While algorithmic pricing tools can enhance market efficiency and competition, concerns persist about their potential to facilitate collusion. Antitrust enforcers in the United States are increasingly focusing on the use of algorithmic pricing mechanisms by competitors. This article examines the aggressive stance of the Department of Justice Antitrust Division (DOJ) and the Federal Trade Commission (FTC) on algorithmic price-fixing, which seek to treat such practices as per se illegal under the Sherman Act. Through an analysis of early key civil cases like RealPage and Cendyn, the article explores the government's arguments for considering algorithmic pricing as concerted action, and argues that courts should continue applying the Rule of Reason to assess the nuanced impacts of these technologies.
I. Introduction
Artificial intelligence (AI) is poised to profoundly transform the global economy. While AI may spur a technological revolution that has the potential to supercharge productivity,1 AI can also present an array of risks.2 Government enforcers around the world have begun to grapple with the challenges presented by AI, and antitrust and competition enforcers are no exception. In the United States, dual competition regulators, the Department of Justice Antitrust Division (DOJ) and the Federal Trade Commission (FTC), are taking particular aim at competitors' use of algorithmic pricing mechanisms. Claiming that "[p]rice fixing using AI is still price fixing," the agencies have promised that prosecutors will seek even harsher penalties where AI technology is used to advance collusive crimes.3 Even groups of private plaintiffs are filing civil lawsuits, alleging that competitors who act on the suggestions of algorithmic pricing programs are no more than AI-assisted colluders.
The DOJ and FTC have responded to these lawsuits by filing aggressive Statements of Interest, asserting that mere use of pricing algorithms by competitors should be treated as illegal per se. But courts seem justifiably skeptical of this approach. After all, using algorithmic pricing mechanisms and agreeing with competitors to price according to those mechanisms are distinct factual scenarios, and algorithmic pricing tools can have a host of procompetitive benefits. While it should not be legal for competitors to "task" AI with otherwise collusive activities, the use of AI should not be allowed to per se supplant the legal prerequisite of a conspiratorial agreement.
II. Innovative Algorithmic Pricing Models
The term "algorithmic pricing practices" generally refers to the use of predefined, rule-based algorithms that can analyze market data (such as cost, competitor prices, and demand) and suggest or even automate a company's pricing decisions. Algorithms can use different parameters to achieve different ends, and can analyze and adjust pricing recommendations based on real-time competitive conditions.4 While some algorithmic models are capable of utilizing only a company's own or public data sources to generate strategies, many models rely on the sharing of competitors' information through a third-party algorithm. Typically, these third-party programs will not provide their users with raw competitive data, but only with recommendations based on their completed analyses.
AI and algorithmic pricing practices are proliferating and evolving quickly, often for procompetitive reasons. These tools can aid in price discovery, which allows companies to better engage in price competition. Computers are capable of ingesting and analyzing vastly larger quantities of data than a person attempting to conduct similar analysis. Algorithms are also capable of quickly adjusting recommended prices in response to market fluctuations, which can ensure that pricing is more competitive. And algorithms are also potentially able to reduce human error or bias in the analysis of market conditions, leading to more accurate pricing.5
Yet, competition enforcers are concerned that algorithmic pricing mechanisms could be used to facilitate explicit or tacit price-fixing. Some believe that algorithm use will lead to higher, more uniform market prices as computers signal and engage in parallel pricing moves. And as advancement in generative AI technology merges with algorithmic models, the fear is that collusion will inevitably occur autonomously – that AI-enabled algorithms will independently "conspire" without human instruction or intervention.6
III. Traditional Legal Framework in the U.S
Pure information exchanges between competitors, even those involving competitively sensitive pricing data, have traditionally been analyzed under the Rule of Reason standard.7
This is due to the courts' long-standing recognition8 that the sharing of "information about prices, costs, capacity and availability can benefit [market participants], by allowing markets to function more efficiently, intelligently and competitively[.]"9
