Bias fears over Uber academic research programme

Uber is facing questions over its joint papers with academic economists, with critics warning that selective data sharing and conflicts of interest could be skewing research to support the ride-hailing company’s business model.

Since 2015, university economists have collaborated with Uber employees, often using the company’s data, on at least 10 papers.

While not all findings have been positive for the company, these papers have concluded variously that Uber drivers are relatively satisfied (albeit anxious) and benefit from Uber’s “flexibility”, and that the company’s controversial “surge pricing” makes economic sense.

Concerns about the collaborations were raised last month by Hubert Horan, an independent transport economist, who accused the company of co-authoring papers to “back the company’s PR and lobbying strategy” using scholarly credibility.

Writing on a University of Chicago Booth School of Business blog, Mr Horan argues that many of these papers are based on inaccessible “proprietary data” and perform “non-replicable analysis”. He dissects what he sees as flaws in four key articles whose Uber-friendly claims were nonetheless reported by mainstream media outlets.

The concerns play into broader worries that technology platforms are distorting research by allowing only sympathetic academics to access data.

“They are explicit that they are only working with people who are going to be congenial,” said Lawrence Mishel, a distinguished fellow at the Economic Policy Institute, a union-affiliated thinktank based in Washington DC. “I’ve had that conversation with them.”

Speaking anonymously, another economist who worked on a paper with Uber told Times Higher Education that the firm was unlikely to approve a potentially damaging project.

Luigi Zingales, a professor of entrepreneurship and finance at Chicago who has warned against corporate “capture” of academia, said the data a researcher can get from Uber “are really cool”.

But if given access to such information, “will I avoid writing other papers…that will piss off Uber? Probably so. I’m not doing anything dishonest, but the result for the literature is terrible,” he said.

A THE analysis of economic research papers on Uber’s website discovered that potential conflicts of interest for university-based co-authors were commonplace, with academics having current or past jobs at the company, holding Uber stock or later leaving academia for the firm.

One author of a paper supporting Uber’s surge pricing, where riders are charged more during peak periods, works simultaneously as an associate professor at Cornell University and an Uber data scientist. Peter Frazier, who wrote the paper with two other Uber employees, said that the findings had gone through a “rigorous peer-review process”.

John Horton, an economist then based at New York University, co-authored a paper with two Uber employees while his spouse was an Uber employee and shareholder. He had also briefly been a paid visiting economist at Uber.

He told THE that his results had actually been used against Uber, and that the underlying data would eventually be released to increase credibility. “This is not to say I don’t appreciate the concern about bias given the policy relevance of Uber,” he said. “One possibility is just not doing this kind of research, but I’m not sure that makes the world more informed.”

Keith Chen, associate professor of economics at the University of California, Los Angeles, co-authored a paper on “the value of flexible work: evidence from Uber drivers” while holding Uber stock options as a former employee.

Dr Chen said he had disclosed his status as a former employee to journals and reviewers. Uber “had no rights to block publication if they did not like our findings or conclusions” and had “agreed that any research economist can replicate our findings on Uber’s data with the help of an Uber data scientist”. “On balance we all thought it helped the work that we were evaluating systems that I helped design and build, and thus understood well,” Dr Chen said.

Not all critics saw financial conflicts of interest as inherently problematic. But Professor Zingales, speaking generally, said that economists sometimes had too much “arrogance” to realise conflicts of interest could sway them – despite researching the impact of such incentives on others. “They cannot see themselves in the mirror,” he said.

Elizabeth Mishkin, senior economist at Uber, said that company employees were included as co-authors “when their intellectual contribution merits this”, for example, in selecting “what data are available or what natural experiments we could leverage”.

“However, it is up to the external academic researcher to decide on the appropriate methodology and interpret the results,” she said.

All papers using Uber data “clearly disclose potential conflicts of interest” and it was “simply not true” to say that Uber would support only projects that were likely to be beneficial for the company.

Uber will also amend its website, said Dr Mishkin, after THE pointed out that “abstracts” of its research did not match the real abstracts in linked papers, and in some cases appeared to draw out Uber-friendly conclusions more explicitly.

“I personally drafted those synopses to be accessible for a less technical audience,” she said, but acknowledged it was “wrong” for “these synopses to be represented as abstracts”. This was the result of an “error in labelling introduced by the website creator”.

david.matthews@timeshighereducation.com