Add a bookmark to get started

27 de fevereiro de 202414 minute read

The Preventing Algorithmic Collusion Act: A swing and a miss?

Senator Amy Klobuchar (D-MN), Chair of the Senate Subcommittee on Competition Policy, Antitrust, and Consumer Rights, has introduced legislation[1] designed to address the prevalence of pricing algorithms.[2] The intent of the bill, introduced on January 30, 2024, is to bar the use of algorithms that may be used to engage in price fixing. A price-fixing agreement would be presumed “when direct competitors share competitively sensitive information through a pricing algorithm to raise prices.”

In this alert, we discuss these key considerations related to the proposed legislation, including the bill’s usefulness, possible consequences, and next steps for companies.

Background

In January 2024, Senator Amy Klobuchar and six colleagues introduced the Preventing the Algorithmic Facilitation of Rental Housing Cartels Act.[3] The goal of the legislation is, among other things, to “[m]ake it unlawful for rental property owners to contract for the services of a company that coordinates rental housing prices and supply information, and designate such arrangements a per se violation of the Sherman Act.”[4]  Days later, many of the same co-sponsors introduced the Preventing Algorithmic Collusion Act of 2024,[5] which applies the same presumption of illegality to pricing algorithms outside of the housing market and, per the sponsors’ press release, “close[s] a loophole in current law by presuming a price-fixing ‘agreement’ when direct competitors share competitively sensitive information through a pricing algorithm to raise prices.”[6]

The current focus on algorithmic pricing traces its genesis to a 2022 ProPublica report on a company called RealPage, whose software program, YieldStar, amalgamated private rent data to “help landlords push the highest possible rents on tenants.”[7] Three days after the story was published, tenants filed one of several class action suits against RealPage alleging price fixing,[8] with more than 30 other filings following suit, leading to a consolidated case to be heard in Tennessee.[9]

Following these legal moves, on October 24, 2023, US Senators Amy Klobuchar and Mike Lee (R-UT), Chair and Ranking Member of the Senate Judiciary Subcommittee on Competition Policy, Antitrust, and Consumer Rights, held a hearing titled, “Examining Competition and Consumer Rights in Housing Markets.”[10] In the hearing, Senator Klobuchar identified four major factors in rising residential rents, focusing on “the widespread use of algorithmic pricing tools designed to raise prices even at the expense of higher vacancy rates.”[11]

Since 2015, academic literature has been addressing the possibility that algorithms could collude.[12] Collusion in such publications, however, has been defined by data scientists and economists and not by lawyers or judges.[13] With the recent focus on artificial intelligence (AI) and pricing algorithms, most of the publications from lawyers and consultants have called attention to the issue, but have not tried to sync technological capabilities with how these activities would be viewed through the law of evidence.[14]

Is the bill necessary?

Consumers’ concerns about AI may have prompted the legislation introduced by Senator Klobuchar.[15] However, the legislation suffers from many infirmities, including lack of support from antitrust and evidentiary case law. While unlikely, if the bill passes during a national election year it is reasonable to expect definitional and other challenging litigation, somewhat reminiscent of the litigation that followed the passage of the civil Racketeer Influenced and Corrupt Organizations Act (RICO).

Is the bill well written?

Perhaps this legislation is deliberately vague – it would not be out of character for Congress to attempt to pass a bill like this and, regardless of whether it passes, leave it to the regulatory agencies to implement those details through rulemaking. The Senate Judiciary Committee has held multiple hearings on antitrust legislation without much in the way of new laws, giving the Federal Trade Commission (FTC) an opening to enact law administratively.[16]

What are the possible negative consequences if the bill passes?

The potential flaws in the legislation fall into three categories: lack of definition, failure to recognize possible procompetitive effects by algorithms that agree (avoiding the term collusion here), and the failure to account for existing and future technologies.

Notably, the bill does not attempt to define collusion.[17] We referred to civil RICO. There have been many cases in search of the meaning of “pattern of racketeering activity.” The failure to be specific, provide definitions, and to create a workable bill causes the kind and volume of litigation that can defeat the very objective of the bill – to act quickly to address a perceived problem.