But courts also recognize that information exchanges (particularly those involving competitively sensitive information) can be used to effectuate the kinds of agreements between competitors traditionally treated as illegal per se (including price fixing agreements).10 Where no direct evidence of an agreement exists, plaintiffs can nevertheless invite an inference of conspiracy by proving a pattern of parallel competitive conduct supported by various "plus factors." Under certain conditions, courts have considered exchanges of price information among competitors to be a plus factor supporting the inference of an anticompetitive agreement. In determining whether such an inference is appropriate, courts have considered, for example, the role and responsibility of the persons engaging in the exchanges,11 the flow of information to decision-makers,12 the temporal proximity of the information exchanges to pricing decisions,13 and proof that the exchanges had an impact on pricing.14
IV. Private Civil Cases
In the United States, the most prominent battle over algorithmic pricing software programs has taken place in real estate markets, starting with In re RealPage, Inc., Rental Software Antitrust Litig. (No. II)15 in 2023. There, renters of multifamily and student housing filed dozens of class action lawsuits against RealPage Inc., a company that developed and sold a suite of revenue management software ("RMS") to property owners, operators, and managers. RealPage's clients submitted their rental pricing and supply data to fuel RealPage's price optimization RMS algorithms, and RealPage produced unit-specific, daily price recommendations for their clients. RealPage promised its clients that implementing its recommendations would "outperform the market" by achieving "both short-and long-term goals of increasing revenues by raising rents."16 However, plaintiffs did not merely allege that competing market players used RealPage's software – plaintiffs alleged that RealPage actively sought compliance with its price recommendations such that clients adopted over 80% of its suggested prices. Thus, the suit alleged that RealPage and its clients had formed an illegal "hub-and-spoke" price-fixing conspiracy based not only on each users' contribution of proprietary data, but the belief of each user that its competitors would also price their properties according to the algorithm's recommendations.
Similarly, in Duffy v. Yardi Systems, Inc., 17 plaintiff renters alleged that property owners and managers used Yardi's pricing algorithm, "RENTmaximizer," to inflate multifamily unit rental prices. Plaintiffs also alleged, as in RealPage, that Yardi encouraged its customers to automatically adopt its calculated price recommendations. Central to the allegations in both RealPage and Yardi is the argument that by ceding independent pricing and supply decisions to an algorithm, defendants created a "hub-and-spoke" cartel that did not require a "rim" – horizontal competitors did not need to agree with one another in order to effectuate price hikes.
Numerous private suits in the hotel and casino space shortly followed the filing of RealPage and Yardi. 18 Suits are also now being filed in the healthcare space, where medical provider plaintiffs have sued a repricing algorithm provider and several major insurance companies for allegedly using thirdparty repricing tools to reduce reimbursement rates paid to out-of-network providers.19
V. The DOJ and FTC Weigh In
Citing the "tremendous practical importance"20 of the development of the law on algorithmic collusion, the DOJ (later joined by the FTC) has filed Statements of Interest in support of plaintiffs in three key private cases. In all three, the government has taken the aggressive position that competitors' common use of pricing algorithms should be treated as a per se illegal price fixing arrangement, in violation of Section 1 of the Sherman Act.
The DOJ's first shot was fired in RealPage, beginning with the common-sense proposition that AI may not legally do what would otherwise be illegal,21 and that the Sherman Act should be malleable enough to reach technologically evolving methods of collusion. But then the DOJ turned to the argument that competitors' use of algorithmic pricing mechanisms, without more, can meet the two key prongs of a Sherman Act Section 1 violation.
First, the DOJ set out an expansive definition of "concerted action." In the DOJ's view, any conduct that "eliminate[s] independent decisionmaking" can be legally cognizable concerted action, regardless of whether there was "any additional subsequent agreement or coordination among the parties."22 In fact, despite the DOJ's view that an agreement can be proven through circumstantial evidence, it argues that neither simultaneous action among conspirators, nor any action close in time, is required. Rather, all that must be shown is an invitation from an alleged co-conspirator and an action in concert with the invitation. If adopted by courts, the DOJ's position would arguably eliminate the need to establish any "rim" of an alleged hub-and-spoke conspiracy where algorithmic pricing mechanisms are used.
