The Supreme Court decision in Ohio v. American Express Co., has sparked more than its share of controversy. Indeed, the decision has been described as “what may be the worst antitrust decision in many decades.” The Court ruled that the product market for use of credit cards needed to be defined to include both cardholders and the merchants that accepted those cards, and that because the plaintiffs did not establish harm in that relevant market, they thereby failed to establish a prima facie case that would require the defendant to offer evidence justifying its conduct. In a manner not contemplated by past Supreme Court precedent, the Amex Court introduced a new measure of market definition, based on the presence of “transaction platforms.”
There are many reasons to believe that Amex is destined to be considered bad law. But, for the moment at least, Amex is the law and antitrust defendants are eager to expand its scope—a phenomenon we both saw first-hand at the Department of Justice when private firms, then in possession of the Secord Circuit opinion that the Supreme Court affirmed, rushed to the Antitrust Division to defend their behavior and discourage further investigation by arguing that any antitrust analysis would require a full effects assessment of all sides of any multi-sided business platform.
Our purpose is to examine the key arguments for applying Amex even where the relevant market participants are not transaction platforms, and the key court decisions that in the last four years have grappled with them. Our view is that any expansion of Amex would threaten basic principles upon which effective antitrust enforcement is built, including that (i) market definition begins with fact-specific inquiries into demand substitution and (ii) market definition is involved in only one process for identifying market power. As we will show, nothing in Amex supports the needless abandonment of these key principles in circumstances different than those considered by the Amex Court, even if one were to take the Amex decision as correct in the circumstances of that case.
Indeed, starting with demand-side analysis and taking each side separately at the outset limits the extent to which Amex could be inappropriately expanded to mask the improper exploitation of market power. Adopting this approach also reduces the risk that defendants would successfully “take all possible steps to shoehorn the facts into the ambit of ‘transaction platform,’ forcing courts and litigants into expensive and difficult analysis that is likely to cause more Type 2 errors (false negatives) in antitrust cases.”
This article briefly reviews what the Amex Court actually did decide. Next it offers thoughts on how, consistent with the Amex analysis, that ruling should be applied and explains why courts should reject application of Amex beyond its facts. A brief conclusion follows.
The Amex Court had a view of the facts that is difficult to square with the district-court’s factual findings and the traditional deference due to the fact-finder—or, indeed with Supreme Court precedent. But, even on its own terms, the decision describes a very limited set of conditions. Recall that the Department of Justice had brought a Section 1 case alleging, in essence, that anti-steering limitations on merchants who accept American Express cards harmed competition. The district court ruled for the government, finding that, for example, the Amex restrictions prevented merchants from providing consumers truthful information about alternative credit cards. Importantly, the government recognized that credit cards impact the interests of both cardholders and merchants that accept them; it alleged, and the district court accepted, that “by disrupting the price-setting mechanism ordinarily present in competitive markets, the [contractual terms] reduce American Express’s incentive—as well as those of Visa, MasterCard, and Discover—to offer merchants lower discount rates” and that the restrictions on the merchants harmed consumers “as inflated merchant discount rates are passed on to all customers—Amex cardholders and non-cardholders alike—in the form of higher retail prices.”
The Amex Court believed that the circumstances before it required a market to be defined even though the Department of Justice at trial had relied on direct evidence of harm (in other words, had not relied on market share to establish market power). The Amex Court asserted that because the alleged restraint was vertical, (i) market definition was required in a manner not required in cases of alleged horizontal restraints and (ii) that in the circumstances before it, market power “cannot be evaluated unless the Court first defines the relevant market.” Both parts of this assertion are dubious because, as Justice Breyer explained in dissent, “proof of actual adverse effects on competition is, a fortiori, proof of market power” and the district court had found such adverse effects. The difficulty is that the Court’s analysis can lead to the view that market definition is a necessary step in showing market power, which unnecessarily reverses the usual relationship between market power and market definition and, if used beyond the boundaries of the Amex opinion itself, would mask, rather than identify, market power.
Nonetheless, and while taking a very different view of the evidence than the district court, the Court moved on to analyze market definition in a manner moored in the Court’s understanding of the relevant competitive conditions. It is worth parsing the steps in the Court’s analysis:
Even if one does not agree with the Court’s conclusion that a single market is appropriate and required in this setting, these are a highly restrictive set of circumstances. Indeed, when the Supreme Court subsequently decided the antitrust implications of the operation of a multi-sided platform in Apple Inc. v. Pepper, it did not even cite Amex, consistent with Tim Wu’s suggestion that Amex is “narrower than some have suggested.”
In a case where a defendant asserts that the appropriate market is composed of transaction platforms, and a court reaches the point where product market definition is appropriate, that court should take a series of steps to carefully evaluate the choices available to the users on each side of the alleged multi-sided platform in order to understand the existence and nature of substitutes that are available to any user on any side. This is critical in order to limit the likelihood that market definition becomes a mechanism that obscures rather than helps to reveal market power, a point to which we return below.
