Can an algorithm break antitrust law?

More than 20 years ago, executives at rival auction houses Sotheby’s and Christie’s were found guilty of coordinating a massive price-fixing scheme. Leaders from the companies held covert meetings in limousines and hotel rooms where they colluded to set identical commission fees, driving prices up while eliminating competition.

Today, active antitrust cases show that the ways in which companies might conspire are changing. U.S. regulators say algorithms have, in some cases, replaced secret meetings, but it’s still collusion, whether it’s a human or a bot pulling the strings.  

Marketplace’s Lily Jamali spoke to Joe Harrington, professor of business economics and public policy at the University of Pennsylvania’s Wharton School, about how antitrust law holds up against new technology.

The following is an edited transcript of their conversation.

Joe Harrington: The antitrust laws have proven to be incredibly resilient. These laws were passed more than a century ago, and they’ve been able to handle all sorts of harmful practices. But I think we might have kind of stretched the limit in terms of what they can do. And the reason why is because a fundamental source of competition among firms, that is, the prices you set, there’s now the possibility of that being set by a common third party through such a thing as a data analytics company, and that’s fundamentally different.

Lily Jamali: Let’s talk about an example of this. There’s a case involving hotel rooms in New Jersey that is getting a lot of attention right now, including from the Department of Justice and the Federal Trade Commission. Can you explain what that case is about?

Harrington: Absolutely. What we have here is that hotels such as Caesars and Hard Rock and Harrah’s have used the services of a third party, a data analytics company called Rainmaker. They’ve provided their data on hotel prices, vacancies, all sorts of information to Rainmaker. Rainmaker has trained a pricing algorithm that uses that data and then provides recommendations for hotel rates back to those companies. So, what this case is all about is that the plaintiffs in it claim that there is an unlawful agreement between Rainmaker and the hotels that use their services to all adopt Rainmaker’s pricing algorithm, and that this pricing algorithm sets prices above that which would have occurred under competition.

Jamali: And that’s by virtue of the fact that they’re all using the same program?

Harrington: That’s correct. It’s a common third party. And I think it’s important to keep in mind here that while all that sounds rather suspicious and nefarious, the fact is there’s an efficiency here. A company always has to decide when it’s going to produce some input internally and when it’s going to use an outside supplier to provide that input. What’s happened in the time of big data and algorithms is the input that can now be outsourced is the pricing algorithm, because that third party, that data analytics company, may be able to use data better than the company itself in terms of finding out the best prices.

Jamali: Is a case like this harder to prove than the Sotheby’s example?

Harrington: Well, it really comes down to whether there’s evidence of communication. And in the case of Sotheby’s and Christie’s, they had firsthand evidence which established that there was communication where they agreed to set higher and common commission rates and other dimensions of competition. So, if you have evidence like that, it doesn’t make any difference whether you’re coordinating on prices or pricing algorithms, that’s clearly a violation of the law.

The real issue here in the New Jersey case is going to be whether there’s evidence that they really coordinated their decisions to adopt Rainmaker’s pricing algorithm. Now I can say related to that, there’s a similar case that’s been taking place in Las Vegas involving the same set of players, and recently a judge ruled in that case and dismissed it with prejudice. The judge found no evidence provided by the plaintiffs that even suggested that there was an agreement among the hotels and Rainmaker.

But I think what’s really important to keep in mind here is two different issues. One is what’s a violation of the law, and that’s all about there being an agreement. But the other one is about, well, are consumers harmed? And consumers can be harmed even when there’s not a violation of the law. And I think that possibility is a real concern here, in that the third party, the data analytics company, in the absence of any sort of agreement with the subscribing firms, could choose to have the algorithm recommend inflated prices. If it gets all its subscribers to charge higher prices, they’ll do better, they’ll earn higher profits, and the third party can get some of that back to the fees they charge.

Jamali: I wanted to visit something that FTC Chair Lina Khan recently told my colleague Rosie Hughes — she’s one of our producers here at Marketplace. Lina Khan talked to her about her approach to algorithmic price fixing, and here’s what she said:

Lina Khan: We’ve been in partnership with the DOJ, filing briefs with the court, explaining to the court that price fixing is still illegal, even if you are achieving it through an algorithm, that there’s no kind of algorithmic exemption to the antitrust laws.

Jamali: So, she’s pretty clear there: This is illegal, whether the collusion is happening in the room with the person in, you know, some fancy hotel room or whatever, or if you’re using a computer program, an algorithm to do this. What’s your response to that?

Harrington: Well, you have to distinguish between the case we’ve been describing, in which it’s an open question whether there’s an agreement, and another example, which clearly falls into what Ms. Khan is saying. So, there was a case involving sellers of wall posters on Amazon Marketplace. These two sellers communicated. They didn’t get in the same room, but they communicated through email and other means, and they coordinated and agreed to the pricing algorithms that they would set, and those pricing algorithms resulted in them not competing against one another. They were still competing with other sellers in the market, but not with each other. That is clearly a violation of existing antitrust law. They’re just not coordinating prices, instead, they’re coordinating the pricing algorithms that take data to then determine prices. But there, what you have is two or more individuals reaching an agreement. With the case of the third parties [as in the New Jersey hotels case], where there’s a data analytics company who might be resulting in supracompetitive prices, the question is: Is there actually an agreement? And it’s quite possible that there’s not an agreement and that this is entirely done at the behest of the third party. And in that case, that gets a little murky, because that’s what’s called unilateral conduct, and that does not fall under Section 1 of the Sherman Act, which makes price fixing agreements illegal.

Jamali: Do you get the sense that these companies being accused of collusion by way of an algorithm can hide behind the algorithm and almost use it as an excuse? There’s that hear no evil, see no evil aspect to this.

Harrington: Well, it certainly has happened in the past, where companies are hiding not necessarily from some kind of legal judgment, but really from consumer retribution. There have been episodes in the airline industry where they’ve been using algorithmic pricing for decades, where all of a sudden, the airfares just skyrocket. So, there was a well-known incident a number of years ago where Amtrak on the Northeast Corridor was shut down, so all of a sudden, there was a spike. I mean, in just a matter of hours, a huge spike in demand for air travel along the Northeast Corridor, and ticket prices just skyrocketed. And you could say that, well, these are managers that are just taking advantage of this increase in demand. But the fact is, it was already programmed in the pricing algorithm that if all of a sudden there’s a scarcity of available seats, that prices would go up. And so, they could say — and I think it was appropriate — that we weren’t intending to set these really high prices, but that’s just what was built into the algorithm. Now that’s just from the perspective of trying to rationalize your conduct with consumers. In terms of protecting yourself from the law, that’s going to be determined.

Stories You Might Like

Are Amazon prices fair or not?

Will a Biden administration try to break up Big Tech?

From prison to pandemic

"We learned the government is often the underdog in a lot of these cases"

White House signals that antitrust enforcement is on its agenda

Will Biden address antitrust concerns with an executive order?