norberthaering (en-GB)

How Uber money dominates and distorts economic research on ride-hailing platforms

Ride-hailing platform operator Uber is often accused of undermining labour market regulations and of overpricing at times of peak demand by “surge pricing”. Uber defends itself against such accusations not only by using high-profile lobbyists, but also with the help of top-notch economists, who cooperate in exchange for exclusive data and lucrative consultancy assignments. Even reputable journals publish such sponsored analysis as if it were scholarly research.

recent paper by renowned MIT professor Joshua Angrist with Sydnee Caldwell and Uber Research Director Jonathan V. Hall comes to the favourable conclusion that drivers benefit greatly from the fact that Uber exists. Also Princeton economist Alan B. Krueger, former chairman of the Council of Economic Advisors of the US president, wrote a paper on Uber, under contract from Uber, together with Uber’s Mr Hall: an essay entitled “An Analysis of the Labor Market for Uber’s Driver Partners in the United States“,which makes the company appear to be a very good employer.

The survey of “Uber driver partners” on which they heavily rely, was done by a firm hired by Uber for the purpose of producing a PR-brochure. The way the research was done, and used by Kruger and Hall, is highly problematic. A potential bias, which is probably quite severe in the case of Uber with its high rate of churn, is not addressed: the survivorship bias. Dissatisfied drivers are likely to have left Uber before this satisfaction survey was taken.

Kruger and Hall do not mention that Uber cut driver compensation in many markets after the end of the investigation period. Kruger, when asked to comment, said that there was no data on that (at the time).

Uber used to publish regular earnings evaluations for New York drivers. However, despite inquiry, it has offered evaluations only up to 2015, the year in which the pay cuts started.

Krueger says he assumes that the drivers would not earn less because of such cuts. In fact, a paper exists which delivers precisely this convenient result. It is co-written by John Horton of New York University and – you guessed it – Uber economists Jonathan Hall and Daniel Knoepfle.

Data from interested parties

Krueger wrote a second paper on Uber (this time not under contract with Uber) with his Princeton colleague Judd Cramer and placed it in the May 2016 issue of American Economic Review. The main takeaway: Uber “driver partners” are used much more than normal taxi drivers. The data on Uber drivers was collected by the company, which could be considered problematic. Uber has a strong interest in the outcome, which finds that rival taxi firms are outperformed by higher efficiency rather than dumping prices, as might be inferred from the billions in losses that Uber posts every year.

In a literature survey called The Ridesharing Revolution: Economic Survey and Synthesis on Uber research for the book More Equal by Design, Robert Hahn and Robert Metcalfe from the Universities of Oxford and Chicago conclude that Uber probably has positive effects overall. As evidence they cite the study Cohen et al (2016). According to this study, Uber users realised a huge consumer surplus, which would be lost if such services were prevented by regulatory means. Only a footnote mentions that Cohen works for Uber, but readers of Hahn’s and Metcalfe’s chapter would not have learned, that Hahn and Metcalfe themselves are co-authors of Cohen et al, together with Uber-economist Hall and Steven Levitt, the author of the best seller Freakonomics.  Asked for a comment on this, Hahn said that he would take care of the disclosure aspects the inquiry had pointed out, when working on the galley proofs for his chapter of the book.

Uber is also busy commissioning research to be published in non-US academic journals. For example, the former head of the German Monopolies Commission, Justus Haucap, co-authored a survey on mobility markets with co-authors from Uber and under contract with Uber, which was published in 2015, among other places, in the journal List Forum, which conveniently is edited by Haucap. In Wirtschaftsdienst 2/2015, a contribution from Haucap appeared that was very positive about Uber and other companies in the sharing economy. There was no reference to Haucap’s contract work for Uber. When questioned about this Haucap said the disclosure had inadvertently been omitted.

Neither I nor an Uber spokesman could find a paper with Uber involvement that did not result in convenient conclusions, which Uber could use to defend itself against public criticism. This shows how problematic it is that Uber, with its financial power, dominates economic research on ride-hailing platforms.