Another potentially significant flaw with the legislation is the collapse of the traditional distinction between bare information sharing and collusion. As drafted, the bill presumes an agreement in restraint of trade if an algorithm trained with nonpublic competitor data is distributed and used even if only to recommend pricing. This raises the question of whether a defendant should be presumed to have violated the antitrust laws if it merely took recommendations, but then made an entirely unilateral pricing decision? Ordinarily, under the Sherman Act,[18] an exchange of information in the absence of an agreement would not necessarily be unlawful, but would be governed by the rule of reason, leaving the burden on the plaintiff or government to show anticompetitive effects.[19] It is unclear why the information exchange should be any different because it is done algorithmically.[20] We believe that this reduction in proof does not withstand scrutiny under current antitrust law.

What should companies do if they are concerned about the passage of the bill?

While the bill contains some flaws, and passage appears unlikely, companies are encouraged to act as if the bill will pass. This is because it seems to track with the Department of Justice’s (DOJ) stated position that current antitrust law “prohibits competitors from fixing prices by knowingly sharing their competitive information with, and then relying on pricing decisions from, a common human pricing agent who competitors know analyzes information from multiple competitors. The same prohibition applies where, as here, the common pricing agent is a common software algorithm.”

Notably, the DOJ recently sued Agri Stats, a data reporting company in the meat industry, for collecting, analyzing, and distributing industry financial data.[21] It is possible DOJ or FTC might bring such an action against any company (agricultural or otherwise) that provides pricing guidance to competitors based upon the application of algorithmic programs to the collection of companies’ competitively sensitive data, even in the absence of an explicit effort to collude.

Will the bill stand the test of time?

While pricing algorithms have been around for a while, new add-on technologies that use greater computing power, such as machine learning and deep learning, are starting to make their way into mainstream pricing strategies.[22] Such technological advances have prompted the DOJ Antitrust Division to implement “Project Gretzky” and hire data scientists and AI experts to understand the new technological advances in order to ensure that DOJ is able to enforce existing antitrust laws.

Moreover, these technological advances are producing procompetitive efficiencies, in addition to some concerning effects. This bill raises the question of whether there is a need for this particular legislation, and whether any new antitrust legislation is warranted, as some believe that the current antitrust laws and associated case law are robust enough to handle the developing technology. In any event, in five or ten years, when the technology has advanced even more, it seems plausible that this bill would be outdated.

What are the prospects for passage?

We suggest that there are multiple problems with the proposed legislation. Most saliently, it conflicts with longstanding antitrust doctrine regarding information exchanges, implicitly presupposes only negative effects while ignoring procompetitive efficiencies, and its approach seems unlikely to keep pace with advancing technology. However, the prospects for its passage are dim. However well-intentioned the bill may have been as an effort to address consumers’ concerns in a time of rising prices, this legislation is no home run – it is a swing and a miss.

For more information, please contact any of the authors.



