The Supreme Court’s landmark decision in Ohio v. American Express (Amex) remains central to the enforcement of antitrust laws involving digital markets. The decision established a framework to assess business conduct involving transactional, multisided platforms from both an economic and legal perspective. At its crux, the Court in Amex integrated both the relevant market and competitive effects analysis across the two distinct groups who interact on the Amex platform, that is, cardholders and merchants. This unified, integrated approach has been controversial, however. The primary debate is whether the Court’s ruling places an undue burden on plaintiffs under the rule of reason paradigm to meet their burden of production to establish harm to competition. Enter Epic v. Apple (Epic): a case involving the legality of various Apple policies governing its iOS App Store, which, like Amex, is a transactional, multisided platform. While both the district court and the Ninth Circuit largely ruled in favor of Apple over Epic, these decisions are of broader interest for their fidelity to Amex.

A careful review of the decisions reveals that the Epic courts operationalized Amex in a practical, sensible way. The courts did not engage in extensive balancing across developers and users as some critics of Amex contended would be required. Ultimately, the courts in Epic (a) considered evidence of effects across both groups on the platform and (b) gave equal weight to both the procompetitive and anticompetitive effects evidence, which, this Article contends, are the essential elements of the Amex precedent. Relatedly, the Epic decisions illustrate that the burden of production on plaintiffs in multisided platform cases is not higher than in cases involving regular, single-sided markets. Additionally, both parties, whether litigating single-sided or multi-sided markets, are fully incentivized to bring evidence to bear on all aspects of the case. Finally, this Article details how the integrated Amex approach deftly avoids potential issues involving the out-of-market effects doctrine in antitrust, which limits what type of effects courts can consider in assessing conduct.

Introduction

Apple’s iPhone is one of the most successful technology products in modern history. Introduced in 2007,¹1.Press Release, Apple, Apple Reinvents the Phone with iPhone (Jan. 9, 2007), https://www.apple.com/newsroom/2007/01/09Apple-Reinvents-the-Phone-with-iPhone (https://perma.cc/AZ9P-93AZ). the smartphone and associated iOS currently commands a fifty-two percent share in the U.S. market²2.US Smartphone Shipments Market Data (Q4 2022 – Q1 2024), Counterpoint (May 27, 2024), https://www.counterpointresearch.com/insights/us-smartphone-market-share (https://perma.cc/V29S-GCZ3). and is responsible for fifty-eight percent of Apple’s revenues.³3.Press Release, Apple, Apple Reports First Quarter Results (Feb. 1, 2024), https://www.apple.com/newsroom/2024/02/apple-reports-first-quarter-results (https://perma.cc/6CFV-2TJX). One does not have to look far to find the opinion that the product has revolutionized mobile communications.⁴4.See, e.g., Rachel Sandler, How the iPhone Changed the Telecommunications Industry, USA Today (July 4, 2017, 1:29 PM), https://www.usatoday.com/story/tech/news/2017/07/04/how-iphone-changed-telecommunications-industry/103154146 (https://perma.cc/54VM-D6P9) (detailing how the iPhone shifted consumer focus away from the mobile carrier to the mobile device itself because iPhone consumers demand a carrier that supports the iPhone more than they demand a specific carrier); Ronald Hamilton, Jr., How the iPhone Changed Society, COM GAP, https://com-gap.org/blog/how-the-iphone-changed-society (https://perma.cc/PZ34-ZNZ2). Yet, since the iPhone’s introduction and the subsequent rollout of the App Store in 2008,⁵5.Press Release, Apple, iPhone App Store Downloads Top 10 Million in First Weekend (July 14, 2008), https://www.apple.com/newsroom/2008/07/14iPhone-App-Store-Downloads-Top-10-Million-in-First-Weekend (https://perma.cc/VC5F-LJQB). Apple has maintained a strict set of governance policies that dictates how developers and users can interact on the iOS platform. These policies are now the subject of antitrust claims alleging that Apple is using these policies not to benefit its users and developers, but to create anticompetitive harm.⁶6.See Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898, 942–52 (N.D. Cal. 2021); Epic Games, Inc. v. Apple, Inc., 67 F.4th 946, 968 (9th Cir. 2023). Further, the growing consensus among government regulators, both foreign and domestic, is that Apple’s platform policies need to be controlled and altered.⁷7.See, e.g., Press Release, Senator Richard Blumenthal, Blumenthal, Blackburn & Klobuchar Introduce Bipartisan Antitrust Legislation to Promote App Store Competition (Aug. 11, 2021), https://www.blumenthal.senate.gov/newsroom/press/release/blumenthal-blackburn-and-klobuchar-introduce-bipartisan-antitrust-legislation-to-promote-app-store-competition (https://perma.cc/42Z2-3MGN) (detailing how “(t)he Open App Markets Act would protect developers’ rights to tell consumers about lower prices and offer competitive pricing; protect sideloading of apps; open up competitive avenues for startup apps, third party app stores, and payment services”); Press Release, European Commission, Digital Markets Act: Commission Designates Six Gatekeepers (Sept. 6, 2023), https://ec.europa.eu/commission/presscorner/detail/en/ip_23_4328 (https://perma.cc/YRZ6-9WGD) (announcing that the European Commission has determined that, inter alia, Apple’s Apple Store is a “gatekeeper” and, thus, subject to the regulatory obligations of the EU’s Digital Markets Act).

Front and center in this debate is Epic Games’ litigation against Apple.⁸8.See Epic Games, 559 F. Supp. 3d at 942–52; Epic Games, 67 F.4th at 968. Epic is the creator of the popular online game Fortnite, which is available on various platforms including PlayStation, Xbox, and Android.⁹9.See Play Fortnite Today!, Epic Games, https://www.fortnite.com/download (https://perma.cc/ZD2T-YYD9); Play Fortnite on Mobile Devices, Epic Games, https://www.fortnite.com/mobile (https://perma.cc/5A9J-7RC2) Prior to August 13, 2020, Fortnite was also available on iOS.¹⁰10. See Juli Clover, Apple Removes Fortnite from App Store (Update: Epic Files Lawsuit Against Apple), MacRumors (Aug. 13, 2020, 11:58 AM PDT), https://www.macrumors.com/2020/08/13/apple-removes-fortnite-from-app-store (https://perma.cc/BU3N-4HVR). However, on that date, Epic filed a complaint against Apple in the Northern District of California alleging that Apple’s various policies governing its App Store violate federal and California antitrust laws.¹¹11. Complaint for Injunctive Relief, Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898 (N.D. Cal. Aug. 13, 2020) (No. 4:20-cv-05640). On the same day, Apple removed Fortnite from the App Store, citing a violation of the terms of service.¹²12. See Clover, supra note 10. Specifically, Epic contends that (1) app developers must exclusively use Apple’s App Store to distribute software on iOS—that is, Epic is prohibited from launching its own app store (a.k.a. “sideloading”); (2) app developers must exclusively use Apple’s in-app-payment (IAP) system for online transactions, which prevents Epic from having a direct financial relationship with users; and (3) app developers are prohibited from “steering” users via notices and advertisements within their apps to encourage users to make purchases outside of the iOS platform,¹³13. Epic Games, 67 F.4th at 968. a policy similar to the issue at the heart of Ohio v. American Express (Amex).¹⁴14. See Ohio v. Am. Express Co. (Amex), 585 U.S. 529, 539-40 (2018) (detailing the steering provision).

Importantly, Epic’s antitrust claims against Apple are not unique. Beyond other litigation against Apple with similar claims,¹⁵15. In re Apple iPhone Antitrust Litig., 846 F.3d 313 (9th Cir. 2017), aff’d sub nom. Apple Inc. v. Pepper, 587 U.S. 273 (2019) (allowing similar claims as Epic but from users rather than developers to proceed beyond a motion to dismiss); Reilly v. Apple Inc. 578 F. Supp. 3d 1098, 1103 (N.D. Cal. 2022) (dismissing claims regarding Apple’s iOS App distribution practice and DPLA); Plaintiffs’ Consolidated Class Action Complaint for Violations of the Sherman Act and California Unfair Competition Law, Cameron v. Apple Inc., No. 4:19-cv-03074 (N.D. Cal. Sept. 30, 2019) (asserting similar claims as Epic from the developer side). Notably, the parties in Cameron v. Apple recently settled. See Cameron, et al. v. Apple Inc., https://smallappdeveloperassistance.com. (https://perma.cc/9V3J-5794). lawsuits contesting various policies of app stores have exploded. Other targets include Google’s Android Play Store, Sony’s PlayStation Store, and Valve’s Steam.¹⁶16. See Complaint, Utah v. Google LLC, No. 3:21-CV-05227 (N.D. Cal. July 7, 2021) (alleging Google Play Store’s 30 percent commission is supra-competitive); Caccuri v. Sony Interactive Ent. LLC, No. 21-cv-03361-RS, 2023 WL 1805137, at *1 (N.D. Cal. Feb. 7, 2023) (challenging Sony’s new distribution practice); Wolfire Games, LLC v. Valve Corp., No. C21-0563-JCC, 2022 WL 1443744, at *1 (W.D. Wash. May 6, 2022) (disputing Valve’s most-favored-nation policy and its 30 percent commission). The parties in Utah v. Google recently settled. See Press Release, Sean D. Reyes, Utah Office of the Attorney General, Attorney General Reyes Announces $700 Million Settlement with Google over Play Store Misconduct (Dec. 19, 2023), https://attorneygeneral.utah.gov/attorney-general-reyes-announces-700-million-settlement-with-google-over-play-store-misconduct (https://perma.cc/A4JX-PFYS). These cases share a common thread: a belief that platform owners are exercising too much control over online exchanges between developers and users. Consequently, the plaintiffs allege that competition is harmed, consumers are denied meaningful choice, and overall innovation is dampened.¹⁷17. See, e.g., Complaint for Injunctive Relief at 3–4, Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898 (N.D. Cal. Aug. 13, 2020) (No. 4:20-cv-05640-YGR) (“Apple’s anti-competitive conduct with respect to iOS app distribution results in sweeping harms to (i) app distributors, who are foreclosed from competing with Apple and innovating new methods of distributing iOS apps to users outside the App Store . . . (ii) app developers, who are denied choice on how to distribute their apps . . . and (iii) consumers, who are likewise denied choice and innovation . . . .”). Decisions regarding these various app store policies will have a profound influence on online commerce as app stores account for trillions in annual revenue.¹⁸18. See e.g., Press Release, Apple, App Store Developers Generated $1.1 Trillion in Total Billings and Sales in the App Store Ecosystem in 2022 (May 31, 2023), https://www.apple.com/cm/newsroom/2023/05/one-point-one-trillion-generated-in-app-store-ecosystem-in-2022 (https://perma.cc/K2S3-NVHK). Further, it seems as if litigation and regulatory proposals are working in lockstep to shape app store polices.¹⁹19. See, e.g., Open App Markets Act, S. 2710, 117th Cong. (2022); Open App Markets Act, H.R. 5017, 117th Cong. (2021); Open App Markets Act, H.R. 7030, 117th Cong. (2022); H.R. 2005, 55th Leg., 1st Reg. Sess. (Ariz. 2021); H.R. 1184, 92d Sess. (Minn. 2021); S. 2577, 31st Leg. (Haw. 2022); S. 2333, 67th Leg. (N.D. 2021); H.R. 4599, 102nd Leg. (Ill. 2022); H.R. 140, 192nd Leg. (Mass. 2021); H.R. 229 (Ga. 2021); S. 4822 (N.Y. 2021).

