DOJ Can’t Prosecute AI for Price Fixing Unless There’s Price Fixing

Among the Biden administration’s attempts to lower the cost of living, actions addressing housing costs have been front and center.

On August 23, the Department of Justice (DOJ), alleged in a lawsuit that RealPage, a company which uses AI to make pricing recommendations to landlords, is violating the Sherman Act. Eight state attorneys general, all Democrats except for Tennessee’s Jonathan Skrmetti, joined in the litigation. 

In announcing the suit, U.S. Attorney General Merrick B. Garland stated, “We allege that RealPage’s pricing algorithm enables landlords to share confidential, competitively sensitive information and align their rents. Using software as the sharing mechanism does not immunize this scheme from Sherman Act liability, and the Justice Department will continue to aggressively enforce the antitrust laws and protect the American people from those who violate them.”

This description of the allegedly illegal act is telling. Why use the word “align” rather than words more traditionally used to describe price fixing? The DOJ doesn’t point to “an agreement to fix prices” because the simple use of a software pricing tool does not itself constitute an agreement, and that is the fatal flaw of the DOJ’s case.

There’s little authoritative case law about AI. Most of it is copyright-related. There’s none related to the Sherman Act. So to analyze the merits of the DOJ’s allegation, we are left only with a long and well-developed understanding of how the courts have interpreted the act’s price fixing ban in less technologically advanced circumstances. 

Syracuse University law professor Shubha Ghosh writes that in the 1940 landmark price fixing case United States v. Socony-Vacuum Oil Co., “the court distinguished between the unlawful behavior, which it defined as any ‘agreement or plan embraced to raise and maintain prices’ among competitors, from the valid mechanisms or tools used to reach and support these price levels.” 

Landlords have always struggled to establish the highest price the market will bear. As Real Clear Markets’ John Tamny explains, accurate pricing is critical to address the shortage of housing because, “[a]bsent pricing clarity, there’s no way for apartment supply to ever match demand.” 

So pricing is critical. And although enabling landlords to charge the highest rent the market will bear doesn’t align with the administration’s policy goal of lowering housing costs, that does not necessarily mean RealPage is engaging in an agreement or plan between competitors to raise and maintain prices in violation of the Sherman Act.

Certainly, one could reason that the only reason landlords pay a fee to RealPage is that it helps them to charge the highest price the market will allow.

Professor Ghosh argues that, in situations like this, “the primary legal concern is whether the entities’ actions constitute a direct interference with the free play of market forces.” This would require the government to show an agreement  to fix rental prices between two or more competing landlords.  

In its explanation of the Sherman Act, the Justice Department states that “agreements among competitors to fix prices or wages, rig bids, or allocate customers, workers, or markets, are criminal violations. Other agreements such as exclusive contracts that reduce competition may also violate the Sherman Antitrust Act and are subject to civil enforcement.”

Neither RealPage nor its clients have an agreement to fix prices or any agreement that reduces competition. They simply have an agreement to share information through the service, which results in pricing suggestions that clients are free to ignore. The algorithm is simply a more efficient way for landlords to do what they’ve always done: gather as much information as possible to establish maximal pricing. Sharing pricing information alone, be it in a smoke-filled room or through an algorithm, is not a violation of the Sherman Act unless there is an accompanying agreement, tacit or explicit, to use that information to fix prices. The use of an algorithm does not relieve the DOJ of its burden of showing an agreement to fix prices.

As the Court wrote in Socony-Vacuum Oil Co., “An agreement to pay or charge rigid, uniform prices would be an illegal agreement under the Sherman Act. But so would agreements to raise or lower prices whatever machinery for price-fixing was used.” What is necessary for a Sherman Act violation, regardless of the method or machinery, is an agreement itself to fix prices. As the court wrote, “They are fixed because they are agreed upon.”

DOJ itself has used this exact formulation in previous proposed jury instructions:

Price fixing is an agreement between two or more persons to fix, control, raise, lower, maintain or stabilize the prices charged or to be charged for their products or services. Although a price-fixing conspiracy is commonly thought of as an agreement to establish the same price, prices may be fixed in other ways. For example, prices are fixed if the range or level of prices is agreed upon, or if, by agreement, various guidelines or formulas are to be used in computing the prices. They are fixed because they are agreed upon. Thus, any agreement to raise prices, to set a specific price, to maintain a specific price, to stabilize prices by eliminating discounts, or to set other terms or conditions of sale relating to price, is illegal.

The DOJ’s own jury instructions make clear that without an agreement related to pricing, there is no Sherman Act violation.

If landlords were contractually required by RealPage to set the price proposed by the software, or a price within a given percentage of the suggested price, the DOJ would have a more persuasive argument. But there is no agreement of any kind to fix prices in any way. So for the DOJ and the state AGs to prevail, they’d have to convince the court to hold—for the first time—that the mere sharing of information via the use of a third party algorithm to inform pricing decisions by landlords constitutes an “agreement or plan (by competitors) to raise prices,” or that landlords had a separate agreement to abide by the algorithm’s recommendations.

Why is there no agreement? Because a RealPage client is (contractually) free to test the market and rent a unit for more than the algorithm suggests. Suggests is the operative word. Conversely, a landlord, perhaps one who is experiencing a cash crunch, may choose to charge less than the algorithm suggests to rent the unit sooner than they could if they held out for the algorithm’s higher suggested price. Without an agreement to fix pricing, there can be no Sherman Act violation.

Senator Ron Wyden (D-OR) may have been thinking the same thing when, in January 2024, he introduced a bill ”[t]o prohibit the use of algorithmic systems to artificially inflate the price or reduce the supply of leased or rented residential dwelling units in the United States.” If the Sherman Act applied here, there’d be no need to introduce legislation that bans the practice that already violates long-existing and established law. 

No algorithm or other innovative AI tools, no matter how intelligent, could violate the Sherman Act without an agreement between market actors to abide by its recommendations. Short of such a meeting of the minds between landlords, the government’s allegation of price fixing under the Sherman Act would require a completely novel interpretation of the act.

Where the DOJ finds bad actors who have come to a meeting of the minds to set prices, with or without an algorithm, the Sherman Act is an appropriate tool to protect the American people from those who violate the law against collusion. But short of that specific evidence, or until Senator Wyden’s bill becomes law, the administration will have to find a better way to lower the cost of rental housing.