[1] The Preventing Algorithmic Collusion Act, S. 3686, 118th Cong. (2024), https://www.congress.gov/bill/118th-congress/senate-bill/3686.
[2] A pricing algorithm is a computerized sequence of rules that “uses price as an input, and/or uses a computational procedure to determine price as an output.” Pricing Algorithms, Economic Working Paper on the Use of Algorithms to Facilitate Collusion and Personalised Pricing 9 (U.K. Competition & Mkts. Auth., Working Paper No. CMA94, 2018), https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/746353/Algorithms_econ_report.pdf.
[3] Preventing the Algorithmic Facilitation of Rental Housing Cartels Act, S. 3692, 118th Cong. (2024), https://www.congress.gov/bill/118th-congress/senate-bill/3692.
[4] Press Release, Sen. Ron Wyden (D-OR), Wyden and Welch Introduce Legislation to Crack Down on Companies that Inflate Rents with Price-Fixing Algorithms (Jan. 30, 2024), https://www.wyden.senate.gov/news/press-releases/wyden-and-welch-introduce-legislation-to-crack-down-on-companies-that-inflate-rents-with-price-fixing-algorithms.
[5] Preventing Algorithmic Collusion Act of 2024, S. 3686, 118th Cong. (2024), https://www.congress.gov/bill/118th-congress/senate-bill/3686.
[6] Press Release, Sen. Amy Klobuchar (D-MN), Colleagues Introduce Antitrust Legislation to Prevent Algorithmic Price Fixing (Feb. 2, 2024), https://www.klobuchar.senate.gov/public/index.cfm/2024/2/klobuchar-colleagues-introduce-antitrust-legislation-to-prevent-algorithmic-price-fixing.
[7] Heather Vogell, Haru Coyne & Ryan Little, Rent Going Up? One Company’s Algorithm Could Be Why, PROPUBLICA (Oct. 15, 2022, 5:00 AM), https://www.propublica.org/article/yieldstar-rent-increase-realpage-rent.
[8] Bason, et al. v. RealPage, Inc., et al., No. 3:22-cv-01611-WQH-MDD (S.D. Cal. Oct. 18, 2022).
[9] Leslie Shaver, RealPage Lawsuit Progresses in Court, MULTIFAMILY DIVE (Jan. 12, 2024), https://www.multifamilydive.com/news/antitrust-suit-RealPage-rent-pricing-software-algorithm/704352/; See In re RealPage, Rental Software Antitrust Litig., 669 F. Supp. 3d 1372 (U.S. Jud. Pan. Mult. Lit. 2023).
[10] Examining Competition and Consumer Rights in Housing Markets: Hearing Before the Subcomm. on Competition Pol’y, Antitrust, & Consumer Rts., 118th Cong. (2023), https://www.judiciary.senate.gov/committee-activity/hearings/examining-competition-and-consumer-rights-in-housing-markets [hereinafter Competition Pol’y Subcomm. Hearing].
[11] Competition Pol’y Subcomm. Hearing, supra note 6.
[12] See, e.g., Ariel Ezrachi & Maurice Stucke, VIRTUAL COMPETITION, THE PROMISE AND PERILS OF THE ALGORITHM-DRIVEN ECONOMY, (Harvard Univ. Press 2016); Zach Brown & Alexander MacKay, Competition in Pricing Algorithms (Nat’l Bureau of Econ. Rsch., Working Paper No. 28860, 2021); Ai Deng, WHAT DO WE KNOW ABOUT ALGORITHMIC COLLUSION NOW? NEW INSIGHTS FROM THE LATEST ACADEMIC RESEARCH, (January 28, 2024); Akhil Sheth, CONSPIRING COMPUTERS: ALGORITHMIC PRICE-SETTING, ARTIFICIAL INTELLIGENCE, AND THE ANTITRUST LAWS, at 3 (Association of Business Trial Lawyers, Vol. XXI, No. 1) (2019) (Current DLA Piper attorney Akhil Sheth explores the interplay between antitrust and artificial intelligence).
[13] For more on the differences between the economic and legal view of collusion, see generally Louis Kaplow, An Economic Approach to Price Fixing, 77 ANTITRUST L. J. 343, 344 n.5, 373 (2011); Louis Kaplow & Carl Shapiro, Antitrust 23–24, 44 (Nat’l Bureau of Econ. Rsch., Working Paper No. 12867, 2007), https://www.nber.org/papers/w12867.
[14] Although there are only a few reported decisions addressing pricing algorithms, none have stated that current antitrust or evidentiary law is insufficient to address the various legal issues associated with the use or misuse of pricing algorithms. In fact, in a recent action, the Department of Justice (DOJ) filed a Statement of Interest declaring that the current framework of antitrust laws is sufficient to address alleged price fixing by algorithmic collusion. See Memorandum of Law In Support of the Statement of Interest of the United States, In re: RealPage, Rental Software Antitrust Litigation (No. II), Case No. 3:23-MD-3071, November 15, 2023, (available at https://fingfx.thomsonreuters.com/gfx/legaldocs/byvrrwqozve/DOJ%20RealPage%202023-11-15%20Memorandum%20dckt%20628_0.pdf).