To that end, the Ninth Circuit’s recent decision in Epic v. Apple, where the court largely affirmed the district court’s ruling in favor of Apple,²⁰20. Epic Games, Inc. v. Apple, Inc., 67 F.4th 946, 973 (9th Cir. 2023). The court only ruled in favor of Epic on its California law claim regarding the anti-steering provision. See id. at 999–1003. represents a critical lesson in how courts consider app store cases specifically and transactional platform cases more generally. The decision constitutes the first substantive ruling at the appellate level on the legality of various app store policies under antitrust law. Given the Ninth Circuit’s judgment, the decision also raises the broader question of the scope and boundaries of the Supreme Court’s ruling in Amex.²¹21. Amex, 585 U.S. 529. Much ink has been spilled over the wisdom of the Court’s Amex decision and the legal bounds of the ruling.²²22. Compare Geoffrey A. Manne, In Defence of the Supreme Court’s ‘Single Market’ Definition in Ohio v American Express, 7 J. Antitrust Enf’t 104, 104–05 (2019) (finding the Court’s conclusion that “both sides of a two-sided market must be considered in defining the relevant market and evaluating the existence and consequences of a firm’s exercise of market—is, indeed, the proper” conclusion), with Tim Wu, The American Express Opinion, Tech Platforms & the Rule of Reason, 7 J. Antitrust Enf’t 117, 122 (2019) (“At bottom, the approach announced (by) the Court is unprecedented, procedurally indefensible, unnecessarily complex, and ultimately incoherent.”). Epic represents a key opportunity to examine the soundness of the Amex decision and the bounds of the precedent. This is especially true given that both cases involve a transactional multisided platform²³23. See Lapo Filistrucchi, Damien Geradin, Eric van Damme & Pauline Affeldt, Market Definition in Two-Sided Markets: Theory and Practice, 10 J. Competition L. & Econ. 293, 298 (2014) (“Two-sided transaction markets, such as payment cards, are instead characterized by the presence and observability of a transaction between the two groups of platform users. As a result, the platform is not only able to charge a price for joining the platform, but also one for using it . . . .”). and an anti-steering provision. Further, both Apple and Epic petitioned the Supreme Court for a writ of certiorari—albeit on separate, targeted issues²⁴24. See Petition for Writ of Certiorari, Epic Games, Inc. v. Apple Inc., No. 23-337 (U.S. Sept. 27, 2023) (petitioning on grounds that the Ninth Circuit failed to properly “balance” the harms and benefits and improperly examined less restrictive alternatives); Petition for Writ of Certiorari, Apple Inc. v. Epic Games, Inc., No. 23-344 (U.S. Sept. 28, 2023) (disputing the scope of the anti-steering injunction imposed by the appellate court to include all developers rather than just Epic). —which the Court ultimately declined.²⁵25. Epic Games, Inc. v. Apple, Inc., 67 F.4th 946 (9th Cir. 2023), cert. denied, No. 23-337, 2024 WL 156473 (US Jan. 16, 2024). Given this situation, the question that future litigants, practitioners, agencies, and judges must address is whether Epic is consistent with the Amex precedent. If so, how? If not, then why not, and what are the implications for other online marketplace cases working their way through the judiciary?

This Article contends that Epic operationalized Amex’s framework for assessing a transactional multisided platform under antitrust’s rule of reason in a practical, sensible way. Part I offers an overview of the rule-of-reason framework and the associated burden-shifting paradigm. It also details the Amex decision and how the Court adapted the rule-of-reason framework to transactional platforms. This Part closes with a discussion of the fallout from the Amex precedent. Next, Part II unpacks both the district and circuit courts’ decisions in Epic and focuses on the fidelity of the decisions to the Amex approach. Finally, in Part III, this Article harmonizes Amex and Epic. This Article argues that the central element of Amex, implemented in Epic, is a balanced view of burdens for procompetitive and anticompetitive effects counted across all the relevant groups on a platform. Namely, both cases applied the rule of reason to (a) credit benefits and harms, even if they accrued to different groups on the platform; and, relatedly, (b) give equal weight to the purported benefits and the alleged harms of a practice. Put simply, the courts did not prejudice the evidence in either direction and considered evidence of effects across all the relevant platform stakeholders. Further, in implementing these critical elements, the courts in both cases (i) did not subject plaintiffs to a heavier burden of production to demonstrate anticompetitive harms—contrary to some characterizations²⁶26.  See infra Section I.C for a fuller discussion of the criticisms of the Amex decision on this point. —and (ii) avoided the tricky issue of dealing with the out-of-markets efficiencies doctrine, which excludes procompetitive effects that occur outside the strict boundaries of the relevant market.²⁷27. See infra Section I.A for further discussion of the doctrine.

Rule of Reason & the Amex Precedent for Platforms

Understanding the impact of Amex on antitrust analysis requires first comprehending the rule-of-reason framework. To that end, this Part reviews the rule-of-reason framework before discussing the Amex decision in depth, including its underlying economic principles and subsequent criticisms.

The Rule of Reason & Burden-Shifting Paradigm

The rule-of-reason framework has been fundamental to the administration of Sherman Act cases since Standard Oil and Chicago Board of Trade.²⁸28. Standard Oil Co. v. United States, 221 U.S. 1, 68 (1911); Bd. of Trade of Chicago v. United States, 246 U.S. 231, 238 (1918) (“The true test of legality is whether the restraint imposed is such as merely regulates and perhaps thereby promotes competition or whether it is such as may suppress or even destroy competition.”); see also Cal. Dental Ass’n v. FTC, 526 U.S. 756, 781 (1999) (“What is required . . . is an enquiry meet for the case, looking to the circumstances, details, and logic of a restraint.”). The objective of this framework is to identify and weigh both procompetitive and anticompetitive effects caused by disputed business conduct.²⁹29. See Herbert Hovenkamp, Antitrust Harm and Causation, 99 Wash. U. L. Rev. 787, 789 (2021) (“Many practices that are challenged under the antitrust laws have effects that can plausibly pull in two directions.”). See generally Oliver E. Williamson, Economies as an Antitrust Defense: The Welfare Tradeoffs, 58 Am. Econ. Rev. 18 (1968). By contrast, for conduct assessed under a per se illegal rule, such as price fixing and territorial allocations, the only question is whether the defendant actually engaged in the prohibited conduct, regardless of procompetitive effects.³⁰30. See, e.g., Herbert Hovenkamp, The Rule of Reason, 70 Fla. L. Rev. 81, 83 (2018) (detailing how, under “antitrust’s ‘per se’ rule . . . anticompetitive effects are largely inferred from the conduct itself”). Thus, the rule of reason raises distinct questions about how to structure the legal proceeding and best identify relevant evidence of both procompetitive and anticompetitive effects, while minimizing error costs and administrative burdens.³¹31. See generally Abraham L. Wickelgren, Determining the Optimal Antitrust Standard: How to Think about Per Se versus Rule of Reason, 85 S. Cal. L. Rev. Postscript 52 (2012) (detailing how the rule of reason’s burden shifting structure can be adapted to the nature of the violation including the likelihood and magnitude of harm); Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1 (1984) (highlighting the various social costs associated with false positives and negatives in antitrust law); Warren F. Schwartz & Gordon Tullock, The Cost of a Legal System, 4 J. Legal Stud. 75, 76 (1975) (“(I)f the enforcement mechanism does not assure perfect accuracy, each party is subject to the risk of a sanction’s being wrongfully imposed even if he does not violate the governing rules (‘the costs of error’).”).

Under the formal rule of reason framework, each side “takes turns”—thus, the framework is a “burden-shifting” exercise between the plaintiff and defendant.³²32. See, e.g., Hovenkamp, supra note 30, at 132; Jonathan B. Baker, The Antitrust Paradigm 68 (2019) (“Antitrust decision rules have continued to evolve and today typically adopt a burden-shifting approach that structures the rule of reason and harmonizes it with per se analysis.”). Specifically, in step one, the plaintiff has the burden to produce evidence of “anticompetitive harm” to consumers (as opposed to competitors³³33. See, e.g., Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 488 (1977) (“The antitrust laws, however, were enacted for ‘the protection of competition not competitors.’” (quoting Brown Shoe Co. v. United States, 370 U.S. 294, 320 (1962))); United States v. Microsoft Corp., 253 F.3d 34, 58 (D.C. Cir. 2001) (“(T)o be condemned as exclusionary, a monopolist’s act must have an ‘anticompetitive effect.’ That is, it must harm the competitive process and thereby harm consumers. In contrast, harm to one or more competitors will not suffice.”); Baker, supra note 32, at 68 (“(A) plaintiff meets a burden of production—it sets forth its prima facie case—by presenting evidence of anticompetitive harm.”). )—which can be shown directly in the form of higher prices, lower output, or reduced quality.³⁴34.  Amex, 585 U.S. at 542; see also Hovenkamp, supra note 30, at 118 (“In rule of reason analysis, the point … is to assess whether the challenged restraint reduces output or increases price from the non-restraint level.”). If the plaintiff meets this prima facie burden, then in step two, the burden shifts to the defense to offer procompetitive justifications, including efficiencies, for the business conduct.³⁵35. See Amex, 585 U.S. at 541; see also Hovenkamp, supra note 30, at 103. See generally Andrew I. Gavil & Steven C. Salop, Probability, Presumptions and Evidentiary Burdens in Antitrust Analysis: Revitalizing the Rule of Reason for Exclusionary Conduct, 168 U. Pa. L. Rev. 2107, 2110 (2017). Then, if the case proceeds to step three, the burden of production returns to the plaintiff, who must “rebut” the defense’s evidence by demonstrating that the purported procompetitive effects could have been achieved through less restrictive alternatives (LRAs).³⁶36. Amex, 585 U.S. at 542; see United States v. Baker Hughes, Inc., 908 F.2d 981, 983 (D.C. Cir. 1990); see also C. Scott Hemphill, Less Restrictive Alternatives in Antitrust Law, 116 Colum. L. Rev. 927, 941 (2016). Finally, some argue that judges add a “step four” to “balance” the various effects and render a judgment.³⁷37. See Michael A. Carrier, The Four-Step Rule of Reason, 33 Antitrust 50, 50 (2019) (“The rule of reason has four steps, not three.”). Throughout the burden shifting paradigm, the plaintiff always has the ultimate burden of persuasion.³⁸38. See Baker Hughes, 908 F.2d at 983 (“(T)he ultimate burden of persuasion . . . remains with the government at all times.”); see also United States, Note, The Standard of Review by Courts in Competition Cases, Organisation Econ. Coop. & Dev. 2 (June 4, 2019), https://one.oecd.org/document/DAF/COMP/WP3/WD(2019)22/en/pdf (“The burden of proof for a violation of law lies with the enforcer or civil plaintiff. The defendant never needs to affirmatively prove their innocence in the United States.”).

