Uber, Lyft, and other ride-sharing services were supposed to be a more egalitarian transportation option than a traditional cab service. There’s a mountain of data that shows how difficult and dangerous something as simple as finding a cab ride home can be, particularly for women and black people. Ride-hailing apps were expected to help fix that.

But a new study finds some of the problems persist. The economists Yanbo Ge and Don MacKenzie, of the University of Washington, along with Christopher R. Knittel of MIT and Stephen Zoepf of Stanford, came up with an experiment to test how well these popular apps were serving groups that had been marginalized by the taxi industry. They sent research assistants out as riders in two cities, Seattle and Boston, to hail nearly 1,500 rides using Uber, Lyft, and another service called Flywheel. The riders took screenshots of the apps to chronicle their experience, recording their estimated wait times, whether a driver accepted them, and where and when they were picked up and dropped off. Through these trials the economists were able to gather data on how the three services differed in their treatment of gender and race. At the end of the trials, the economists found “significant evidence of racial discrimination,” meaning that black riders faced longer wait times and more frequent cancellations than white riders.

The data the researchers collected was especially damning for Uber. In Seattle, black riders who requested a ride using either Uber and Lyft waited substantially longer for their ride to be confirmed—they waited about 16 to 28 percent longer than white riders did. For black Uber riders in particular, wait times in Seattle were as much as between 29 to 35 percent longer. In the Boston experiment, black Uber riders were much more likely than white riders to have a driver cancel on them after confirming, and the effect is especially pronounced for black men, whose cancellation rate was three times as high as white males. And riders with “black-sounding names” were significantly more likely to be canceled on than either white riders ors black riders with “white-sounding names.” This effect was worse in lower-density areas, where finding a ride is often harder in the first place.

“Ridesharing apps are changing a transportation status quo that has been unequal for generations, making it easier and more affordable for people to get around,” Rachel Holt, the head of Uber’s North American operations, wrote in a statement to Bloomberg. “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.”

That Uber generated some of the most troubling data doesn’t necessarily mean that Lyft is friendlier to black customers. Instead, it might reflect the ability of Lyft drivers to discriminate much earlier in the process of finding riders. When riders used Lyft—which allows drivers to see a prospective rider’s name and photo before accepting a trip—it was harder to identify potential discrimination before and during rides, since drivers could simply choose never to confirm black passengers to begin with, the researchers note. But the prolonged wait times that did show up among black Lyft users, the authors say, might be consistent with drivers viewing and skipping over a pending request from a black customer. (The report found that despite delays in waiting for a ride to be confirmed, the data collected about Lyft was too imprecise to determine that delay’s impact on the overall wait time between ride requests and pick-ups.) The study found little evidence of discrimination among Flywheel drivers, a fact that the researchers say might be due to the fact that the service doesn’t have users add photos to their profiles.

Adrian Durbin, a spokesperson for Lyft, said in a statement, “We are extremely proud of the positive impact Lyft has 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.”

The report also found that some women faced discriminatory treatment as well, often in the form of longer and more expensive routes than the ones drivers took with male riders, despite there being preset pick-up and drop-off locations. Women who participated in the study reported drivers who were very talkative on these prolonged routes, which sometimes involved a driver passing through the same intersection multiple times. “The additional travel that female riders are exposed to appears to be a combination of profiteering and flirting to a captive audience,” the authors find. That’s especially troubling given reports of sexual misconduct and harassment on the part of drivers in the past.

Many hoped that arrival of Uber and similar apps would reduce denials of service based on race, a possibility that Doug Glanville, a former pro baseball player and a commentator for ESPN, wrote about in a 2015 essay for The Atlantic. Of one episode at a Los Angeles airport, he wrote,

When the driver saw me approaching to put my luggage in the trunk, he froze and his entire demeanor shifted. His English was not strong, but it was clear. “Go across the street! You take the bus! It is $19!” ...I knew what was going on: 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 low 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.

Glanville said encounters like this helped him understand that many might choose to use ride-hailing apps in hopes of avoiding unfair treatment. But it appears that these platforms do little to mitigate such discrimination.

Sadly, the researchers note, ride-hailing apps aren’t any worse than attempting to hail a cab the old-fashioned way. When the researchers sent their assistants out to catch a ride sans app, they found that white riders never had more than four taxis pass before they were offered a ride, while black riders were usually skipped over by six or seven before finally being offered a ride.

This isn’t to say that Uber or Lyft are doing something discriminatory, but that their platforms can be used by their drivers in ways that are discriminatory. Still, given that this is both a current and a historical pattern, these companies ought to invest in figuring out what more they could do to diminish it.