 Welcome to the latest edition of the MIT Sloan Expert Series. I'm your host, Rebecca Knight. Our topic today is racial bias in the sharing economy, how Uber and Lyft are failing black passengers and what to do about it. Here to talk about that is Chris Knittle. He is a professor of applied economics here at MIT Sloan. And he's also the co-author of a study that shows how Uber and Lyft drivers discriminate based on a passenger's skin color. Thanks so much for joining us. Oh, it's great to be here. Before we begin, I want to remind our viewers that we will be taking your questions live on social media. Please use the hashtag MIT Sloan Expert to pose your questions on Twitter. Chris, let's get started. So there is a lot of research that shows how difficult it is to hail a cab, particularly for black people. Uber and Lyft were supposed to represent a more egalitarian travel option, but you didn't find that. That's right. So what we found in two experiments that we ran, one in Seattle and one in Boston, is that Uber and Lyft drivers were discriminating based on race. And we've already seen, actually, some racial evidence of racial discrimination in the sharing economy, not just with ride sharing apps. Sure. So there's evidence for Airbnb. And what's interesting about Airbnb, actually, is the discrimination is two-sided. So not only do white renters of properties not want to rent to black renties, but white renters do not want to stay at a home of a black homeowner. So do your findings and the findings of that other research you just talked about, does it make you discouraged? Partly. So I was an optimist. We went into this, at least I went into this, hoping that we wouldn't find discrimination. But one thing that has helped, or at least put, shined a more positive light, is that there are ways that we can do better in this sector. You've talked about this study, which you undertook with colleagues from the University of Washington in Stanford, shows the power of the experiment. Can you talk a little bit about what you mean by that? Sure. So what we did was actually run two randomized control trials. So just like you would test whether a blood pressure medication works, you would have a control group that wouldn't get the medication and a treatment group that would, we did the same thing where we sent out in Seattle, both black and white RAs that hailed Uber and Lyft rides. And we randomized whether or not it was a black RA calling the ride or a white RA at that particular time. And they all drove the same exact route at the same exact times of the day. So what did you find? Let's talk about first what you found in Seattle. Sure. So in Seattle, we measured how long it took for a ride to be accepted, and also how long it took for, once it was accepted, for the driver to show up and pick up the passenger. And what we found is if you were a black research assistant that inhaling an Uber ride, it took 30% longer for a ride to be accepted and also 30% longer for the driver to show up and pick you up. 30% seems substantial. Well, so for the time it takes to accept a ride, we're talking seconds. But for the time it takes for a passenger to actually be picked up, it's over a minute longer. And I'll mention also for Lyft, we found a 30% increase in the amount of time it took to be accepted. But there was no statistically significant impact on how long it took for the driver to actually show up. So the thing about the minute difference, I mean, that can be material, particularly if you're trying to catch a cab, an Uber, or a Lyft for a job interview or to get to the airport. Yeah, this is introspection. But I always seem to be late. So even a minute can be very costly. I hear you. I hear you. So why do you think there was the difference between Lyft and Uber? So what's interesting, and we learned this while we were doing the experiment, a Lyft driver sees the name of the passenger before they accept the ride, whereas an Uber driver only sees the name after they've accepted. So in order for an Uber driver to discriminate, they have to first accept the ride and then see the name and then cancel, whereas a Lyft driver can just pass it up right away. So it turns out because of that, the Lyft platform is more easily capable of handling discrimination because it pushes it to another driver faster than the Uber platform. I want to come back to that. But I want to say, also, that difference caused you to change the way you did the experiment in Boston. So in Boston, a couple of differences. One is that we sent out RAs with two cell phones, actually. So each RA had an Uber and Lyft account under a stereotypically white-sounding name, and then also an Uber and Lyft account under a stereotypically black-sounding name. So that was one difference. And then also what we measured in Boston that we didn't measure in Seattle is cancellations. So an Uber driver accepts the ride and then cancels on the RA. Let's go back to the stereotypically black-sounding name versus white-sounding name. You're an economist. How did you determine what those names are? So we relied on another published paper that actually looked at birth records from the 1970s in Boston. And the birth records tell you not only the name, but also the race of the baby. So they found names that actually 100% of the time were African-American or 100% of the time were not African-American. So we relied on those names. And the names were? So you could imagine Jamal, for example, compared to Jerry. All right. OK, Ayesha and Allison. Sure. Yeah. So what was your headline finding in Boston? So in Boston, what we found is if you were a black male calling an Uber ride, that you were canceled upon more than twice as often as if you were a white male. And what about Lyft? So for Lyft, there is no cancellation effect. And that's not because there's no discrimination. It's just that they don't have to accept and then cancel the ride. They can just pass up the ride completely. So it's actually a nice little experiment within the experiment. We shouldn't find an effect of names on cancellations for Lyft. And in fact, we don't. And also, the driver network is much thicker in Boston than in Seattle. So in Boston, although we found this cancellation effect, we didn't find that it has a measurable impact on how long you wait. And this is somewhat speculation. But we speculate that that's because the driver network is so much more dense in Boston that although you were canceled upon, there's so many other drivers nearby that it doesn't lead to a longer wait time. How do you think what you found compares to hailing traditional cabs? We started our conversation talking about the vast body of research that shows how difficult it is for black people to hail cabs. Yeah, we are quick to point out that we are not