While the above blueprint sounds like a linear protocol, the reality of litigation is quite different. First, cases rarely follow such an ordered approach, and the various “steps” are more of a conceptual paradigm. For instance, judges may use backwards induction and incorporate the likely procompetitive arguments when considering the prima facie case.³⁹39. See Herbert Hovenkamp, Antitrust Balancing, 12 N.Y.U. J.L. & Bus. 369, 381 (2016) (“(W)hen the government makes a prima facie case, it already takes into account what might be considered ‘ordinary’ or typical efficiency gains that mergers are likely to produce.”). Second, and relatedly, some courts have adopted a “totality-of-the-circumstances approach” to consider the impact of a disputed business practice.⁴⁰40. Cf. Baker Hughes, 908 F.2d at 984 (detailing how, in assessing mergers, “(t)he Supreme Court has adopted a totality-of-the-circumstances approach . . ., weighing a variety of factors to determine the effects of particular transactions on competition”). See also Chi. Bridge & Iron Co. v. FTC, 534 F.3d 410, 424 (5th Cir. 2008) (“The Ninth and Eleventh Circuits interpret Baker Hughes’ burden-shifting language as describing a flexible framework rather than an air-tight rule.”). This is not to suggest that the burden shifting paradigm does not matter or that it does not serve to guide courts. Rather, the point is that judges rarely explicitly or quantitatively “balance” the various anticompetitive and procompetitive effects.⁴¹41. See, e.g., Hovenkamp, supra note 39, at 373 (“(B)alancing is a very poor label for what courts actually do.”); Gregory J. Werden, Cross-Market Balancing of Competitive Effects: What Is the Law, and What Should it Be?, 43 J. Corp. L. 119, 139-40 (2017) (“(T)he rule of reason asks only which competitive effect from a restraint predominates . . . the determination of a restraint’s predominant effect on competition need not be quantitative or precise.”); Daniel C. Fundakowski, The Rule of Reason: From Balancing to Burden Shifting, Perspectives in Antitrust, ABA Section of Antitrust L., Jan. 22, 2013, at 2 (“Although this is the theoretical methodology, the reality is that courts very rarely reach the balancing exercise.”); Andrew I. Gavil, Burden of Proof in U.S. Antitrust Law, in 1 Issues in Competition Law and Policy 125, 147 (ABA Section of Antitrust Law 2008) (“Such ’rule of reason balancing’ is perhaps the greatest myth in all of U.S. antitrust law. It is almost always described as the final step in the rule of reason analysis, yet few, if any, decisions turned on a true balancing of pro- and anticompetitive effects. Instead, most cases turn on the strength and weight of the evidence of effects or efficiencies.”); William Kolasky, Reinvigorating Antitrust Enforcement in the United States: A Proposal, 22 Antitrust 85, 87 (2008) (“(T)he balancing occurs at each preceding step of the analysis, rather than at the end.”). Judges are continually updating their beliefs based on the flow and totality of the evidence.⁴²42. See, e.g., Easterbrook, supra note 31, at 11 (expressing that judges must make decisions where “everything is relevant, nothing is dispositive”). Along the same lines, the Ninth Circuit in Epic noted it is not even clear there is a balancing step.⁴³43. Epic Games, Inc. v. Apple, Inc., 67 F.4th 946, 993 (9th Cir. 2023) (“Supreme Court precedent neither requires a fourth step nor disavows it. In the Court’s two most recent Rule of Reason decisions, it discussed only the three agreed-upon steps.”).

A related corollary to the rule of reason is the out-of-market efficiencies doctrine.⁴⁴44. See, e.g., Amex, 585 U.S. at 574 (Breyer, J., dissenting) (“A Sherman Act §1 defendant can rarely, if ever, show that a pro-competitive benefit in the market for one product offsets an anticompetitive harm in the market for another.”); United States v. Am. Express Co., 88 F. Supp. 3d 143, 229 (E.D.N.Y. 2015) (observing that “(a)s a general matter . . . a restraint that causes anticompetitive harm in one market may not be justified by greater competition in a different market.”). See also Ted Tatos & Hal Singer, The Abuse of Offsets as Procompetitive Justifications: Restoring the Proper Role of Efficiencies after Ohio v. American Express and NCAA v. Alston, 38 Ga. St. U. L. Rev. 1179, 1188 (2022) (“(T)treatment in merger cases generally rejects offsetting harms in the relevant market with some exogenously derived justifications.”). The doctrine, first established in Philadelphia National Bank (PNB), is that courts should only count effects—whether anticompetitive or procompetitive—strictly within the bounds of the relevant market.⁴⁵45. United States v. Phila. Nat’l Bank, 374 U.S. 321 (1963). This interpretation of PNB is not universally accepted, however. See, e.g., Werden, supra note 41, at 140 (suggesting that PNB did not involve a question of cross-market balancing as the case did not involve “multiple markets across which the court could have balanced”). While this doctrine has some intuitive appeal, critical questions remain regarding both its scope and its wisdom given its current formulation.⁴⁶46. See Daniel A. Crane, Balancing Effects Across Markets, 80 Antitrust L.J. 397, 397 (2015) (arguing that the PNB principle is best operationalized as a rebuttable presumption); see also John M. Yun, Reevaluating Out of Market Efficiencies in Antitrust, 54 Ariz. St. L.J. 1261 (2022) (proposing an alternative structure to consider out-of-market efficiencies). For instance, whether the doctrine applies beyond merger cases under Section 7 of the Clayton Act is unclear.⁴⁷47. See Werden, supra note 41, at 126 (“Philadelphia National Bank did not create a rule applicable in Sherman Act cases, and no subsequent merger decision by the Court has been cited as authority for the merger-specificity rule.”). Additionally, the economic justification for the doctrine is exceedingly weak given the potential interrelationships across various relevant markets.⁴⁸48. See Yun, supra note 46, at 1291 (“(H)arms and benefits are two levers in antitrust. They may correspond perfectly such that the relevant market to examine harms precisely captures all the benefits as well. Or . . . the economics of the conduct may not follow such a neat mapping.”). While the debate over “counting” out-of-market effects is beyond the scope of this Article, the lack of clarity about the state of the doctrine is relevant. Indeed, the Ninth Circuit in Epic openly wrestled with this question.⁴⁹49. In considering Epic’s argument that Apple’s procompetitive effects impacted iOS (that is, a market outside of the relevant market) and not the App Store, the Ninth Circuit ultimately rejected the argument because it found Apple’s effect did impact the App Store. However, in reaching that conclusion, the court highlighted the doctrinal mess. See Epic Games, Inc. v. Apple, Inc., 67 F.4th 946, 989 (9th Cir. 2023) (“The Supreme Court’s precedent on this issue is not clear. . . . While we have never expressly confronted this issue, we have previously considered cross-market rationales when applying the Rule of Reason. . . . We decline to decide this issue here.”).

All that said, what role does the rule of reason paradigm really play? Arguably, the primary value of the framework is simply to offer courts a coherent structure to organize the evidence.⁵⁰50. See Fundakowski, supra note 41, at 3 (asserting that, while the rule of reason is rarely strictly followed in sequence in court cases, the value is in “maintaining the court’s focus,” providing “predictability,” and allowing for “judicial economy”). To that end, how the rule of reason is actually implemented is, to a degree, fungible. Based largely on the nature of the conduct and the degree to which anticompetitive and procompetitive effects must be demonstrated, the rule of reason is similar to a sliding scale: the slide can move from close to per se condemnation to nearly per se legality.⁵¹51. See Cal. Dental Ass’n v. FTC, 526 U.S. 756, 779 (1999); Nat’l Soc’y of Pro. Eng’rs v. United States, 435 U.S. 679 (1978); Verizon Commc’ns, Inc. v. Law Offs. of Curtis V. Trinko, LLP, 540 U.S. 398, 411 (2004) (finding there are “few existing exceptions from the proposition that there is no duty to aid competitors”); see also Hovenkamp, supra note 30, at 128 (“While the Court (in Engineers) did not speak of a ‘quick look’ or articulate its mode of analysis, it was clearly applying something that fell between per se and full rule of reason analysis.”). Consequently, in the following Section, this Article examines how the Court in Amex adapted the rule of reason to assess conduct on transactional platforms.

The Amex Decision & the Impact on the Rule of Reason

The emergence of online platform businesses with network effects is one of the central stories in commerce over the past thirty years.⁵²52. See, e.g., Michael A. Cusumano, Annabelle Gawer & David B. Yoffie, The Business of Platforms 8–12 (2019). As these business organizations have proliferated, the economic literature has tracked their growth. First came the economic research on network effects.⁵³53. See, e.g., Ronald Artle & Christian Averous, The Telephone System as a Public Good: Static and Dynamic Aspects, 4 Bell J. Econ. & Mgmt. Sci. 84 (1973); Jeffrey Rohlfs, A Theory of Interdependent Demand for a Communications Service, 5 Bell J. Econ. & Mgmt. Sci. 16 (1974); Michael L. Katz & Carl Shapiro, Network Effects, Competition, and Compatibility, 75 Am. Econ. Rev. 424, 424 (1985). Second came the rich literature on multisided platforms.⁵⁴54. See, e.g., Michael R. Baye & John Morgan, Information Gatekeepers on the Internet and the Competitiveness of Homogeneous Product Markets, 91 Am. Econ. Rev. 454, 454, 470 (2001); Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 1 J. Eur. Econ. Ass’n 990, 1018 (2003) (“The quest for ‘getting both sides on board’ makes no sense in a world in which only the total price for the end user interaction, and not its decomposition, matters.”); David S. Evans, The Antitrust Economics of Multi-Sided Platform Markets, 20 Yale J. on Regul. 320, 334–35 (2003); Jean-Charles Rochet & Jean Tirole, Two-Sided Markets: A Progress Report, 37 RAND J. Econ. 645, 664–65 (2006). Eventually, the economic learning on multisided platforms came to a head in a 2018 Supreme Court case, Amex.⁵⁵55. Amex, 585 U.S. 529.