[15] Fictional dystopian scenarios have been invoked. In contrast, more serious literature is proliferating. See Salil Mehra, Price Discrimination-Driven Algorithmic Collusion: Platforms for Durable Cartels, 26 STAN. J. L. BUS. & FIN. 171 (2021); Ariel Ezrachi & Maurice Stucke, Sustainable and Unchallenged Algorithmic Tacit Collusion, 17 NW. J. TECH. & INTELL. PROP. 217 (2020). Another view is represented by Thibault Schrepel, The Fundamental Unimportance of Algorithmic Collusion for Antitrust Law, HARV. J. LAW & TECH. (February 7, 2020).
[16] We await the outcome of litigation challenging Chevron deference. See, e.g., Corner Post v. Board of Governors of the Federal Reserve System, S. Ct., Case No. 22-1008 (argued February 20, 2024); See generally Fed. Trade Comm’n, Rulemaking, https://www.ftc.gov/enforcement/rulemaking (providing a current catalog of existing, rescinded, and proposed FTC rules).
[17] When asked “can pricing algorithms collude?”, Chat GPT-4 (Copilot) offers contradictory answers: (1) “Certainly! The use of pricing algorithms has become increasingly prevalent, especially in the realm of e-commerce. These algorithms dynamically adjust prices based on various factors such as market conditions, competitor prices, and consumer behavior, all with the goal of maximizing profits;” versus, (2) “In summary, while pricing algorithms haven’t fully colluded yet, ongoing research and vigilance are essential to ensure fair pricing practices in the digital age.” Chat GPT sources various papers iin support of each position. Other AI models provide different answers when responding to the same query. For example, Perplexity AI responds that “[p]ricing algorithms have the potential to tacitly collude, leading to supracompetitive prices. Research has shown that algorithms can learn to charge high prices without communicating with each other.”
[18] Most scholars agree that the Sherman Act should control but other ideas are advancing. See Aneesa Mazumdar, Algorithmic Collusion: Reviving Section 5 of the FTC Act, 122 COLUM. L. REV. 449 (2022) (arguing that Section 5 of the FTC Act is a better framework compared to the Sherman Act.). This would fit progressive thinking, especially as the FTC has redefined its interpretation of unfair competition under Section 5 to more heavily favor enforcement. Fed. Trade Comm’n, Policy Statement Regarding the Scope of Unfair Methods of Competition Under Section 5 of the FTC Act (July 9, 2021), https://www.ftc.gov/system/files/ftc_gov/pdf/P221202Section5PolicyStatement.pdf.
[19] In a recent podcast, David Brown, Daniel Cajueiro, Andrew Eckert, and Douglas Silveira discussed their academic research on collusion and price fixing through their study of the Alberta restructured wholesale electric market, which prices and reprices repeatedly. The authors agree that pricing algorithms can increase competition but can also be used for anticompetitive effects such as price fixing. Stanford Computational Antitrust Podcast, Episode 21: Using Machine Learning to Detect Tacit Collusion (January 29, 2024), https://podcasters.spotify.com/pod/show/stanford-computational-antitrust/episodes/Episode-21-Using-Machine-Learning-to-Detect-Tacit-Collusion-e2bf45s (citing David Brown, Daniel Cajueiro, Andrew Eckert & Douglas Silveira, Information and Transparency: Using Machine Learning to Detect Communication Between Firms (Stanford Computational Antitrust, Vol. III) (2023)); see also OECD, Algorithmic Competition, OECD Competition Policy Roundtable Background
Note, at 10–11 (2023), https://www.oecd.org/daf/competition/algorithmic-competition-2023.pdf (noting the pro-competitive effects of pricing and other algorithms).
[20] Note the typo in Section 5 on p. 9, lines 21–22, which confusingly states both “or” and “and.”
[21] Press Release, Dep’t of Just., Off. Of Pub. Aff., Justice Department Sues Agri Stats for Operating Extensive Information Exchanges Among Meat Processors (Sept. 28, 2023), https://www.justice.gov/opa/pr/justice-department-sues-agri-stats-operating-extensive-information-exchanges-among-meat.
[22] “Machine learning is a specific sub-field of AI that tries to create intelligent machines using algorithms that iteratively learn from data and experience. Deep learning is a sub-field of machine learning (and thus a sub-field of AI) that broadly replicates neurons in the human brain through an artificial neural network.” See OECD, supra note 19, at 9 (internal citations omitted).

Print