at all saying that Uber and Lyft are worse than the traditional status quo system. And we want to definitely make that clear. In fact, in Seattle, we had our same research assistant stand at the busiest corners and hail cabs. And what we found there is if you were a black research assistant, the first cab passed you 80% of the time. But if you were a white research assistant, it only passed you 20% of the time. So just like the previous literature has found, we found discrimination with the status quo system as well. You've talked to the companies about your findings. What has the response been? So that's been actually heartening. So both companies reached out to us very quickly. And we've had continued conversations with them. And we're actually trying to design follow-up studies to minimize the amount of discrimination that's occurring for both Uber and Lyft. But those are off the record. So we're not talking specifics. But what I can say is that the companies understand this research and they definitely want to do better. In fact, the companies both have issued statements about this. The first one is from Lyft. We are extremely proud of the positive impact Lyft has had on communities of color. Because of Lyft, people living in underserved areas, which taxis have historically neglected, are now able to access convenient affordable rides. And we provide this service while maintaining an inclusive and welcoming community and do not tolerate any form of discrimination. Uber has also responded. Ride-sharing apps are changing a transportation status quo that has been unequal for generations, making it easier and more affordable for people to get around. Discrimination has no place in society and no place on Uber. We believe Uber is helping reduce transportation inequities across the board, but studies like this one are helpful in thinking about how we can do even more. So let's talk about solutions to this. What do you and your colleagues who undertook this research suggest? So we've been brainstorming. We don't know for sure if we have the silver bullet. But a few things could change. For example, you could imagine Uber and Lyft getting rid of names completely. We realize that that has a trade-off in the sense that it's nice to know the name of the driver. Sure, you can strike up a conversation. It makes it more social, it makes it more personal, more peer-to-peer, if you will. But it would eliminate the type of discrimination that we uncovered. Another potential change is to delay when you give the name to the driver, so that the driver has to commit more to the ride than he or she previously had to. And that may increase the cost of discrimination. So that would be changing the software or? Right, so you could imagine now, like I said, with Lyft that you see the name right away. Maybe you wait until they're 30 seconds away from the passenger before you give them the name. What about the dawn of the age of autonomous vehicles? Might that have an impact? We already know that Uber is experimenting with driverless cars in Pittsburgh and Arizona. That would obviously solve it. So that would take the human element out of things. And it's important to point out that these are the drivers that are deciding to discriminate. So if provided you didn't write the autonomous vehicle software to discriminate, you would know for sure that that car is not going to discriminate. What about a driver education campaign? Do you think that would make a difference? I'm reminded of an essay written by Doug Glanville, who is an ESPN commentator and former pro ball player. He writes on talking about his experience being denied service by an Uber driver. The driver had concluded I was a threat, either because I was dangerous myself or because I would direct him to a bad neighborhood or give him a lower tip. Either way, given the circumstances, it was hard to attribute his refusal to anything other than my race. Shortly after we walked away, I saw the driver assisting another passenger who was white. Well, we all hope that information helps and eliminates discrimination. It's certainly possible that Uber and Lyft could have a full information campaign, where they show the tip rates for different ethnicities. They show the bad ride probabilities for different ethnicities. And my hope is that once the drivers learn that there aren't differences across ethnicities, that the drivers would internalize that and stop discriminating. Policy. Senator Al Franken has weighed in on this urging Uber and Lyft to address your research. Do you think that there could be policies too? Does government have a role to play? Potentially. But what I'll say, again, is that Uber and Lyft, I think, have all the incentive in the world to fix this and that they seem to be taking active steps to fixing this. So what could policy makers do? They can, obviously, it's already outlawed. They could come down and maybe potentially find the companies if there's more evidence of discrimination. But I would at least allow the companies some time to internalize this research and respond to it and see how effective they can be. Many think tanks and government advocacy groups have weighed into the MIT Sloan Expert Series recently sat down with Ava Malona of the Massachusetts Immigrant and Refugee Coalition. She will talk about this research in the context of immigration. Let's roll that clip. We're an advocacy organization and we represent the interest of foreign born and our mission is to promote and enhance immigrant and refugee integration. So anecdotally, yes, I would say that the research and given the impressive sample of the research really leads to such belief that, you know, discrimination is still out there and there is a lot that needs to be done across sectors to really address these issues. We are really privileged to live in such fantastic Commonwealth with, you know, the right leadership and all sectors together to really making our Commonwealth a welcoming place and I do want to highlight the fantastic role of our Attorney General for standing up for our values. But, you know, Massachusetts is one state and it could be an example but the concern is nationwide given a very divisive campaign and also not just the campaign but also what is currently happening at the national level that the current administration is really rejecting this welcoming effort and the values of our country as a country who welcomes immigrants. All sectors needs to be involved in an effort to really making our society a better one for everyone and it's going to take political leadership to really set the right tone, set the right message and really look into the integration and the welcoming of the newcomers as an investment in our future of our nation. Uber and Lyft have an opportunity here to provide