The case involved a challenge by various states to Amex’s anti-steering policy,⁵⁶56. The Department of Justice (DOJ) was involved through the Court of Appeals but did not join the states in their petition for certiorari to the Supreme Court. which governs how merchants on the Amex network can promote other credit cards to Amex cardholders.⁵⁷57. Amex, 585 U.S. at 533. Specifically, at the point of sale, the policy prohibits merchants in the network from “steering” an Amex cardholder to another card,⁵⁸58. The policy did not prohibit steering to other payment methods, such as cash, credit, and debit cards. See id. at 539. such as the Discover card, by—for instance—offering a price discount.⁵⁹59. Id. In lieu of the policy, merchants may have incentive to steer Amex cardholders because Amex charges higher “merchant fees,” which represent the percentage of the transaction that the merchant pays to the credit card company.⁶⁰60. Id. Expressly, the states alleged that the policy violated Section 1 of the Sherman Act because it raised prices by preventing steering discounts.⁶¹61. Id. Amex countered with the arguments that: (1) merchants are free to accept or decline participation in the network; (2) Amex does not have monopoly power; (3) without the policy, merchants could free-ride on Amex’s promotional services; and (4) steering would unravel the source of differentiation between other credit cards and Amex, which offers generous rewards and premium services.⁶²62. Id. at 538-39; see also Brief for Respondents American Express Company and American Express Travel Related Services Company, Inc. at 10, Ohio v. Am. Express Co., 585 U.S. 529 (Jan. 16, 2018) (No. 16-1454) (“(M)erchant steering undermines the investment that Amex makes — through Membership Rewards and other benefits — to encourage its cardholders to use Amex rather than one of the other cards that almost all Amex cardholders also carry. Merchant steering thus ‘interferes with a network’s ability to balance its two-sided net price.’”).

In deciding the case, the Supreme Court defined the relevant product market by using an integrated credit card transactions market⁶³63. See Amex, 585 U.S. at 546 (“In two-sided transaction markets, only one market should be defined.”); id. (explaining the Court analyzed “the two-sided market for credit-card transactions as a whole to determine whether the plaintiffs have shown that Amex’s antisteering provisions have anticompetitive effects”). rather than delineating two separate markets—one for cardholders and one for merchants.⁶⁴64. In contrast, some have strongly advocated for a standard where evidence of harm to only one group is sufficient to find an antitrust violation, largely based on procedural grounds. See, e.g., Brief of 28 Professors of Antitrust Law as Amici Curiae Supporting Petitioners at 22, Ohio v. Am. Express Co., 585 U.S. 529 (2017) (No. 16-1454) (hereinafter Brief of 28 Professors) (“(T)his Court held, in a case where the defendant operated a two-sided platform, that each side represented a ‘separate . . . market’ and that injuring competition in the restrained market alone was sufficient to violate the Sherman Act.”). The central logic motivating the Court’s market definition is that, for transactional platforms such as a credit card network, cross-group network effects are significant and relevant to understand conduct.⁶⁵65. Amex, 585 U.S. at 535 (“A credit card, for example, is more valuable to cardholders when more merchants accept it, and is more valuable to merchants when more cardholders use it.”); see generally Catherine Tucker, Digital Data, Platforms and the Usual (Antitrust) Suspects: Network Effects, Switching Costs, Essential Facility, 54 Rev. Indus. Org. 683, 685 (2019) (“(E)conomists classically think of . . . ‘indirect network effects’ or a ‘cross-side’ network effect. Indirect network effects occur when the value of the product or service is larger for a certain group of users, when more users who are in a different category of users are using the service.”). Further, the incentive of the platform is to optimize over all the participating groups on the platform—rather than profit maximize over one or the other.⁶⁶66. See Amex, 585 U.S. at 547 (“(C)ompetition cannot be accurately assessed by looking at only one side of the platform in isolation.”); see also Andrei Hagiu, Proprietary vs. Open Two-Sided Platforms and Social Efficiency 19 (AEI-Brookings Joint Ctr., Working Paper No. 06-12, 2006) (“By being able to balance the interests of the two sides through its pricing structure, a proprietary platform may come closer to the socially optimal level of adoption than a platform simply pricing at marginal cost on both sides.”); Benjamin Klein, Andres V. Lerner, Kevin M. Murphy & Lacey L. Plache, Competition in Two-Sided Markets: The Antitrust Economics of Payment Card Interchange Fees, 73 Antitrust L.J. 571, 574, 595, 598, 626 (2006). Due to this binding thread of network effects and a common output from transactions, the Court deemed it necessary, under the rule-of-reason framework, for the plaintiff to consider the welfare impact of the policy on both merchants and cardholders. It was not enough to demonstrate impacts on, for instance, merchant welfare.⁶⁷67. Amex, 585 U.S. at 547 (“Evidence of a price increase on one side of a two-sided transaction platform cannot by itself demonstrate an anticompetitive exercise of market power.”).

Operating under this structure, the Court decided that the plaintiffs failed to meet their burden of production to demonstrate anticompetitive effects,⁶⁸68. Id. at 552 (“In sum, the plaintiffs have not satisfied the first step of the rule of reason.”). which, importantly, is not the same as concluding that the policy is procompetitive. In particular, the Court pointed to a lack of evidence of effects on quantity, quality, or innovation.⁶⁹69. Id. at 548. There was some evidence of price effects, but the Court viewed this evidence as noisy and questioned any causal claims that the anti-steering provision was responsible for a price increase.⁷⁰70. Id. 548-49 (“(T)he cause of increased merchant fees is not Amex’s antisteering provisions, but rather increased competition for cardholders and a corresponding marketwide adjustment in the relative price charged to merchants.”). This conclusion aligns with the economic literature that finds platforms structure prices differently than single-sided markets due to the presence of network effects.⁷¹71. See, e.g., Rochet & Tirole, supra note 54, at 646. In doing so, the side that experiences a price increase may be worse off, but overall welfare could increase by generating greater network effects and ultimately more output and innovation.⁷²72. For example, consider an advertising platform’s decision to limit, at the top of a search results page, the number of ads. This benefits consumers who dislike search results populated with ads but harms advertisers. Alternatively, consider a platform’s decision to require developers to pass robust privacy standards. If developers must receive consent from users to collect geolocation data, then such a policy may cause net harm to developers, but this could potentially bring greater net welfare to users. The Court also pointed to a lack of evidence that the policy caused a reduction in output—the number of transactions.⁷³73. See Amex, 585 U.S. at 548. Some of the output evidence that pointed to an expansion of transactions was not demonstratively causal. This evidence was also immaterial to the decision that the plaintiffs did not carry their burden and is thus dicta.⁷⁴74. See id. at 548–49. Importantly, the Court’s comments on the benefits of output evidence do not create a precedent that proof of anticompetitive effects generally requires output evidence, nor did the Court establish such a precedent in transactional platform cases specifically.⁷⁵75. See, e.g., John M. Newman, The Output-Welfare Fallacy: A Modern Antitrust Paradox, 107 Iowa L. Rev. 563, 565 (2022) (advancing that “(i)n AmEx, a 5-4 majority announced that the government needed to demonstrate an output reduction, despite abundant evidence that the challenged restraints had stifled innovation, increased the prices of nearly every good and service sold at retail in the United States”). Notably, the Court merely listed output evidence as one of several potential avenues to demonstrate anticompetitive effects. See Amex, 585 U.S. at 542 (“Direct evidence of anticompetitive effects would be ‘proof of actual detrimental effects (on competition),’ . . . such as reduced output, increased prices, or decreased quality in the relevant market.”).

Ultimately, Amex adjusted the rule of reason for transactional platforms when cross-group network effects are relevant to understand a business practice. In these cases, the relevant market must integrate all the relevant groups on the platform. Consequently, antitrust liability requires more than a demonstration of a policy’s harm to one group on a platform without reference to the impact on the other relevant groups.

The Fallout from the Amex Decision

The Court’s adaptation of the rule of reason to transactional, multisided platforms has been controversial. Supporters find the case properly incorporated the economic literature on network effects and platforms and created a sensible rule for transaction platforms when network effects are relevant.⁷⁶76. See, e.g., David S. Evans & Richard Schmalensee, Antitrust Analysis of Platform Markets: Why the Supreme Court Got It Right in American Express (2019); Manne, supra note 22. Critics condemn the decision for improperly defining an integrated relevant market;⁷⁷77. See Herbert Hovenkamp, Platforms and the Rule of Reason: The American Express Case, 2019 Colum. Bus. L. Rev. 35, 37, 49; John B. Kirkwood, Antitrust and Two-Sided Platforms: The Failure of American Express, 41 Cardozo L. Rev. 1805, 1837 (2020). unreasonably requiring the assessment of welfare across all relevant groups on the platform;⁷⁸78. See Brief of 28 Professors, supra note 64, at 11. undermining antitrust’s common law process;⁷⁹79. See Michael L. Katz & A. Douglas Melamed, Competition Law as Common Law: American Express and the Evolution of Antitrust, 168 U. Pa. L. Rev. 2061, 2081 (2020). and hurting tech workers.⁸⁰80. See Lina Khan, The Supreme Court Just Quietly Gutted Antitrust Law, Vox (July 3, 2018, 9:40 AM), https://www.vox.com/the-big-idea/2018/7/3/17530320/antitrust-american-express-amazon-uber-tech-monopoly-monopsony (https://perma.cc/NZH6-SZTH).

The most common thread is that the decision raised the evidentiary burden on plaintiffs.⁸¹81. See, e.g., Fiona Scott Morton, Kartikeya Kandula & Karissa Kang, Do We Need a New Sherman Act?, 2022 Colum. Bus. L. Rev. 42, 61 (arguing that Amex exemplifies the “excessively high standards of liability”); Katz & Melamed, supra note 79, at 2016 (“As it happens, every one of the Court’s controversial rulings (in Amex) and departures from sound common law adjudication benefitted the defendants.”); Wu, supra note 22, at 122 (“American Express takes the less-appealing approach of using complex economic theory to create near-impossible burdens of proof—burdens particularly hard to meet when they emerge on appeal.”). Specifically, a key criticism is that defendants are in best the position to offer procompetitive evidence, yet Amex requires the plaintiff to provide this evidence instead.⁸²82. See, e.g., Hovenkamp, supra note 77, at 57 (“Because the defendant is the creator of its restraint and presumably knows what its motives were, it is in a far better position to provide proof of its rationale and effects.”); Kirkwood, supra note 77, at 1813 (“It is inefficient because the defendant possesses the relevant information, not the plaintiff.”). As a result, critics contend that Amex (a) elevated the burden on plaintiffs to show some degree of net harm in step one of the rule-of-reason framework,⁸³83. See, e.g., Steven C. Salop, Daniel Francis, Lauren Sillman & Michaela Spero, Rebuilding Platform Antitrust: Moving on from Ohio v. American Express, 84 Antitrust L.J. 883, 884 (2022) (“The practical inheritance of Amex is plain to see: the burdens faced by plaintiffs have been needlessly increased, and enforcement efforts have been obstructed.”); Kirkwood, supra note 77, at 1813 (“Under the American Express Court’s version of the rule of reason, the plaintiff must show, in the very first step, that the challenged conduct produces net harm across the entire platform.”). and (b) simultaneously lowered the burden on defendants in step two. At the extreme, critics suggest Amex rendered step two moot given all the work that plaintiffs must do in step one.⁸⁴84. See, e.g., Harry First, American Express, the Rule of Reason, and the Goals of Antitrust, 98 Neb. L. Rev. 319, 336 (2019) (“Plaintiffs should not be tasked with dreaming up and disproving all possible justifications that a defendant could raise in support of restraints that have proven adverse effects on competition. If there is a really good justification for a restraint, let the defendant prove it.”); Kirkwood, supra note 77, at 1823 (“The burden of establishing a market failure, however, should rest on the defendant, not the plaintiff. . . . If the burden of disproving the existence of any market failures rested on the plaintiff, it would have to raise and rebut every reasonable possibility, which would be inefficient.”). In other words, plaintiffs must effectively fold steps one and two into an integrated effects analysis that anticipates the defendant’s procompetitive arguments to arrive at some net effect on competition.

