Algorithmic collusion is a possibility to watch out for
In 2011, a biology book was listed on Amazon for $1.7 million, eventually reaching $23 million due to two sellers’ algorithms reacting to each other’s pricing. This incident underscores the challenges and potential regulatory issues as businesses increasingly rely on algorithms. It has gained attention in legal contexts, including Indian courts, where cases challenge the legality of pricing algorithms used by ride-hailing companies.
This article was first published in The Mint. You can read the original at this link.
In April 2011, Michael Eisen, a biologist at the University of California in Berkeley noticed that Peter Lawrence’s classic on development biology, The Making of a Fly, was selling on Amazon for $1.7 million. The book was, at the time, out of print and being offered by just two sellers, but even this scarcity could hardly account for the astronomical price. After all, this was a dry, academic book about the drosophila fly. Over the course of the next week, Eisen watched as the price kept rising, till it reached a price of $23 million for a brand new copy on Amazon.
As he tracked the prices, Eisen observed something odd about the way in which they were rising. Each day, one of the two sellers of the book would set their price at 0.9983 times the price of the other. Later in the day, the second seller would increase the price of the book to exactly 1.270589 times the price of the first seller. After thinking about this for a while, he came up with the only answer he believed could explain this oddness.
Eisen believed that it was the first seller who actually had the only available copy of the book. In order to maximize profits, this seller set the price just marginally below the market. The second seller did not have a copy of the book, and so had to list a price that incorporated the cost of purchasing the book from the first seller along with a modest markup. The trouble was that both sellers left it entirely up to their respective algorithms to set the price, and so, each day, the algorithms reacted to each other, colluding in a strange counterproductive manner until the cost of the book reached astronomical levels.
The more we incorporate algorithms into our businesses, the more will they challenge the ways in which business has traditionally been regulated. As we increasingly move our business online, courts and regulators will be forced to revisit historical precedents in order to come up with new, more appropriate ways of regulation that address the challenges that technology poses. This will prove to be particularly difficult in fields where algorithms are interacting with each other, as the mistakes they make will be harder to spot and the deleterious outcomes greatly exaggerated.
Until recently, Indian courts hadn’t yet had the opportunity to engage with any sort of collusion involving algorithms, but I knew that it was only a matter of time before they were called upon to do so. That is why it came as no surprise to me when Samir Agarwal, an Indian lawyer, set out to challenge the legality of the pricing algorithms used by app-based ride-hailing companies in India.
At first glance, it seems that Samir’s case was heavily inspired by Spencer Meyer vs. Travis Kalanick, a case brought before the Southern District of New York a few years ago, alleging app-enabled collusion between drivers on Uber. But there were significant differences. In the first place, the US case had not been filed against the company, but against its chief executive officer who, as it happened, was also a registered driver with the service. Since Kalanick was a driver, Spencer Meyer could file the price fixing complaint against him not as CEO but in his capacity as a driver. This avoided the mandatory arbitration clause contained in the terms of service, but more importantly gave him the ability to raise a novel competition argument.
Drivers on Uber have no option but to charge the fares that the algorithm tells them to. While Uber says that drivers can charge less than the suggested fee, in reality there is no mechanism for them to do so in the app. The fares suggested by the app tend, for the most part, to be on par and often lower than those you would have to pay for a traditional taxi. But Uber also has surge pricing—a feature that, in New York, could increase the fare up by up to 10 times the normal rate if taxis are scarce on the roads. This pricing mechanism, Meyer suggested, offers a motivation for drivers to collude among themselves.
The digital marketplace, Meyer argued, gave Uber a platform through which Kalanick was allegedly able to organize a horizontal conspiracy among its drivers. Uber frequently organized picnics and other meetings for its drivers—events where they were exhorted to log themselves onto the app during busy periods to increase their earnings. This, Meyer claimed, was evidence of how Kalanick encouraged drivers to exploit the algorithm to take advantage of higher prices available to them during peak hours.
The facts are somewhat different in India. In the first place, there is no dominant ride hailing platform—the market is mostly split between Ola and Uber. Price discrimination is only relevant when it’s implemented by a dominant player. Not only was it impossible to argue that Uber and Ola were jointly dominant, the court found that they were not even independently dominant.
Samir Agarwal then tried to suggest that the platform allowed drivers to collude among themselves on price. However, once again, the fragmented nature of the market in India defeated that argument. If there is no dominant platform and drivers can (and do) frequently switch between apps, it is impossible for them to collude among themselves.
For all these reasons, the case was dismissed. By cleaving blindly to the form of complaint that had succeeded in the US, Agarwal lost the opportunity to set a precedent. Even though our courts still haven’t passed judgment on collusion using algorithms, it’s only a matter of time.
