 percent can be explained by time and location, right?
LIST: I think that’s right. But now after looking at time and location, that analysis actually hinted at a deeper effect, which I will call driver experience.
HALL: Yeah. So there are pretty large returns to what we call experience, which is literally the number of trips that you have done. This is an area that’s pretty well-studied in economics, and it’s learning-by-doing. We estimate that the more trips you do as a driver, the more you learn about how to make money on the platform.
DIAMOND: So obviously this is not getting a raise from Uber, in the sense that the formula of pay is changing. Drivers are just getting better at figuring out when and where to drive, a little bit about how fast to drive, and also, how to strategically accept or cancel rides that they think are a bad match.
HALL: And we estimate that men and women learn identically quickly in trips. So a man or a woman in the data who have done the same number of trips will have accumulated the same amount of learning. However …
LIST: When you look at the experience of our drivers or the average tenure, this is heavily tilted in men’s direction. Men are far more likely to have been driving on Uber for over two years. Women are likely to have just joined in recent months, and this is because women leave the platform much more often than men.
DUBNER: What is the overall driver attrition rate? I don’t know whether it’s measured in six months or a year, or whatever.
DIAMOND: Yes, six months is what we’ve been looking at.
LIST: More than 60 percent of those who start driving are no longer active on the platform six months later.
DIAMOND: So the six-month attrition rate for the whole U.S. for men is about 63 percent, and for women it’s about 76 percent.
DUBNER: Wow. So that would connote to me, an amateur at least, that maybe this gender pay gap among Uber drivers is reflected in the fact that women leave it so much more. Maybe it’s just a job that on average, women really don’t like. Is that measurable?
LIST: That’s a good question. I like to think about people liking to be a ride-share partner, rather than disliking it. But it is measurable. When you look at the attrition rates, it is true that women do fall off the platform more. But they’re also earning less. So it’s not clear whether it’s because of preferences for not liking to drive as much as men like to drive, or if it’s simply an earnings effect.
LIST: It’s likely the combination of both those two.
DUBNER: Yeah, but does this higher female attrition rate mean that the average female is likely to be less experienced than the average male driver, and therefore will earn less. Yeah?
LIST: No, that’s right. When you look at experience, really men are more experienced than women because of two primary reasons. One, women drop off the platform more often than men. But, two, even for those who are on the platform for the same amount of time, since the average man drives about 50 percent more trips per week than the average woman, you still have the experience effect for those who have been on the platform the same number of months.
DIAMOND: So at any given day or time, the men driving for Uber have a higher level of experience under their belt than women, and that plays an important role in compensation.
HALL: And that explains about 30 percent of the pay gap that we measure.
DUBNER: Okay, a third of the gap can be explained by returns to experience. You said about 20 percent of the gap can be explained by time and location of work. But that leaves almost half that can be explained by the third factor. What is that?
LIST: That’s right. So after we account for experience now we’re left scratching our heads. So, we’re thinking, “Well, we’ve tried discrimination. We’ve done where, when. We’ve done experience. What possibly could it be?” What we notice in the data is that men are actually completing more trips per hour than women. So this is sort of a eureka moment.
DUBNER: They’re driving faster, aren’t they?
HALL: Yeah. So the third factor, which explains the remaining 50 percent of the gap, is speed.
DIAMOND: So men happen to just drive a little bit faster, and because driving a little bit faster gets you to finish your trips that much quicker, and get on to the next trip, you can fit more trips in an hour, and you end up with a higher amount of pay.
DUBNER: Now how did these Uber driver data for male/female speed compare to male/female driver speed generally? Do we know for a fact that men generally drive faster than women?
LIST: Yeah, what you find is that in the general population men actually drive faster than women.
DUBNER: Okay, so male Uber drivers drive faster than female Uber drivers, and therefore that helps them make more money. Is that, however, more dangerous, the faster driving?
DIAMOND: So the gap is small — men drive about 2 percent faster than women. So it doesn’t suggest that that’s leading to big differences in risk.
DUBNER: But I also did see that the University of Michigan transportation research unit, they looked at a big, nationally representative sample of police-reported crashes, and they did seem to find that females, on average — on a per-mile driven basis — have more crashes than males. In your data, certainly you could — you have all of that data, right? You have miles driven, you have crashes, presumably. Could you look if you wanted to, and see if, on a per-mile basis, women crash more or less than men?
DIAMOND: I haven’t worked with that data. We’ve been just working with the labor-market data. Uber maybe could look at that but that hasn’t been something we’ve worked at.
HALL: We don’t have an answer to that. It’s something that we would like to study, but we do not have any answer to it.
DIAMOND: I think on the flip side of the — if you look at — of the female having — women having more accidents, I think men have more fatal accidents. So there’s sort of a quality/quantity tradeoff, so it’s not clear who’s actually driving safer. One thing I can say is we’ve looked at, like, the ratings of customers on faster versus slower rides. And if anything, it’s marginally correlated with a higher rating. So it looks like riders do value getting there faster.
DUBNER: So in summary, this is a labor ecosystem — Uber drivers — that would seem to remove all gender discrimination, and yet women earn 7 percent less for doing essentially the same work.
DIAMOND: I mean, I think they’re not doing the same, right? That’s what we’re showing, they’re doing different — they’re making different choices in the labor market. I think it’s — really the whole point is that they’re not doing the same. And once you control for the differences, they are paid the same.
LIST: That’s right. We’ve stripped away all of the factors that we thought were underlying determinants of the gender pay gap, and we go to this new vibrant gig economy that promises worker flexibility and labor flexibility and equal pay for equal work. When you analyze the mounds and mounds of data, it ends up that we have a 7 percent difference. Now, what’s interesting and intriguing is that after you unpack those differences, what you find is that there are perfectly reasonable explanations for what’s happening on the Uber platform.