Unpacking the Epic Decision

Fundamentally, the criticisms of Amex raise fears that Amex disrupted the standard rule of reason paradigm and placed a thumb on the scale to favor a defendant’s ability to avoid liability. After all, if the plaintiff must incorporate the welfare of all the groups on a platform, then what does the defense do? To begin to address that question, this Part examines how the district and appellate courts in Epic operationalized Amex. It focuses particularly on the burdens for both plaintiffs and defendants to demonstrate effects from Apple’s policies.

The District Court’s Decision

The dispute between Epic Games and Apple began on August 13, 2020, when Epic filed a complaint against Apple in the Northern District of California.⁸⁵85. Complaint for Injunctive Relief, Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898 (N.D. Cal. Aug. 13, 2020) (No. 4:20-cv-05640-YGR). Epic’s primary position is that Apple is a monopolist over locked-in users of iOS, and, in turn, Apple uses its monopoly power over users and developers who exchange on the iOS’s App Store, in violation of Sections 1 and 2 of the Sherman Act and California’s Unfair Competition Laws (UCL). Specifically, Epic challenges Apple’s requirement that all developers distribute software via the App Store (“exclusivity of distribution”), receive payment for in-app-purchases (IAPs) using Apple’s IAP system (“exclusivity of payment”), and abstain from steering app users to alternative platforms (“prohibition on steering”).

On September 21, 2021, the district court ruled on the merits of these theories of harm, finding largely in favor of Apple.⁸⁶86. Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898 (N.D. Cal. 2021). The exception is a finding that Apple violated California’s UCL with the anti-steering provision. While this Article focuses largely on the federal antitrust claims, the district court’s finding on anti-steering does not demonstrate that Epic is inconsistent with Amex. Crucially, the standards for a Sherman Act violation are materially different from a UCL violation.⁸⁷87. The court invoked both a “tethering test” and a “balancing test” to assess whether Apple’s anti-steering provision violated the UCL. See id. at 1053. Both tests are unrelated to the rule of reason used to assess most federal antitrust violations. See id. at 1055 (“Thus, although Epic Games has not proven a present antitrust violation, the anti-steering provisions ‘threaten() an incipient violation of an antitrust law’ by preventing informed choice among users of the iOS platform . . . .”). Additionally, Epic’s fidelity is Amex is not on the legality of steering provisions per se, which can be implemented in differentiated ways,⁸⁸88. In fact, the court highlights the differences that it perceives between Apple’s and Amex’s steering policies. Id. at 1056. but on the procedural aspects of considering the benefits and harms from a disputed business practice on a transactional platform.

The Relevant Market

The district court began its analysis with an assessment of the relevant market.⁸⁹89. Id. at 954. Epic alleged a monopoly market where the only relevant area of commerce is Apple’s “own system” or iOS.⁹⁰90. Id. at 921 (“Epic Games structured its lawsuit to argue that Apple does not compete with anyone; it is a monopoly of one.”). In the parlance of antitrust law, the alleged “Apple-only” markets are “aftermarkets,” that is, “an aftermarket for the distribution of iOS apps” and “an aftermarket for payment processing for iOS apps.”⁹¹91. Id. at 954. Broadly, the aftermarkets doctrine, derived from Eastman Kodak v. Image Technical Services,⁹²92. 504 U.S. 451 (1992). is that, in some instances, consumers who may enjoy competitive options in the “foremarket” may nonetheless be subject to various monopoly aftermarkets once the consumer is locked into one of the foremarket products. The aftermarket monopolist may then engage in opportunistic conduct, depending on user switching costs. Epic’s aftermarket theory is that, even though Apple may compete with Android in the foremarket for mobile operating systems, Apple has monopoly power over the locked-in users in the iOS aftermarket. Consequently, Apple has an incentive to engage in ex post opportunistic conduct affecting these locked-in users. In contrast, Apple alleged a market that includes all digital video game platforms, including the Android Play Store, the Sony PlayStation Store, and the Microsoft Xbox Store.⁹³93. Epic Games, 559 F. Supp. 3d at 972-87. Apple’s market focused on online games rather than a broader app download market because gaming apps represent an outsized percentage of commerce on the App Store.⁹⁴94. Id. at 953 (“In 2016 for instance, despite game apps only accounting for approximately 33% of all app downloads, game apps nonetheless accounted for 81% of all app store billings that year.” (emphasis removed)).

Yet, the district court rejected both Epic’s aftermarkets and Apple’s expansive “all digital video games” market.⁹⁵95. Id. at 921. First, the court decided Epic’s “proposed foremarket [of Apple’s iOS] is entirely litigation driven, misconceived, and bears little relationship to the reality of the marketplace.”⁹⁶96. Id. at 955. The district court rejected Epic’s aftermarkets because the aftermarkets doctrine requires not just lock-in and the existence of switching costs, but some degree of consumer “bait and switch” or persistent lack of knowledge about the lock-in.⁹⁷97.  Id. at 958 (“From a broad perspective, Epic Games did not conduct any analysis of whether consumers know that they are buying into a walled garden. . . . Without a consumer survey, there is no evidence that consumers are unaware of walled garden before purchasing the smartphone. Thus, there is no ‘bait-and-switch.’”); id. at 960 (“Apple’s evidence strongly suggests that low switching between operating systems stems from overall satisfaction with existing devices, rather any ‘lock-in.’”). On this count, the record lacked evidence that iOS consumers are suffering from misinformation or a general lack of understanding about IAP.⁹⁸98. Id. at 958-60.

Instead of using the proposed markets, the court assessed Apple’s policies using a “digital mobile gaming transactions” market.⁹⁹99. Id. at 921. The court found network effects to be highly relevant to understanding the disputed conduct.¹⁰⁰100.   Id. at 994 (“Thus, indirect network effects often dominate and create a ‘winner-take-all’ system that allows only a few large platforms to survive.”); id. at 1007 (“This is consistent with the indirect network effects identified by Dr. Schmalensee (Apple’s economic expert): the small burden on developers maintains a healthy ecosystem that ultimately benefits both sides. Thus, the evidence shows that developers both benefit and suffer from app distribution restrictions.”); id. at 1011 (“By providing a consistent and trusted user experience, IAP encourages users to spend freely, which benefits developers through indirect network effects and has resulted in millions of dollars of revenue.”). Consequently, the court included both the platform’s users and developers in defining the relevant market, in line with Amex.¹⁰¹101.   Id. at 987.

The Competitive-Effects Analysis

Turning to the core assessment, the court had to decide how to structure the competitive-effects analysis within the relevant market. On this point, the court reasoned that “[i]n two-sided transaction markets, an anticompetitive price or restriction on one side may well reflect a competitive equilibrium on the other side. Thus . . . competitive effects can only be determined after carefully considering both sides of the transaction (developers and users), including any indirect network effects.”¹⁰²102.   Id. at 994. This reasoning is firmly in line with the logic of the Amex Court.

The district court first addressed the Section 1 claim regarding exclusivity in distribution and payment within the App Store.¹⁰³103.   Id. at 1033. In doing so, the district court explicitly referenced Amex’s rule of reason framework.¹⁰⁴104.   Id. at 1034. Importantly, the court also noted that “[t]he three steps ‘do not represent a rote checklist’ and are not ‘an inflexible substitute for careful analysis.’”¹⁰⁵105.   Id. Rather, they serve ‘‘to furnish ‘an enquiry meet for the case, looking to the circumstances, details, and logic of a restraint.’”¹⁰⁶106.   Id.

Starting with the prima facie case of anticompetitive effects, the court cited Amex’s litany of possible metrics that could be used to demonstrate harm, including supracompetitive pricing, output restrictions, and other indicators of stifled competition.¹⁰⁷107.   Id. at 1036. Ultimately, the district court found “some” evidence of anticompetitive effects.¹⁰⁸108.   Id. at 1037. First, the court noted “direct evidence” that Apple has not lowered its 30% commission rate over time despite developer complaints.¹⁰⁹109.   Id.; see also id. at 1036 (“Apple’s commission rate has remained static throughout even though Google, Apple’s main competitor (and who also charges a 30% commission rate), does not have the same app distribution restrictions.”). This is the type of direct evidence that Amex suggested as possible evidence of harm. Moving to the indirect evidence, the court considered (a) the 30% commission remaining stable and not falling as Apple’s market share grew to 55% and (b) a “but for” world where developers and users have more choice and lower commissions.¹¹⁰110.   Id. at 1037. The court later explains that conduct reducing consumer choice is insufficient to prove anticompetitive effects. Id. at 1038. The court found that this indirect evidence supported the conclusion that Apple’s exclusionary policy of in-app distribution generates anticompetitive effects under Section 1.¹¹¹111.   Id. at 1037.

The court then moved to Apple’s procompetitive justifications.¹¹²112.   Id. Apple argued its governance policies produced three primary procompetitive effects: (1) security, (2) promotion of intrabrand competition, and (3) protecting intellectual property investments.¹¹³113.   Id. More or less, the court found that all three rationales were valid.¹¹⁴114.   Id. at 1040. Notably, the court examined these procompetitive effects in step two of the analysis, with Apple bearing the burden of production. Finally, the court examined Epic’s proposed LRAs and rejected all of them.¹¹⁵115.   Id. at 1040-41.

The court then moved to consider Apple’s policies requiring use of the IAP for payment.¹¹⁶116.   Id. The analysis largely follows the app store exclusivity analysis, and, similarly, the court concluded that the IAP policies do not violate Section 1.¹¹⁷117.   Id. at 1043. Next, the court addressed the Section 2 monopolization claim and found that Apple does not possess the requisite level of market power for a Section 2 violation.¹¹⁸118.   Id. at 1043-44. Finally, the court examined the Section 1 tying claim for the IAP system and the App Store.¹¹⁹119.   Id. at 1044. The court dismissed the claim. It found that the IAP system was integrated, not tied, with the iOS and App Store, and thus was not a separate product.¹²⁰120.   Id. at 1046. Although the Ninth Circuit noted this analysis was erroneous, the error was harmless because the appellate court affirmed the dismissal of the tying claim.¹²¹121.   See Epic Games, Inc. v. Apple, Inc., 67 F.4th 946, 994-6 (9th Cir. 2023).