Second, the DOJ insisted that algorithmic price recommendation was an unreasonable restraint that should be condemned as horizontal per se unlawful price fixing. Even though the agreements between algorithm providers and their customers are vertical, the DOJ argued that the effect of the agreement was to eliminate independent pricing decisions by horizontal competitors. The statement nods to the proposition that information exchange is not per se unlawful, but goes on to argue that competitors "knowingly combined their sensitive, nonpublic pricing and supply information in an algorithm," and then relied on the algorithm's pricing recommendations when making decisions, "with the knowledge and expectation that other competitors will do the same." According to the government, this is sufficient for per se liability, even if algorithm users never communicated about prices directly.
Subsequently, in two additional Statements of Interest, the DOJ and FTC23 jointly pushed for a per se rule. In Yardi, the government claimed that even where users of a common pricing algorithm retain discretion to make independent pricing decisions based on the recommendations of the algorithm, the practice should be per se illegal where competing companies "jointly delegate key aspects of their pricing."24 Likening it to an agreement among competitors to fix list prices where final sales could still deviate, the agencies took the position that algorithmic recommendations are nevertheless agreements to fix the "starting point" of prices, which "corrupt[s] the decentralized price-setting mechanism in the market, whether or not [competitors] ultimately succeed in raising or stabilizing prices." In other words, the DOJ argued for a presumption of per se illegality based on the mere use of algorithmic pricing, even if competitors did not follow the algorithm's recommendations.
Later, in Cornish-Adebiyi v. Caesars Entertainment, 25 the DOJ and FTC argued that no direct information exchange between competitors, nor any direct communications whatsoever, are required to prove a "rim" agreement linking conspiring defendants. They posited that the very act of centralizing pricing information "with a common pricing agent" should be considered the legal equivalent of sharing information directly with a competitor, because "the alleged scheme is designed to obviate the need for competitors to share information directly with each other." This argument, if adopted by courts, would eliminate the need for a plusfactor analysis where a provider's invitation to delegate pricing decisions to its algorithm "inherently contemplates concerted action."
VI. The Real Page Court's Response
While the DOJ and FTC's positions on algorithmic pricing are aggressive, the real question is how the courts will receive those arguments. So far, the courts evaluating these issues have either directly rejected the agencies' key arguments or otherwise were reluctant to embrace the agencies' view of the law.
In RealPage, the district court rejected the application of the per se standard, but allowed the case to continue under the Rule of Reason analysis.26 The court not only appreciated that the per se standard should be applied only to "clearcut" horizontal conduct with "no plausibly procompetitive features," it viewed the vertical relationship between the algorithm provider and its customers as something other than "a straightforward conspiracy."27
The court also found it important that defendants did not follow RealPage's pricing recommendations between 10- 20% of the time, noting that without "absolute delegation" of pricing decisions to RealPage, the agreement could not clear the per se bar.28 The court noted that plaintiffs did not allege direct communications between RealPage's clients to join or maintain a conspiracy, nor was there any alleged mechanism to require acceptance of price recommendations or punish uncooperative clients. According to the court, while "some level of horizontal conspiracy"29 may have existed among RealPage clients to contribute data, that alone was not enough to merit per se treatment. Finally, the court found that plaintiffs did not sufficiently allege parallel conduct, given that each defendant began using RealPage's services at different times, often years apart from one another.
It is also interesting to note that a recent decision in a case involving the aggregation and dissemination of publicly available data also belies certain of the agencies' arguments. In Gibson v. Cendyn Grp., LLC, 30 the court dismissed with prejudice a civil class action alleging that Las Vegas hotel operators and an algorithmic pricing software engaged in a hub-and-spoke conspiracy to fix hotel room prices.
