The struggles people of color have when trying to hail a cab have been well-documented — taxis pass by the upraised hand of a black man far more often than a white man. Uber and Lyft were supposed to help level the playing field. But has discrimination just moved from street corners to behind the scenes? A new study indicates that maybe it has.
Researchers at Stanford University, the Massachusetts Institute of Technology and the University of Washington sent research assistants (RAs) on nearly 1,500 Uber, Lyft and Flywheel rides in Seattle and Boston to see how and whether racial discrimination in the ride-sharing economy is a thing. Not surprisingly, it is.
Study 1: Seattle
The first study in Seattle was designed to look at whether there was a difference in the overall amount of time black people and white people needed to get where they were going using ride-sharing companies. (To establish a baseline, students tried to hail taxis on the street. They found that the first taxi they saw stopped 60 percent of the time for white RAs but less than 20 percent of the time for black RAs.)
"Initially what we had been focused on was trying to gather time measurements over numerous rides," says study co-author Stephen Zoepf, executive director of the Stanford University Center for Automotive Research.
The researchers identified key points along the experience: when the ride request was made, when it was accepted, when the driver showed up and when the ride ended. They found that African-Americans waited 30 percent longer for rides than whites when using Uber. But there was little difference in waiting times for Lyft and Flywheel customers of both races.
Why might this be? Uber drivers only see a passenger's location and star rating before accepting a trip. After acceptance, a passenger photo and name show up. By contrast, Lyft drivers see a passenger's name and photograph up front. Flywheel, an app that works with existing taxi services, provides no photos of customers.
The longer wait time for Uber might be attributed to something that the researchers noticed over the course of the Seattle study. "The students started to report that some of the rides were being canceled," Zoepf says. This would mean that new drivers would have to be assigned, increasing the wait time for a ride.
Study 2: Boston
So, when the study moved on to Boston, it was tweaked to look into cancelation rates. The researchers found that on Uber, people with "African-American sounding names" (like Aisha and Hakim) had double the cancelation rates of people with "white-sounding names" (like Allison and Brendan) — 10.1 percent versus 4.9 percent. The rate was even higher for men with "African-American sounding names": 11.2 percent versus 4.5 percent.
For Lyft, the cancelation rate was 6 percent for people with "African-American sounding names" and 7.7 percent for people with "white-sounding names". (While the cancelation rate for men of both races was about the same, the cancelation rate for black women was much lower than for white women on Lyft).
Although Uber has a policy of letting go of drivers who cancel rides too often, Don MacKenzie, study co-author, noted that a lot of Uber drivers got around that by simply not showing up.
"We ended up calling these de facto cancelations," says MacKenzie, an assistant transportation engineering professor at the University of Washington. "A driver did not officially cancel the trip. They would just sit there and not bother to come and pick up. Or in some cases would even drive away in the opposite direction."
Tracking drivers is easy. The apps have a map that allows riders to see where drivers are. The research assistants who were left high and dry were told to cancel the trip themselves after 20 minutes. "I suspect that's what the drivers were trying to do because if you cancel the trip as a passenger you pay a fee and it's not going to count against the driver," he says.
So, what's a company to do with this information? Researchers stressed that both Uber and Lyft have antidiscrimination policies and that the discrimination seems to be an individual action. But, they did offer up suggestions for ways the companies could step up antidiscrimination efforts. These include using pass codes rather than names to identify passengers, increasing disincentives for drivers to cancel and performing driver audits.
"One of the other things that really surprised me," says Mackenzie, "is how much stronger the evidence for discrimination and the impact of discrimination seemed to be on the Uber platform as opposed to the Lyft platform. It was surprising because you think if you convey less information about race and ethnicity up front, then they'd be less likely to discriminate. The charitable interpretation is that having a name and photo puts people at ease.
"The alternative explanation is that there's at least a subset of drivers who would screen out or decline trip requests from black passengers. Because that information is presented up front and is so fast, the discrimination occurs so quickly and seamlessly and efficiently that the trip request gets passed to another driver. So it might be that discrimination is happening [on Lyft], but a person can't tell that they've been discriminated against and the car still gets them to their destination just as quickly.”