Takeaways from the District Court Decision

What are some big-picture observations about the role of two-sided platforms in the district court’s ruling? The district court repeatedly referenced and cited Amex, and it considered the welfare of both developers and users. The key conceptual link between the district court and the Amex opinions is the existence of cross-group, or indirect, network effects. While users are the direct beneficiaries of the App Store’s policies protecting security and privacy, the court also found that “developers benefit from the safe environment created by the App Store.”¹²²122.   Epic Games, 559 F. Supp. 3d at 1007. Specifically, “[b]ased on a trusted environment, users download apps freely and without care, which benefits small and new developers whose apps might not be downloaded if users felt concern about safety.”¹²³123.   Id. Thus, ultimately, “the evidence shows that developers both benefit and suffer from app distribution restrictions.”¹²⁴124.   Id. This is a key finding because, without consideration of these network effects, a one-sided analysis would have likely concluded that developers were harmed.

Critically, the court operationalized the Amex framework to arrive at the above conclusion. The court did not impose herculean hurdles on the plaintiff to demonstrate harm, nor on the defendant to demonstrate benefits. Rather, the court considered both anticompetitive and procompetitive effects, acknowledged the network effects that created trade-offs—particularly for developers—and determined that there is no sound basis to conclude that the platform policies harmed the competitive process or the platform participants.

The Appellate Court’s Decision

The appellate court largely affirmed the district court’s ruling that Apple’s practices are not anticompetitive.¹²⁵125.   Epic Games, 67 F.4th at 966. The appellate court also affirmed that the anti-steering policy violated California’s UCL. Id. Notably, the appellate court also highlighted errors in the district court’s opinion on product market; however, the court labeled these errors as “harmless” and tangential to the core analysis. Id. at 973. Like the district court, the appellate court decision begins with market definition.¹²⁶126.   Id. To analyze Epic’s proposed aftermarkets, the appellate court aptly explained the aftermarkets doctrine requirements established in Kodak.¹²⁷127.   Id. at 976–77. Notably, the court highlighted that consumers must be unaware of the aftermarket consumable or service requirements and must in some way be hindered in their understanding and ability to calculate life-cycle pricing.¹²⁸128.   Id. Thus, the court concluded Epic failed to prove its aftermarkets given the lack of evidence that consumers are unaware of Apple’s app distribution and IAP policies when they purchase an iOS device.¹²⁹129.   Id. at 973. Ultimately, the court upheld the district court’s “middle-ground” market definition of mobile games transactions.¹³⁰130.   Id. at 981.

Next, the appellate court turned to the Section 1 claims of exclusivity in app distribution and payment systems.¹³¹131.   Id. The court began by addressing Apple’s argument that the district court erred in finding anticompetitive effects in step one of the analysis.¹³²132.   Id. at 983. Similar to the district court, the appellate court delineated the discussion into assessments of both direct and indirect evidence. In terms of direct evidence, Apple argued that the finding of anticompetitive effects lacked support because Epic did not show that Apple reduced output.¹³³133.   Id. at 984. The appellate court appropriately noted that Amex did not actually require a reduction in output to meet this burden.¹³⁴134.   Id. Apple’s second argument was that the finding of supracompetitive pricing cannot hold as a matter of law because Apple never raised its commission.¹³⁵135.   Id. Apple’s argument is fundamentally that, if Apple entered the market with a 30% rate when it had no market share, then how can the rate represent a supracompetitive overcharge when its market share grew?¹³⁶136.   Apple’s argument assumes the quality of the service to developers has remained constant. The argument is stronger if quality has increased; if so, then a constant rate means the quality-adjusted price decreases as quality grows. Nonetheless, the appellate court rejected Apple’s argument because that argument has no support in legal precedent.¹³⁷137.   Epic Games, 67 F.4th at 984 (“Apple cites no binding precedent in support of its proposition that the charging of a supracompetitive price must always entail a price increase, though we recognize that it ordinarily does.”). Apple’s third argument was that the 30% rate is similar to competitors’ rates.¹³⁸138.   Id. at 984–85. While the appellate court rejected Apple’s argument because of the difference between charging list prices and effective, negotiated prices, the larger point is that commonality, while relevant, does not disprove a finding of supracompetitive rates. Finally, Apple argued that Amex requires Epic to establish anticompetitive effects for both sides of the platform.¹³⁹139.   Id. at 985. On this vital point, the appellate court rightly dismissed this reading of Amex:

We have previously held: “Amex does not require a plaintiff to [show] harm to participants on both sides of the market. All Amex held is that to establish that a practice is anticompetitive in certain two-sided markets, the plaintiff must establish an anticompetitive impact on the ‘market as a whole.’”¹⁴⁰140.   Id. (quoting PLS.com, LLC v. Nat’l Ass’n of Realtors, 32 F.4th 824, 839 (9th Cir. 2022)).

Turning to the indirect evidence, the appellate court dismissed Apple’s challenge that the finding of harm was “speculative.”¹⁴¹141.   Id. Doubling down on the district court’s proposed “but for” world where alternative app stores would flourish, the appellate court explained that developers and users would enjoy the fruits of a more differentiated app store market.¹⁴²142.   Id. Thus, citing North American Soccer League to support the proposition that reducing consumer choice satisfies step one, the appellate court found that the district court’s indirect evidence finding was sufficient.¹⁴³143.   Id. at 985. Notably, North American Soccer League did not turn on reduced consumer choice; the plaintiff sought to prove its case via indirect evidence of barriers to entry. N. Am. Soccer League, LLC v. U.S. Soccer Fed’n, Inc., 883 F.3d 32, 42–43 (2d Cir. 2018). This conclusion is consistent with the court’s position that, for indirect evidence, “[t]his inquiry need not always be extensive or highly technical.”¹⁴⁴144.   Epic Games, 67 F.4th at 983.

After discussing the evidence of harm, the appellate court then turned to the procompetitive effects.¹⁴⁵145.   Id. at 985–86. The court first examined Epic’s claim that Apple’s procompetitive rationales (security, privacy, and IP compensation) occur outside of the relevant market.¹⁴⁶146.   Id. at 989. Specifically, Epic argued that the effects occur in the iOS foremarket and not the district court’s mobile gaming transaction market.¹⁴⁷147.   Id. As an initial matter, the court commented on the state of the law on out-of-market effects.¹⁴⁸148.   Id. Intriguingly, the court asserted that the law is unsettled about counting cross-market or out-of-market efficiencies—citing Supreme Court and Ninth Circuit cases for this assessment.¹⁴⁹149.   Id. This assertion runs contrary to some scholarship that contends that out-of-market efficiencies are excluded under Philadelphia National Bank and Topco,¹⁵⁰150.   See, e.g., Tatos & Singer, supra note 44, at 1187–88 (“United States v. Topco Associates illuminated the rule-of-reason analysis and discarded the logic of attempting to balance cross-market economic harms.”); Baker, supra note 32, at 190 (“Consistent with the case law involving harms to suppliers, antitrust law does not permit courts to offset competitive harms in one market with competitive benefits in another . . . . The same rule (not permitting benefits in one market to offset harms in another) applies in non-merger litigation.”). though others challenge this interpretation of these cases.¹⁵¹151.   See, e.g., Werden, supra note 41, at 126; Yun, supra note 46, at 1288 (“Therefore, we should be cautious when interpreting the out-of-market efficiencies principle established in PNB (and, for some, Topco) as fundamentally being about disallowing the comparison of welfare across groups.”). Regardless, the court concluded that Apple’s procompetitive effects occur within the relevant market and met Apple’s rebuttal burden.¹⁵²152.   Epic Games, 67 F.4th at 990.

After upholding the procompetitive effects analysis, the appellate court decision tackled Epic’s challenge to the district court’s LRAs conclusion. Again, the appellate court agreed with the district court’s analysis.¹⁵³153.   Id. Finally, the court considered a balancing step, in part because Epic alleged that the district court erred by not explicitly conducting a fourth “balancing step” under the rule of reason framework.¹⁵⁴154.   Id. at 993. The appellate court decided, despite a lack of an explicit Supreme Court precedent, that there does need to be a balancing step; however, the district court’s lack of explicit balancing was “harmless.”¹⁵⁵155.   Id. Ultimately, the court asserted that this step, while needed, can merely reaffirm the prior steps.¹⁵⁶156.   Id. at 994.

The appellate court also addressed the Section 1 tying claim¹⁵⁷157.   Id. and the Section 2 monopolization claim.¹⁵⁸158.   Id. at 998. On the tying claim, the appellate court held that the district court erred in finding that app distribution and IAP payments are not separate markets.¹⁵⁹159.   Id. at 994, 996. However, the court invoked the D.C. Circuit’s holding that per se illegality for ties is inappropriate for ties involving software, platforms, and third-party apps.¹⁶⁰160.   Id. at 997. Without extensive analysis, the court simply concluded that “[a]pplying the Rule of Reason to the tie involved here, it is clearly lawful.”¹⁶¹161.   Id. at 998. Relatedly, the court explained that the tying arguments just rehash the prior Section 1 claims. Finally, for the Section 2 monopolization claim, the appellate court again affirmed the district court.¹⁶²162.   Id. at 999.

In summary, other than noting several harmless errors, the appellate court upheld the district court’s decision.¹⁶³163.   Id. at 1004. In doing so, the appellate court did not require immense levels of evidence of the various alleged harms and benefits on the platform given the nature of the disputed conduct.

The Fallout from the Epic Decision

In the wake of both the district court and Ninth Circuit rulings, various stakeholders filed a flurry of amicus briefs on behalf of both sides criticizing the rulings. After the district court decision, those in support of Epic’s claims argued that the court did not give sufficient weight to the anticompetitive evidence.¹⁶⁴164.   See, e.g., Brief of the American Antitrust Institute as Amicus Curiae in Support of Plaintiff, Counter-Defendant–Appellant at 3, Epic Games, Inc v. Apple, Inc., 67 F.4th 946 (9th Cir. Jan. 27, 2022) (No. 21-16506) (“The district court failed to appreciate the significance of its finding of direct anticompetitive effects.”); Brief of Amici Curiae the Consumer Federation of America and Developers in Support of Epic Games, Inc.’s Brief on Appeal at 1, Epic Games, Inc v. Apple, Inc., 67 F.4th 946 (9th Cir. Jan. 27, 2022) (No. 21-16506) (“The district court was wrong. If it had conducted the requisite balancing under the rule of reason, it would have found that the massive harm caused by Apple’s policies dwarfs the purported justifications that Apple put forth.”); Brief of Amici Curiae 38 Law, Economics, and Business Professors in Support of Appellant/Cross-Appellee at 2, Epic Games, Inc v. Apple, Inc., 67 F.4th 946 (9th Cir. Jan. 27, 2022) (No. 21-16506) (“The court below found that Epic satisfied this burden by showing that Apple’s restrictions on competition from other means of app distribution had substantial anticompetitive effects. That is no small accomplishment.”). This position is perhaps not surprising, as there is a prevailing presumption by some that, once harm is found, the defense faces an “uphill battle” to demonstrate offsetting procompetitive effects.¹⁶⁵165.   See, e.g., Evans & Schmalensee, supra note 76, at 27 (“(C)ourts seldom give much weight to pro-competitive benefits in the second stage.”). Further, given that the second step in Epic involves the welfare of users instead of developers (at least holding aside potential cross-group effects), even the dissent in Amex recognized the challenge that defendants face in this scenario. See Amex, 585 U.S. at 574 (Breyer, J., dissenting) (“American Express might wish to argue that the nondiscrimination (anti-steering) provisions, while anticompetitive in respect to merchant-related services, nonetheless have an adequate offsetting procompetitive benefit in respect to its shopper-related services . . . American Express might face an uphill battle. A Sherman Act § 1 defendant can rarely, if ever, show that a pro-competitive benefit in the market for one product offsets an anticompetitive harm in the market for another.”). Yet, the district court gave equal weight to both the anticompetitive and procompetitive effects. In this context, equal weight does not mean any piece of procompetitive evidence offsets any piece of anticompetitive evidence no matter the relative quality. Instead, it means that procompetitive evidence does not have, for instance, half the weight of the anticompetitive evidence—all else equal. Thus, while the district court did not engage in explicit balancing, as the Ninth Circuit noted,¹⁶⁶166.   Epic Games, 67 F.4th at 993. it is not apparent—as the critics would contend—that the court did not engage in implicit balancing. In other words, in the absence of quantitative evidence specifying the magnitude of the effects, the court simply considered the procompetitive and anticompetitive effects to have similar qualitative weight rather than engaging in a superficial balancing exercise with evidence inadequate for such an exercise.