leadership and come up with promotion of policies and that integrate the newcomers or that are welcoming to the newcomers, provide education and training and train their people and as troubling as this, as a result of this research are, we like to believe that this is the attitude of the drivers but not really what the corporate represents so we see an opportunity for the corporate to really step in and work and promote policies of integration, policies of improvement and betterment for the whole society and provide an example. Let me say thank you to Professor Kanido for his leadership and MIT for always being a leader and looking into these issues but if we can go deeper into A, the size, B, the geography but also looking into a wider range of all communities that are represented in looking into the Latino community, looking into the Arab communities and other parts of the nation and the more rigorous, more deep and the larger size of research will be very helpful in terms of promoting better policies and integration for everybody who truly is America to be their home. That was Eva Malona of the Massachusetts Immigrant in Refugee Advocacy Coalition. Chris, are you confident this problem can in fact be remedied? I think we can do better for sure and I would say we need more studies like what we just performed to see how widespread it is. We only study two cities. We also haven't looked at all how the driver's race impacts the discrimination. Now we're gonna turn to you. Questions from our viewers. Questions have already been coming in this morning and some overnight. Lots of great ones. Please use the hashtag MIT Sloan Expert to pose your question. The first one comes from Justin Wang who is a Sloan and MIT Sloan MBA student. He asks, what policies can sharing economy startups implement to reduce racial bias? Well, I would say the first thing that is to be aware of this. I think Uber and Lyft and Airbnb potentially were caught off guard with the amount of discrimination that was taking place. So the research that we've performed in the research on Airbnb gives new startups a head start on designing their platforms. Just knowing that this is an issue. Knowing it's an issue and potentially designing their platforms to think of ways to limit the amount of discrimination. Another question, did you look at gender bias? Do you have any indication that drivers discriminate based on gender? So we did look at gender bias. The experiments weren't set up to necessarily nail that. But one thing that we found, for example, in Boston is that there is some evidence that women drivers were taken on longer trips. Again, both the male and the female RAs are going from the same point A to the same point B. That was a controlled part of the study. That was a controlled part of the experiment. And we found evidence that women passengers were taking on longer trips. And in fact, one of our RAs commented that she remembers going through the same intersection three times before she finally said something to the driver. And you think, so you didn't necessarily study this as part of it, but do you have any speculation, conjecture about why this was happening? Well, there's two potential motives. One is a financial motive that by taking the passenger on a longer drive, they potentially get higher fare. But I've heard anecdotal evidence that a more social motive might also be at play. For example, I have a colleague here at Sloan who's told me that she's been asked out on dates three times while taking Uber and Lyft rides. So drivers taking the opportunity to flirt a little bit. Sure. Another question, can you comment on the delete Uber, the hashtag delete Uber campaign? This, of course, is about the backlash against Uber, responding that it was intending to profit from President Trump's executive order, the banning immigrants and refugees from certain countries from entering the United States. Uber maintains that its intentions were misunderstood, but it didn't stop the hashtag delete Uber campaign. Yeah, I haven't followed that super closely, but to me it seems like Uber is getting a bit of a bad rap. One potential reason why they allowed Uber drivers to continue working is that maybe they wanted to bring protesters to the airports to protest. So from that perspective, actually having Uber and Lyft still in business would be beneficial. Another question, did your study take into account the race of the drivers themselves? So we actually were not allowed to. So anytime you do a randomized control trial in the field like this, you have to go through a campus committee that approves or disproves the research. And they were worried that if we collected information on the driver that potentially Uber and Lyft could go back into their records and find the drivers that discriminate and then have penalties assigned to those drivers. So it just wouldn't be allowed to? At least in this first phase. Yeah, they didn't want us to collect those data. Last question, we have time for one more. Why aren't there more experiments in the field of applied economics like this one? That's a good question. That's a great question. And in fact, I think many of us are trying to push experiments as much as possible. My other line of research is actually in energy and climate change research. And we've been designing a bunch of experiments to look at how information impacts consumers' choices in terms of what cars to buy, how it impacts their use of electricity at home. And experiments randomized control trials actually started in developmental economics, where MIT has actually pioneered their use. And again, it's the best way to actually test, the most rigorous way to test whether intervention actually has an effect because you have both the control group and the treatment group. So why aren't they done more often? Well, it's tough. Often you need to find a third party. For example, we didn't need a third party in the sense that we could just send RAs out with Uber and Lyft. But if we wanted to do anything with the drivers, for example, an information campaign, or if we wanted to change the platform at all, we would have needed Uber and Lyft to partner with us. And that can sometimes be difficult to do. And also experiments, let's be honest, are pretty expensive. Expensive because you obviously weren't partnered with Uber and Lyft for this one. Right, but we had research assistants take 1,500 Uber and Lyft rides. So we had to pay for each of those rides, and we also had to give them an hourly rate for their time. Well, Chris Connell, thank you so much. This has been great talking to you, and you've given us a lot to think about. It's been fun. Thanks for having me. And thank you for joining us on this edition of the MIT Sloan Expert Series. We hope to see you again soon.