Unlike in RealPage, the defendants in Cendyn were not alleged to have pooled confidential or proprietary information in Cendyn's system; Cendyn used publicly available pricing to fuel its database. Without an "exchange of confidential information from one of the spokes to the other through the hub's algorithms," the court found, "there is no rim," even though the algorithm used data about Cendyn's customers for its customers' benefit.31 Thus, the "mere use of algorithmic pricing based on artificial intelligence by a commercial entity, without any allegations about agreement between competitors – whether explicit or implicit – to accept the prices that the algorithm recommends does not plausibly allege an illegal agreement[.]"32 The court also found that plaintiffs had not sufficiently alleged collusion because Cendyn's customers did not begin using its services at, or even near, the same time; initial use by defendants was staggered over a 10-year period. The court also held that complaint failed to adequately allege that the hotel defendants delegated pricing authority to the algorithm despite allegations that pricing recommendations were accepted over 90% of the time.
VII. The Rule of Reason Should Remain the Standard for Algorithmic Pricing Tools
The fight to define the applicable legal standard for algorithmic pricing cases is still in its early stages. But despite efforts by both public and private enforcers, courts thus far seem more comfortable with the default framework of the rule of reason. Each of the recent cases is alleged to involve individual, vertical agreements by competitors to use third-party algorithm providers; no case alleges the pure horizontal relationship that historically has been the hallmark of per se application.33 And while enforcers wish to classify such relationships as hub-and-spoke conspiracies, all lack the "rim" of direct competitor communications. None of these algorithmic pricing practices has been subject to sufficient testing by the judicial system to merit being declared "so obviously anticompetitive that it has no plausibly procompetitive features."34
Where users of algorithmic pricing mechanisms retain and exercise individual pricing discretion, per se liability simply should not apply. Even a company's following an algorithmic pricing recommendation cannot mean harm to a market occurred per se, because information on the actual balance of demand and supply can aid individual efforts to compete. The uniform imposition of the per se rule at this early stage of case development could also threaten future innovation. The agencies' attempts to equate merely receiving and/or taking algorithmic pricing suggestions with per se "price fixing" would render the use of nearly any rule-based pricing algorithm illegal. The specific algorithms in the current civil cases only scratch the surface of the markets in which algorithms are now and can be used for procompetitive purposes. Courts should therefore continue to undertake more than a simple per se analysis before weighing on questions of liability.
Finally, courts should take special care with applying per se illegality because doing so may lead to premature efforts at criminalization. Even when the DOJ appears to have taken a long and deliberate path to criminal enforcement, court and jury acceptance has not been smooth. For example, the DOJ's early labor market enforcement efforts have been met with not only opposition from juries, but adverse court rulings that arguably muddy the nature of the per se standard.35 A record of civil adjudication, along with the full analysis required by the rule of reason standard, is best suited to help all market participants to determine the true nature and effect of emerging algorithmic analytics.
Footnotes
1 Sam Altman (CEO of OpenAI, developer of ChatGPT) stated: "I think [AI] will be the most powerful technology humanity has yet invented ... It's the world that sci-fi has promised us for a long time – and for the first time, I think we could start to see what that's gonna look like." Simmone Shah, Sam Altman on OpenAI, Future Risks and Rewards, and Artificial General Intelligence, Time (Dec. 12, 2023), https://time.com/6344160/a-year-in-time-ceo-interview-sam-altman/.
2 Aaron Mok, Sam Altman doesn't think we are worried enough about how AI will impact the economy, Business Insider (May 8, 2024) ("The thing I'm most worried about right now is, the sort of, the speed and magnitude of the socioeconomic change may have, and what the impacts on what that will be."), www.businessinsider.com/sam-altman-says-ais-economic-impact-top-concern-2024-5.