On the other hand, critics in support of Apple questioned the district court’s finding that Epic met its prima facie anticompetitive-effects burden in the first place.¹⁶⁷167.   See, e.g., Brief of Amici Curiae International Center for Law & Economics and Scholars of Law and Economics in Support of Appellee/Cross-Appellant at 2, Epic Games, Inc v. Apple, Inc., 67 F.4th 946 (9th Cir. Mar. 31, 2022) (No. 21-16506) (“Epic’s case fails at step one of the rule of reason analysis. Indeed, Epic did not demonstrate that Apple’s app distribution and IAP practices caused the significant market-wide effects that the Supreme Court in . . . Amex . . . deemed necessary to show anticompetitive harm in cases involving two-sided transaction markets.”). While these amicus briefs contend that the court’s ruling was inconsistent with Amex,¹⁶⁸168.   Id. at 6–7 (“(B)y failing to discuss the net effects on consumers, neither Epic nor the district court properly analyze the alleged anticompetitive effects of Apple’s conduct in the manner prescribed by Amex.”). it is important to separate the underlying legal framework from the specific evidence brought to bear in the Epic case. This Article takes the position that the district and appellate court opinions were procedurally consistent with the Amex precedent because they considered evidence of effects on both platform groups due to material cross-group network effects between the platform participants. But this Article leaves open the question of whether the evidence was sufficient to demonstrate harm to the competitive process. One can hold both the view that the Epic court properly considered evidence of output, price, and innovation, which, when suitably formulated, can address platform-wide impacts, and the view that the court came to the wrong conclusion in terms of the adequacy of specific evidence presented. The district court clearly acknowledged the need to consider platform-wide effects when cross-group network effects are material.

The Ninth Circuit’s opinion drew a similar flurry of filings on behalf of Apple when the parties petitioned the Supreme Court for review. Amici supporting Apple focused on the scope of injunctive relief for the anti-steering provision rather than liability issues.¹⁶⁹169.   See, e.g., Brief of Washington Legal Foundation and TechFreedom as Amici Curiae Supporting Petitioner, Apple Inc. v. Epic Games, Inc., No. 23-344 (U.S. Nov. 1, 2023); Brief of International Center for Law & Economics as Amicus Curiae in Support of Petitioner, Apple Inc. v. Epic Games, Inc., No. 23-344 (U.S. Oct. 27, 2023). Epic’s arguments at this stage took on a similar, but slightly different flavor; Epic focused on the purported fourth balancing step of the rule of reason and the courts’ LRAs analysis.¹⁷⁰170.   Petition for a Writ of Certiorari, Epic Games, Inc. v. Apple Inc., No. 23-337 (U.S. Sept. 27, 2023). Regardless, the Court denied certiorari, ending this specific legal saga.¹⁷¹171.   Epic Games, Inc. v. Apple, Inc., 67 F.4th 946 (9th Cir. 2023), cert. denied, No. 23-337, 2024 WL 156473 (US Jan. 16, 2024). Yet, the denial did not end the overarching conversation about the proper antitrust assessment of app platforms at the heart of this Article.

Harmonizing the Amex and Epic Decisions & the Path Forward

Overall, both the Amex and Epic decisions continue to stoke the flames regarding the proper assessment of alleged anticompetitive conduct involving platforms with material network effects. This Part offers a distilled assessment of what Amex did and did not establish before offering a few final observations regarding Epic’s fidelity to this interpretation of Amex.

The Key Takeaways from Amex

Amex boils down to a central legal point that decreases in the welfare of one group of consumers (e.g., merchants, developers) on a transactional platform is not enough to establish anticompetitive harm when that same policy creates a potential benefit to another group of consumers (e.g., cardholders, users) on the other side of the transaction. One could also add that, in certain situations, initial benefits to one group (e.g., merchants who successfully steer users at the point of sale to another credit card that allows the merchant to keep more of the sales revenue) could be illusory if cross-group effects result in the network unraveling to the degree that merchants ultimately face a lower level of overall demand. Similarly, consider a platform policy change that reduces the amount of consumer data available to online merchants—including those who compete with the platform in some manner. Even if merchants experience an initial reduction in welfare from having less data, the benefit to users remains highly relevant to the assessment of whether the privacy policy is impairing the competitive process. Merchants may eventually benefit from the policy as more users complete purchases on the platform due to greater data protection.

Notably, the above approach to platforms differs from a “standard” Sherman Act case that weighs potential harms to consumers (e.g., a resale price maintenance policy which prevents merchants from offering discounts off the list price)¹⁷²172.   See generally Kenneth G. Elzinga & David E. Mills, The Economics of Resale Price Maintenance, in 3 Issues in Competition Law and Policy 1841 (ABA Section of Antitrust Law 2008). against the potential benefits from efficiency gains to those same set of consumers (e.g., greater overall levels of retail service in the market). This distinction between a standard case and a transactional platform case is critical because, under current antitrust law, when one group receives the harm while another group enjoys the benefits, there is the specter that the benefits will be challenged as impermissible on legal grounds because they are out-of-market efficiencies.¹⁷³173.   See, e.g., Kirkwood, supra note 77, at 1823 (“The Court’s famous rejection of the claim that ‘anticompetitive effects in one market could be justified by procompetitive consequences in another’ accords with this analysis. It suggests that if the two sides of a platform are separate antitrust markets, as they should normally be, a platform cannot justify a restriction on one side by benefits it provides to customers on the other. It cannot use anticompetitive conduct to extract wealth from one customer group and then justify that conduct by funneling the proceeds to another customer group.”). Setting aside the scope and wisdom of the doctrine, the question becomes, given the presence of cross-group network effects and concerns with counting out-of-market benefits, how should judges assess the impact of platform policies? Even critics of the Amex decision acknowledge the importance of cross-group effects and the need to assess the welfare of various groups on the platform “somehow.”¹⁷⁴174.   See, e.g., Michael Katz & Jonathan Sallet, Multisided Platforms and Antitrust Enforcement, 127 Yale L.J. 2142 (2018); Hovenkamp, supra note 77, at 49 (“Without relying on an economically incoherent conception of a relevant market, the Court could simply have said that when power is sought to be proven by direct effects all relevant effects should be considered.”) The question is how to accomplish this assessment without defining an integrated relevant market, which is the Amex solution to this problem.

The main alternative is to define two separate relevant markets. Yet, the division among the critics of Amex highlights the potential pitfalls of defining two separate relevant markets for multisided platforms. These critics diverge on the question of whether procompetitive effects in one group can be compared with anticompetitive effects in another group. Some take the position that out-of-market efficiencies cannot offset a finding of anticompetitive effects in a different relevant market.¹⁷⁵175.   See, e.g., Brief of 28 Professors, supra note 64, at 22 (“(T)his Court held, in a case where the defendant operated a two-sided platform, that each side represented a ‘separate . . . market’ and that injuring competition in the restrained market alone was sufficient to violate the Sherman Act.”). Others support defining separate markets but recognize the inherent link between two sides of a multisided platform.¹⁷⁶176.   See, e.g., Katz & Sallet, supra note 175; Hovenkamp, supra note 77. Such an approach departs from the standard, and intuitive, practice of defining a market and assessing the competitive effects within that market.

These proposed alternatives have several problems. The first is that bifurcating merchants and cardholders into separate relevant markets for the purpose of litigation is artificial and at odds with the fundamental reality that transactional platforms profit maximize and set optimal prices, quantity, and polices by jointly considering both groups.¹⁷⁷177.   See, e.g., Rochet & Tirole, supra note 54. To exemplify this point, both relevant markets in a transactional platform share a common output. The second problem is the lack of clarity regarding the scope and applicability of the out-of-markets efficiencies doctrine. To settle this dispute, the Supreme Court would need to revisit, clarify, or perhaps overturn the doctrine. These observations are not to suggest that antitrust cannot change or adapt to different circumstances. After all, as long as the welfare of all the impacted platform participants is considered and given equal weight, what does it really matter how the relevant market is defined? However, the Court in Amex solved these issues by integrating the analysis into one relevant market. This undoubtedly makes the analysis self-contained and highlights the importance of considering the impact of a platform’s conduct on the platform participants as a whole.

Reconciling Amex and Epic

Given this context and controversy surrounding Amex, how do the Epic decisions fit in? This Article asserts that Epic demonstrated that operationalizing Amex’s integrated approach to transactional platforms does not require fundamental departures from the familiar rule of reason scaffolding or place extraordinary burdens on plaintiffs to demonstrate anticompetitive effects. Specifically, the Epic courts found anticompetitive harm based on two observations: (1) Apple rarely discounted from its long-standing 30% commission and (2) the “but for” world would have had more app stores and IAP systems.¹⁷⁸178.   Specifically, working under Amex, the Ninth Circuit found reduced consumer choice plus market power was sufficient indirect evidence to meet Epic’s prima facie burden of anticompetitive effects, and this indirect evidence theory has some support in sister circuit cases. See N. Am. Soccer League, 883 F.3d at 42; MacDermid Printing Sols. LLC v. Cortron Corp., 833 F.3d 172, 183 (2d Cir. 2016). Cf. FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 459 (1986) (“(A)n agreement limiting consumer choice by impeding the ‘ordinary give and take of the market place’ cannot be sustained under the Rule of Reason.”). While there may be questions regarding the persuasiveness of this evidence, this criticism is not the same as saying the courts should have required a heavier burden on plaintiffs.¹⁷⁹179.   On this point, the quality of Epic’s prima facie evidence is indeed questionable. Third-party app stores have never existed on the iOS and is not a norm in the market. More broadly, choice reduction is arguably inherent in the entire idea of a platform—even an “open” one. See Hanno F. Kaiser, Are “Closed Systems” an Antitrust Problem?, 7 Comp. Pol’y Int’l 91, 97 (2011) (“Every platform needs rules, if only to define the boundary between the system and its environment.”). For instance, Apple’s refusal to license its iOS to independent OEMs reduces choice. Apple’s decision to preload iOS with Apple Maps instead of Google Maps reduces choice. Every default setting reduces choice. See Richard H. Thaler et al., Choice Architecture, in The Behavioral Foundations of Public Policy 428, 430 (2014) (“Defaults are ubiquitous and powerful. They are also unavoidable in the sense that for any node of a choice architecture system, there must be an associated rule that determines what happens to the decision maker if she does nothing.”). Further, Apple screens out low-quality sellers (who a segment of consumers might want on the platform if they are willing to trade-off quality for price) or sellers of products that offend the sensibilities of a significant portion of the user base—even though other users may have a strong preference for their inclusion (e.g., suppliers of indecent images). Consequently, reducing consumer choice is a poor proxy for anticompetitive harm. For instance, suppose Apple had originally allowed third-party app stores on the iOS, but, as its market power grew, tightened the restrictions on these app stores to the point of eventual exclusion. Then, under certain scenarios, it would be highly plausible that Apple’s policy would have harmed the competitive process and the welfare of the platform participants. At the very least, there would be serious questions regarding whether the efficiency arguments were pretextual. This hypothetical scenario would not have placed a greater burden of production on plaintiffs. Additionally, the Epic courts were open to considering effects that impacted one group differently than another, which is another core element of the Amex legal framework: the welfare of various groups must be considered and, implicitly, groups’ gains or losses should not be weighted in a way that favors one group over another.

