The study, led by economists, examined data from more than 1.8 million drivers and 740 million Uber trips in the United States between January 2015 and March 2017. The paper, called “Gender Earnings Gap in the Gig Economy: Evidence From Over a Million Rideshare Drivers,” was written by five economists, including two employed by Uber, two Stanford professors, and John List, chairman of the University of Chicago economics department. List is also the chief economist at Uber.

The gender wage gap has persisted—with women making just under 89 cents on a man’s dollar in 2016—even as females have been returning to the U.S. labor force in greater numbers.

But the gig or “sharing” economy has been viewed as an opportunity for equal pay between women and men largely because the lack of flexibility in traditional workplaces has driven much of the disparities in pay. Gig economy jobs, like driving for Uber, are flexible. And that’s led to all kinds of speculation that more job flexibility will narrow the gender pay gap.

The Freakonomics podcast discussed the study’s findings with a few of the economists who worked on it. List says that his assumption going into the project was that the findings would show equal, or close to equal pay.

“If there was a difference, I think the pay gap would slightly favor women,” List said during the podcast. “And this is kind of for two reasons. One, I knew that they had worked fewer hours per week so they had a chance to cherry-pick the better hours during the week. Point number two was, if there was discrimination on the platform, I was thinking that riders would actually prefer female drivers to male drivers.”

The economists found it’s not so simple.

First, it’s important to understand how Uber pays its drivers. This isn’t a traditional workplace where managers are in control of how their workers are paid. Uber provides the tech, or app, that connects drivers with riders. It pays drivers based on a formula that takes into account the length of the ride in miles, how long the ride takes, and, on occasion, the so-called “surge” multiplier, where higher demand can push up rates. The fare is determined by a gender-blind algorithm.

So why are female Uber drivers earning less even when there is no evidence of discrimination? The economists found three explanations: location of pickups, experience, and driving speed.

Here are some of the key takeaways:

• 77% of women quit the Uber platform after 6 months compared to 65% of male drivers.
• Experience accounts for one-third of the earnings gap, according to the paper. Drivers who have taken more than 2,500 trips earn an average of $3 more per hour than those with less than 100 trips. Men, on average, accumulate more experience  by working more hours each week and being less likely to stop driving with Uber.
• About half of the earnings gap is explained by differences in driving speed, according to Uber. The researchers found that, on average, men drive 2.2% faster than women. The researchers noted that there is a positive expected return to driving faster. However they also note returns may turn negative at excessive speeds. This difference in driving speed is not unique to Uber. Data gathered from the National Highway Travel Survey indicates that a gender gap in driving speed exists in the wider population as well.
• The remaining one-sixth of the gap in earnings  is explained by differences in where people choose to drive, with men, on average, driving in locations with higher surge and lower wait times.