3 Deputy Attorney General Lisa Monaco, Keynote Remarks at the American Bar Association's 39th National Institute on White Collar Crime (Mar. 7, 2024), available at www.justice.gov/opa/speech/deputy-attor ney-general-lisa-monaco-delivers-keynote-remarks-american-bar-asso ciations.
4 See Renato Nazzini/James Henderson, Overcoming the Current Knowledge Gap of Algorithmic "Collusion" and the Role of Computational Antitrust, 4 Stan. Computational Antitrust 1 (2023), available at https://law.stanford.edu/wp-content/uploads/2024/02/Algorithmic-Col lusion.pdf.
5 Sophie Calder-Wang/Gi Heung Kim, Coordinated vs Efficient Prices: The Impact of Algorithmic Pricing on Multifamily Rental Markets (July 24, 2023), available at https://ssrn.com/abstract=4403058.
6 Sara Fish, Algorithmic Collusion by Large Language Models, available at https://arxiv.org/pdf/2404.00806.
7 United States v. Citizens & So. Nat'l Bank, 422 U.S. 86, 113 (Jun. 17, 1975) ("But the dissemination of price information is not itself a per se violation of the Sherman Act.").
8 See e.g., United States v. U.S. Gypsum Co., 438 U.S. 422, 443 n.16 (Jun. 29, 1978) ("The exchange of price data and other information among competitors does not invariably have anticompetitive effects; indeed such practices can in certain circumstances increase economic efficiency and render markets more, rather than less, competitive.")
9 Remarks of J. Thomas Rosch, Antitrust Issues Related to Benchmarking and Other Information Exchanges, at 9 (May 3, 2011), www.ftc. gov/sites/default/files/documents/public_statements/antitrust-issues-rela ted-benchmarking-and-other-information-exchanges/110503rosch benchmarking.pdf.
10 The DOJ has relied on evidence of competitor information exchanges to support criminal price-fixing charges under Section 1. See United States v. Swanson, No. 4:06-cr-00692 (N. D. Cal. Oct. 18, 2006) (charging individual for conspiring to fix prices of DRAM chips based on role in obtaining and disseminating competitor price information).
11 See In re Petroleum Prods. Antitrust Litigation, 906 F.2d 432, 445-50 (9th Cir. Dec. 13, 1991); In re Flat Glass Antitrust Litig., 385 F.3d 350, 368-69 (3d Cir. Sep. 29, 2004) (allowing price-fixing complaint alleging that high-level executives systematically exchanged pricing information that influenced pricing decisions).
12 In re Static Random Access Memory (SRAM) Antitrust Litig., 2010 WL 5138859, at *6 (N.D. Cal. Dec. 10, 2010).
13 In re Blood Reagents Antitrust Litig., 266 F. Supp. 3 d 750, 778 (E. D. Pa. Jul. 19, 2017) (holding that "the close temporal link between the transfer [of price information] and announcement of the 2001 price increases raise an inference of conspiracy"); In re Platinum & Palladium Antitrust Litig., 2017 WL 1169626, at *13 (S. D. N. Y. Mar. 28, 2017) (information exchanges shortly followed by price drops supported inference of conspiracy).
14 Stanislaus Food Prod. Co. v. USS-POSCO Indus., 2013 WL 595122, at *12 (E. D. Cal. Feb. 15, 2013), aff'd, 803 F.3d 1084 (9th Cir. 2015) (quoting In re Baby Food Antitrust Litig., 166 F.3d 112, 125 (3d Cir. Jan. 12, 1999) ("there must [also] be evidence that the exchanges of information [actually] had an impact on pricing.")).
15 In re RealPage, Inc., Rental Software Antitrust Litig. (No. II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023).
16 In re RealPage, Inc., Rental Software Antitrust Litig. (II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023) at *2. Similar state law claims have been brought against RealPage and its customers by the District of Columbia and Arizona Attorneys General.