Economic Journals play along

It is not uncommon for a company blessed with many billions of dollars of investors’ money to use commissioned “scholarly” studies for its public relation goals. Unlike most such commissioned studies, however, the studies commissioned by Uber are disseminated through renowned scientific publication channels – to which the prominent and well-connected editors and authors have access. In this way, texts that border on company PR are dressed up as objective scholarly findings. After all, they appear in the prestigious series “NBER Working Papers”, or reputable journals such as Industrial & Labor Relations Review (ILRR) or in the American Economic Review. Even if the results are unassailable, which is not always the case, it is Uber who is deciding who receives which exclusive data for what purpose. When the results of these papers are cited by other researchers, quite often the important fact of Uber involvement, which might have been mentioned in the original, is omitted.

Chicago economist Levitt said in a podcast, in which he talked about his Uber-related research: “I love Uber,” and waxed lyrical that Uber had created an ideal market from the perspective of economists. It was clear that Uber did not have to worry about overly critical interpretation of exclusive data when Levitt approached them to gain access to it for his research. Nevertheless, Levitt reports, the company did one year refuse to provide the data, because he denied Uber a right of veto on the publication of the results. “In this case, everything was completely kosher,” he concludes. That is, however, only one way of looking at it. After all, two Uber economists did act as his co-authors. They had a whole year to check and select the data that Levitt would eventually see, and ensure that it led to the desired result.

The head of the National Bureau of Economic Research (NBER), James Poterba, acknowledged that, when companies pay for studies or control access to data, it does raise difficult issues. The NBER makes sure that conflicts of interest are revealed and relies on the integrity of the associated researchers. Alan Krueger seems to see less of a danger that Uber might be able to bias research in their own favor if they are collecting and providing the data and their employees participate in the analysis. In response to my critical questions, Prof. Krueger starts by stating: “I am glad you are writing on this topic since many important studies by Joshua Angrist (professor at MIT),  Steven Levitt (Clark Prize winner and professor at University of Chicago), John List (University of Chicago) and other scholars would not be possible without Uber's involvement. It would be good if you could point out that other companies should also follow Uber’s example and make data available to researchers.”

Krueger, who has always correctly disclosed his financial relationship with Uber, emphasises that he had contractually full control over what he writes. However, there is no such statement by his co-author Hall, who is an Uber employee and shareholder. Krueger explains that he could have cut out Hall as a co-author at any time if they had disagreed about something. Whether removing the name of the co-author of a scientific text is compatible with scientific ethics is questionable, though. Lawrence Kahn, the editor of the ILRR Review, in which the text of Hall and Krueger was published, was also invited to comment on this issue. He had no answer, apart from inviting me to write a comment for ILRR, which would then be examined by reviewers for publication.

Change notice (12 December 2017): An edited version of this article provided by Brave New Europe, which crossposted it, replaced the original article.

Change Notice (31.12.2017/1.1.2018): Prof. Krueger wrote a second paper ON Uber, not FOR Uber. The data in his study was not collected by drivers but by the Uber-team via the Uber app. I apologize for these errors, which slipped in in the process of translating the original German blogpost. In paragraph three I have toned down the statement that the paper of Hall and Krueger does not adress response bias at all. I apologize for the exaggeration in the original post. In paragraph 9, I have changed "buying favorable research" into "comissioning research" to bring the text in line with the German original. In paragraph 4, I have eliminated the word "reliable" in front of "data", as Prof. Krueger had not used that word in his response. In the second to the last paragraph I have changed the following phrases: “Alan Krueger, however, sees no problem in the fact that Uber can choose who it gives access to which data, potentially ensuring that a scientist who is critical will always be at a severe disadvantage to an Uber friend, like Mr Levitt. For Krueger, Uber provides an example that should be emulated by other companies”. The changes are meant to make it clear that Prof. Krueger’s remark regarding Uber being a good example was not made in response to a specific question regarding the ability of Uber to be selective in providing data.

profilbild

________________