Returning to the point that the adequacy of a given set of evidence is a separate issue from the procedural precedent, perhaps a source of confusion is that, in Amex, the burden never shifted to the defendant because the Court decided that the plaintiff did not meet its burden of production in step one to establish anticompetitive harm.¹⁸⁰180.   Amex, 585 U.S. at 552 (“In sum, the plaintiffs have not satisfied the first step of the rule of reason. They have not carried their burden of proving that Amex’s antisteering provisions have anticompetitive effects.”). The Court did not procedurally eliminate step two, render it moot, or shift the burden of production to show procompetitive effects to the plaintiffs. In Amex, the primary evidence of harm was that the anti-steering provision prevented discounting at the point of sale to disincentivize the use of the Amex card. The Amex Court suggested output would have been a better metric to assess the impact on the competitive process and decided the plaintiffs did not meet their step one burden. Consequently, reading Amex as raising the burden on plaintiffs to get out of step one in multisided platform cases misrepresents the opinion.¹⁸¹181.   Of course, defendants themselves may wish to hold the position that Amex raised the burden on plaintiffs to prove harm in transaction platforms cases. For instance, in Epic, Apple raised the argument that Amex required plaintiffs to demonstrate harm to both groups on a platform to meet their burden. The Ninth Circuit properly clarified that this is not an accurate reading of the Amex precedent. See supra Section II.B.

In a way, Epic showed us what happens at step two under the Amex framework. In doing so, the Epic case dispelled the claim that Amex required plaintiffs to both prove anticompetitive harms and disprove procompetitive benefits. This is because, once the district court found that the plaintiff met its burden of production for anticompetitive effects, the burden did, in fact, shift to the defense.¹⁸²182.   See Epic Games, Inc. v. Apple Inc., 559 F. Supp. 3d 898, 1038 (N.D. Cal. 2021). Further, as mentioned, Amex did not require a reduction in output to meet the burden at step one,¹⁸³183.   Epic Games, 67 F.4th at 984. which reinforces the flexibility of step one and suggests that Amex may be less defendant-friendly than it appears at first glance.

Much has been written about the undue burden that the Amex decision has placed on plaintiffs to bring and win antitrust cases. Yet, those burdens remain regardless of the structure of the court’s inquiry.¹⁸⁴184.   Relevant to this debate, evidence indicates that most antitrust cases before Amex were dismissed because plaintiffs failed to prove anticompetitive effects. See Michael A. Carrier, The Real Rule of Reason: Bridging the Disconnect, 1999 BYU L. Rev. 1265, 1360 (1999) (“In the modern era, courts dismissed 84% of Rule of Reason cases because plaintiffs failed to demonstrate a significant anticompetitive effect.”); Michael A. Carrier, The Rule of Reason: An Empirical Update for the 21st Century, 16 Geo. Mason L. Rev. 827, 828 (2009) (“(T)he burden-shifting trend has continued and, in fact, has increased. Courts dispose of 97% of cases at the first stage, on the grounds that there is no anticompetitive effect.”). Thus, the idea that Amex somehow represented a seismic shift is not consistent with the decision itself, the Epic decisions, or the prior empirical evidence which indicates that plaintiffs routinely fail to demonstrate anticompetitive harm. In litigation, the burdens of production are arguably similar for both plaintiffs and defendants. The reality for rule of reason cases is that both parties bring to bear evidence on all aspects of the case. Louis Kaplow explains that “[a]lthough production burdens can help in small stakes cases and to organize proceedings, in heavily contested battles this is likely to be rounding error.”¹⁸⁵185.   Louis Kaplow, Efficiencies in Merger Analysis, 83 Antitrust L.J. 557, 615 (2021). Similarly, Ronald J. Allen asserts that “[w]hich party bears what burdens of production is not important in a system with adequate discovery. In a system with discovery, each side has access to essentially all the relevant evidence and can produce it at trial, leading to a decision on the merits.”¹⁸⁶186.   Ronald J. Allen, Burdens of Proof, 13 L. Prob. & Risk 195, 202–03 (2014). Thus, Allen argues that “[t]here is, accordingly, no justification for complex rules allocating burdens of production in such a system, and typically the only complexity that one finds resides in the decision to list certain issues as defences rather than elements.”¹⁸⁷187.   Id. Allen further details that “(p)rior to the creation of robust discovery systems, allocations of burdens of production could significantly affect the outcome of cases, and complex sets of considerations were articulated to guide such allocations. . . . In modern American jurisdictions, these considerations are now largely an irrelevancy.” Id. at n.9. Relatedly, some argue that defendants should have the burden to show procompetitive effects because they are the low-cost providers of that information. While this is certainly true, defendants are the low-cost providers of almost all information relevant to a case.¹⁸⁸188.   Kaplow, supra note 186, at 615 (“Moreover, the underlying logic regarding who possesses information is incomplete. The merging parties, after all, have much internal information relevant to potential anticompetitive effects (such as data on past prices and quantities that are central to the government's analysis).”). Yet, the burden is still on the plaintiff to demonstrate anticompetitive harm and to define the relevant market. Both parties are fully incentivized to bring to bear evidence that informs the question of whether there are harms or benefits from a practice.¹⁸⁹189.   Id. (explaining that defendants “if they have a serious efficiency claim, would be inclined to offer substantial evidence thereof in any event.”). For instance, in merger analysis, while the burden is on the parties to bring an efficiencies defense, this evidence is presented early in the investigation process to agencies.¹⁹⁰190.   Id. (“Moreover, in serious challenges merging parties commonly—and voluntarily, before any analysis much less demonstration of anticompetitive effects—offer significant information on the legitimate business justifications for the merger, hoping to convince the agency not to delay, impose costs on, or disrupt their deal.”). Thus, plaintiffs are generally well informed of the arguments that defendants will bring. Ultimately, both parties must develop evidence on all aspects of a case, including the market definition, competitive effects, entry analysis, and procompetitive justifications. Antitrust law does not demand that burdens are equivalent or symmetrical, and various other dimensions impact burdens, including the use of presumptions.¹⁹¹191.   Consider, for instance, predatory pricing claims. The plaintiff must show both below cost pricing and a dangerous probability of recoupment. The defense has much better information on both these issues. The same could be said for the plaintiff’s burden to demonstrate LRAs. Likewise, presumptions play a large role in antitrust and can materially impact the burden placed on either of the parties. Andrew I. Gavil, Burden of Proof in U.S. Antitrust Law, in 1 Issues in Competition Law and Policy 125 (ABA Section of Antitrust Law 2008) at 128 (“A final traditional procedural element involves the use of presumptions. Presumptions can be irrebuttable, as is true with per se rules, or rebuttable, as with the Philadelphia National Bank ‘presumption.’ They can also be of various degrees of strength, even when rebuttable. Establishing presumptions is critical to the process of allocating burdens of production and, perhaps most importantly, to the process of shifting burdens from one party to another.”). Rather, the key contribution of Amex and Epic is the fair application of existing burdens of production to a specific type of business—transactional multisided platforms.

Finally, the out-of-market efficiencies doctrine does not complicate Epic because, in fidelity to Amex, the Epic decisions defined the relevant market across the entire platform, rather than separately for each group. In contrast, if we define a separate market for each group on the platform, then a strict reading of the out-of-market efficiencies doctrine would prohibit cross-market welfare comparisons.¹⁹²192.   See, e.g., Brief of 28 Professors, supra note 64, at 11 (“As an example, this Court in NCAA v. Board (of) Regents of University of Oklahoma . . . did not permit the NCAA to defend a restriction on televising college football games on the theory that it would ‘protect live attendance.’ That justification rested on the view that exercising market power and restricting output (i.e., limiting broadcasts) would lead to benefits elsewhere in the economy, and so was ‘inconsistent with the basic policy of the Sherman Act.’”); Katz & Sallet, supra note 175, at 2145–46, 2162 (proposing that, while a court should give “careful consideration to any significant linkages between the markets on the different sides of a platform,” ultimately, citing PNB, “harm to a group of users on one side of a platform due to anticompetitive conduct cannot be offset by gains to a user group on another side that are a consequence of that conduct”). Such complications may explain why the Amex court avoided the separate markets approach and instead used the integrated approach.¹⁹³193.   See, e.g., Yun, supra note 46, at 1291 (“Arguably anticipating this potential for antitrust gerrymandering in platform markets, Justice Thomas in Amex preempted the possibility by including both sides of a transactional platform—that is, credit card merchants and cardholders—into one unified “relevant market” to consider the alleged anticompetitive conduct.”).

Conclusion

App stores are increasingly part of the fabric of online commerce. As more transactions flow through this distribution channel, the stakes grow higher for antitrust decisions that impact app store governance. As antitrust claims and regulatory proposals continue to proliferate, developing a legal and economic method to assess the various types of conduct on these platforms becomes critical. This Article advances the view that Epic and Amex provide a clear template for decisionmakers. Those decisions focus on two core principles: (1) give equal weight to the positive and negative effects of a policy or conduct and (2) include the effects irrespective of which group on the transactional platform is impacted.

†Associate Professor of Law, Antonin Scalia Law School, George Mason University. Previously served as an Acting Deputy Assistant Director in the Bureau of Economics, Antitrust Division, at the U.S. Federal Trade Commission. I thank Thyme Hawkins, Kimberly Brooking, and Jaehee (Amy) Jung for excellent research assistance. I am also grateful to the editors of the Yale Journal on Regulation Bulletin for their careful and considered work on this Article’s manuscript.