First, a court must consider whether a relevant product market is composed solely of multi-sided businesses. That requires scrutiny of the competitive substitutes available for each distinct group of sellers or buyers for which the defendant is a trading partner. For sellers, market definition begins with demand-side analysis: Where do buyers look for alternatives to a given product? That makes sense because when we analyze a seller’s potential market power, market definition is used to identify the competitors and circumstances under which a firm has the ability to profitably increase its price by a small but significant amount. A similar, mirrored analysis applies when examining market power of a buyer or group of buyers: where do sellers or suppliers look for alternatives to a given buyer? In this case, market power enables buyers to suppress the price they pay to suppliers, creating harm in upstream markets. In other words, relevant substitutes, identified using the kind of analysis set forth in the 2010 Horizontal Merger Guidelines, must be assessed from the perspective of each set of users on each side of the multi-sided platform.
If for any distinct group, substitutes exist that demonstrate the existence of markets that are not solely composed of multi-sided businesses serving those distinct user groups, then, by definition, Amex is inapplicable. This is consistent with Amex’s recognition, noted above, that a “nontransaction platform” exists when “[a] newspaper that sells advertising, for example, might have to compete with a television network, even though the two do not meaningfully compete for viewers.”
The failure to understand the appropriate starting point for the Amex market-definition analysis is at the heart of the error in United States v. Sabre Corp., where the court concluded that, having found Sabre to be serving both travel agents and airlines, then as a matter of law, Farelogix, which supplied only airlines, could not be a competitor. The problem is that the Sabre court applied the reasoning backwards; first asking whether Sabre was a transaction platform and, having concluded that it was, then ruling that no non-transaction platform could be a competitor. But this is not what the Amex Court did; it first asked whether the market was composed solely of transaction platforms and only then concluded that the market definition should be limited to such business models. Indeed, it unequivocally rejected the approach the Sabre district court adopted, expressly noting that “[n]ontransaction platforms…often do compete with companies that do not operate on both sides of their platform.” Had it proceeded in the correct order, the Sabre court would have first asked whether Sabre and Farelogix were providing competing services to airlines by, for example, looking at whether introduction of Farelogix’s product resulted in lower negotiated Sabre fees from airlines. If the court concluded, as the government urged, that they were competitors, then, again by definition, the Amex analysis would be inapplicable.
Indeed, the Sabre conclusion would surely come as news to any consumer who has chosen between an advertising-supported streaming service, like those offered by Hulu or Pandora, and those that provide content on a subscription basis free from advertising, like Apple TV or Apple Music or Pandora ad-free subscriptions. As commentators have explained, competitive alternatives are not necessarily the same on different sides of even what might resemble a transaction platform, as in the example of Uber’s matching of drivers and riders where alternatives exist for both riders (e.g., Zipcar or public transportation) and drivers (e.g., working in other occupations).
Second, even if all of the businesses competing for the same defined set of distinct user groups are multi-sided, the key inquiry remains: Do they all qualify as transaction platforms? A transaction platform as defined by the Court requires the simultaneous presence of users on each side to effectuate a transaction that jointly consumes a single unit of output and displays the necessary indirect network effects. The requirement of joint consumption deserves emphasis; it exists “only if a merchant and cardholder both simultaneously choose to use the network.”
The firms under examination in a different case may not qualify. Some may be, for example, newspaper-like. In such circumstances, the business models would fail the requirement that they “cannot sell transaction services to either cardholders or [users on another side of the platform] individually.” Even in situations where consumption is joint and simultaneous on both sides, non-transaction platforms can compete with a “transaction” platform. Cash might compete with credit card transaction services. Direct bookings might compete with a hotel booking platform.
To be a market comprised solely of transaction platforms, that business model must not only be adopted by all of the competitors for the same set of users, and not only be a service that cannot be sold separately to one set of users individually, but that service must also require the participants from each side of the market to use the platform at the exact same time in order the effectuate the transfer of “one transaction’s worth” of output.
Thus, the court in In re Delta Dental Antitrust Litig. (N.D. Ill. 2020) held that health insurance markets are not two-sided markets under Amex because those markets “[lack] the ‘key feature’ of a transaction platform: simultaneity of the exchange.” As the court explained:
As common experience teaches, consumers of dental services typically pay insurers fixed premiums at regular intervals, regardless of when or even whether they visit the dentist. And the amount of the insured’s premium generally depends on the terms and coverage of her plan, not on the cost of the goods or services she receives on any particular visit. Yet as plaintiffs allege, insurers reimburse dental providers based on the goods and services they actually provide to patients. So a dental provider receives no payments at all on behalf of an insured who paid her premiums in full but did not actually receive dental care during the plan year. And reimbursements paid on behalf of an insured who does receive covered services during her plan year are untethered in both time and cost from the insured’s premium payment. In these ways, dental insurance operates decidedly differently from the “two-sided transaction platform” in Amex.
