Since Uber and Lyft transformed urban transportation with ride-hail services, there’s been a big outstanding question: How much do drivers make?

Dueling papers released Monday—one from Cornell University commissioned by the ride-hail companies, and one by researchers from The New School and UC Berkeley commissioned by the city of Seattle—will do little to resolve the question. One says the average Seattle driver makes $23.25 an hour. The other says $9.73.

Why the big disparity? Driver pay is complicated. One of ride-hail’s sharpest innovations—off-loading many costs associated with a car service, including car payments, some insurance, and fuel—makes it even more so.

Seattle officials commissioned the New School–Berkeley study to help them set a minimum wage for ride-hail drivers there. The study proposes a wage of $28.19 per hour, which the authors say would allow drivers to earn the city’s minimum wage of $16.39 after accounting for their expenses. Minimum-wage laws for drivers, and driver pay in general, is crucial to the ride-hail companies’ finances as they seek to prove out the basic economics of the business.

The Cornell paper examines the pay and expenses of more than 14,100 Seattle Uber and Lyft drivers during one week in October 2019, using data from the ride-hail companies. The researchers say that their access to real (and anonymized) driver data allowed them to make more-precise calculations about how much drivers earn per hour. Their conclusion—$23.25 an hour, even after accounting for expenses—is well above most other researchers’.

“I was very surprised,” says Louis Hyman, an economic historian at Cornell’s School of Industrial & Labor Relations who worked on the paper. “That’s pretty good money, especially if you compare it against service economy jobs.” In Seattle, the paper estimates, a full-time driver would make a median annual income of $47,400, close to the city’s median income for all occupations, $52,945.

But the devil is in the details, and Hyman and his team made what he acknowledges are some controversial decisions when building their model. For example, they excluded many costs associated with owning a car, such as insurance. Their reasoning: 96 percent of the drivers on the apps drove less than 40 hours a week for work, so they would have faced these costs anyway. The researchers did account for depreciation, some maintenance costs, and fuel.

Uber and Lyft commissioned the Cornell study; Hyman and the companies say the money helped offset some research expenses. The companies also provided feedback on initial drafts. The paper has not been peer-reviewed, though the authors say they intend to eventually publish it in an academic journal.

Drivers who spoke to WIRED were skeptical. “$23.25 an hour is BS. There’s no way,” says Sergio Avedian, who drives part-time in Los Angeles and coaches other ride-hail drivers on how to make the most of their gig. “If people were making that money, everyone and their uncle would be driving for Uber.”

Avedian says wages are higher in Seattle than Los Angeles, where he lives, in part because the market is smaller. But he takes issue with the Cornell researchers’ decision to dismiss some repair and insurance costs. Avedian advises drivers to buy insurance specifically designed for ride-hail drivers, which can cost more than $100 a month. He estimates he takes home about $20 an hour, but only by obsessing over details like which rides to accept and decline. He thinks most drivers make minimum wage or less.

“To make more than $23 an hour, you not only have to be good. You have to be lucky,” says Jay Morran, who drove for both companies in the Los Angeles area until early this year.

The second paper, commissioned by Seattle and written by economists James Parrott and Michael Reich, calculated driver pay very differently. It’s based on an online survey completed by more than 6,500 licensed Seattle ride-hail drivers and reflects one week in December 2019. Whereas the Cornell work calculates driver expenses at 19 cents a mile, the Parrott and Reich one pins it closer to 52 cents a mile, including insurance, a cell phone, and vehicle cleaning. “This is the biggest source for how [the Cornell researchers] came up with net pay being so huge: They’re just not accounting for business costs,” says Reich, an economics professor at Berkeley’s Institute for Research on Labor and Employment.