17 Duffy v. Yardi Systems, Inc., No. 2:23-cv-01391 (W.D. Wash. May 6, 2024).
18 Gibson v. Cendyn Group, LLC, No. 2:23-cv-00140 (D. Nev. Jan. 25, 2023), ECF No. 1 (alleging hub-and-spoke conspiracy among hotel casinos on the Las Vegas Strip); Cornish-Adebiyi v. Caesars Entertainment ("Caesars"), No. 1:23-cv-02536 (D. N.J.Jan. 29, 2024), ECF No. 1 (alleging conspiracy among hotel casinos in Atlantic City); Portillo v. CoStar Group, Inc., No. 2:24-cv-00229 (W. D. Wash. Feb. 20, 2024), ECF No. 1 (alleging price fixing through algorithms in the luxury metropolitan hotel market); Dai v. SAS Institute, Inc., No. 4:24- cv-02537 (N. D. Cal. April 26, 2024), ECF No. 1 (alleging algorithmic price fixing of hotel guest rental rooms).
19 Live Well Chiropractic PLLC v. Multiplan, Inc., No. 1:24-cv-03680 (N. D.Ill. May 6, 2024), ECF No. 1.
20 Cornish-Adebiyi v. Caesars Entertainment, No. 1:23-cv-02536 (D. N.J. Mar. 28, 2024), ECF No. 96.
21 In re RealPage Rental Software Antitrust Litigation (II), No. 3:23-MD03071 (M.D. Tenn. Nov. 15, 2023), ECF No. 627 ("Put another way, whether firms effectuate a price-fixing scheme through a software algorithm or through human-to-human interaction should be of no legal significance. Automating an anticompetitive scheme does not make it less anticompetitive.").
22 In re RealPage Rental Software Antitrust Litigation (II), No. 3:23-MD03071 (M. D. Tenn Nov. 15, 2023), ECF No. 628, at 6.
23 Upon joining the DOJ's efforts, the FTC elaborated on its position in a concurrent blog post. See Hannah Garden-Monheit/Ken Merber, Price fixing by algorithm is still price fixing, Fed. Trade Comm'n, (Mar. 1, 2024), www.ftc.gov/business-guidance/blog/2024/03/price-fixing-algorithm-still-price-fixing.
24 Duffy v. Yardi, Case No. 2:23-cv-01391 (W.D Wash. Mar. 1, 2024), ECF No. 149, at 3.
25 Cornish-Adebiyi v. Caesars Entertainment, No. 1:23-cv-02536, (D. N.J. Mar. 28, 2024), ECF No. 96.
26 In re RealPage, Inc., Rental Software Antitrust Litig. (II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023) at *22-24.
27 In re RealPage, Inc., Rental Software Antitrust Litig. (II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023) at *23.
28 In re RealPage, Inc., Rental Software Antitrust Litig. (II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023), at *23.
29 In re RealPage, Inc., Rental Software Antitrust Litig. (II), No. 3:23- MD-03071 (M.D. Tenn. Dec. 28, 2023), at *23.
30 See generally Gibson v. Cendyn Grp., LLC, No. 2:23-CV-00140, 2024 WL 2060260 (D. Nev. May 8, 2024), at *1.
31 Gibson v. Cendyn Grp., LLC, No. 2:23-CV-00140, 2024 WL 2060260 (D. Nev. May 8, 2024), at *5.
32 Gibson v. Cendyn Grp., LLC, No. 2:23-CV-00140, 2024 WL 2060260 (D. Nev. May 8, 2024), at *6.
33 United States v. Brewbaker, 87 F.4th 563, 573-75 (4th Cir. Dec. 1, 2023)
34 Med. Ctr. at Elizabeth Place, LLC v. Atrium Health Sys., 922 F.3d 713, 718 (6th Cir. Apr. 25, 2019).
35 Daniel Oakes/Tiffany Rider, Losing per se: Potential fallout from the U.S. Department of Justice's no-poach enforcement, Concurrences N° 4 (2023), available at www.axinn.com/assets/htmldocuments/_03.concurrences_4-2023_on-topic_no-poach%20007.pdf.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