The fidelity of the Delta Dental court to the Amex reasoning is critical because, of course, many business models have some form of simultaneous conduct. Dentists generally provide dental services like teeth-cleaning at the precise moment that a patient receives the benefit of the service. Contracts becomes effective simultaneously for the contracting parties. For a newspaper advertisement to be effective, the reader and the newspaper have to agree at a point in time that the reader will purchase a subscription, the advertiser and the newspaper also have to reach an agreement at a point in time and, of course, a reader has to be looking at the page at the same time the advertisement appears for the advertiser to receive value from its purchase. But in none of those cases are the reader and the advertiser “simultaneously choosing to use” the newspaper to interact with one another. Moreover, for Amex to apply the defendant must facilitate the transfer of “one transaction’s worth” of services to both (assuming two) audiences to both audiences for joint consumption. But, as one commentator has concluded, “it is hardly true that a one-to-one correspondence exists between a viewer’s activity and the purchase of advertising” because, for example, “[i]n a market such as free television, advertising volume and rates might be based on Nielsen or other surveys that assess the size and composition of the audience.”
Third, and as a matter of law, the Amex decision was also quite specific to the circumstances of the case, in which the government alleged a violation of Section 1 of the Sherman Act. Nothing in the Court’s opinion reflects the distinct jurisprudence of Section 2 or, as in the Sabre case, Section 7 of the Clayton Act. In this vein, the court in FTC v. Surescripts, LLC, expressly distinguished the Section 1 claim in Amex from the Section 2 claim it was considering. Similarly, the court in In re NCAA distinguished the conduct before it from Amex because it involved a horizontal restraint, whereas Amex was a vertical restraint.
There is a natural tendency for antitrust defendants to seek a broad application of favorable precedent. In this section we review writings that seek to expand Amex beyond the factual and legal boundaries of the Court’s reasoning. In essence, they invoke characteristics of the business models of multi-sided platforms in order to place unwarranted burdens on antitrust enforcers, to replace facts with doctrine, to misdescribe the economic principles underlying the operation of multi-sided platforms and, in general, to shrug off the actual reasoning of Amex. We move now to three specific subjects where the broadening of Amex has been urged.
Pricing: First are suggestions that pricing by multi-sided platforms cannot or is unlikely to be harmful. Joshua Wright and John Yun assert that: “. . . the very definition of the exercise of monopoly power—the reduction of market-wide output and increase in the market price—cannot be satisfied by evidence of a price effect on only one side of a given platform.”
But remember that a non-transactions platform is, by definition, active in two or more distinct product markets. Consider the example employed by the Amex Court itself in which a newspaper and television station compete for advertisers while not competing for users. It is certainly possible to imagine circumstances in which the newspaper would face significant competition from other newspapers such that “an increase in subscription prices may lead to a fall in the number of subscribers, which would then adversely affect advertisers’ demand for ads and, thus, advertising revenues.” But suppose the newspaper faces little if any competition from other newspapers and holds market power in the user-side market for newspaper subscriptions. That would mean that it could increase prices to its readers above competitive levels—an outcome that would not be thwarted by the presence of the television station as a competitor for advertisers. Readers without a competitive choice would be less likely to flee in the face of small but significant price increases, thus allowing the newspaper to charge supra-competitive prices to its subscribers without harming its ability to attract advertisers. In other words, neither the characteristics of a multi-sided platform nor possible indirect network effects inherently eliminate adverse price effects for any group of a platform’s customers.
Further problems arise from the suggestion that courts look only at the total price charged across the multiple sides of a platform (sometimes referred to as the “net price,” for a multi-sided “market”), rather than looking to the individual prices charged to each group of customers for the platform (the “price structure”). As illustrated by the example in the previous paragraph, the economic literature on multi-sided platforms highlights the essentiality of the role played by the price structure (that is allocation of prices across different users of a platform), not merely the net price (or price level) for analysis. While that literature emphasizes the importance of mutual feedback loops (or externalities) from pricing on one side of a platform to others in defining a multi-sided platform, it makes clear one cannot collapse the analysis into a single net price independent of the price structure. Moreover, for many if not most platforms, there may be no economically meaningful way to define a “net price.” Even for a simultaneous transaction platform like credit cards, the two-part transaction fee charged to merchants (typically a fixed fee plus an amount proportional to the amount of the transaction) appears simple compared to the highly nonlinear and heterogeneous pricing structure for cardholders, which may have elements of annual fees (lump sums) that may or may not be waived, varying interest rates and fees that depend primarily on balances rather than transactions, cardholder points or cash back that may be earned linearly in transaction volume but valued nonlinearly (e.g., for airline affinity cards whose points earn frequent flyer rewards), and myriad other cardholder benefits (e.g., insurance, concierge services, travel fee reimbursements, access to travel perqs and upgrades, etc.) that vary with “tier” of the card and are either fixed or nonlinear in transaction volume. Any proposal for how to compute a “net price” per transaction from this complexity is likely inherently misguided and doomed to be wrong.
Output: Finally, some commentators have suggested that plaintiffs should have to demonstrate output reduction not only in the two-sided simultaneous transactions markets to which the Court cabined its Amex decision, but more generally. For example, Evan Chesler and David Korn argue that “[t]he Supreme Court made clear that, in the context of two-sided platforms, courts should carefully scrutinize evidence of increased prices to ensure that the price effects are linked to reduced output.” Judge Douglas Ginsburg and Koren Wong-Ervin suggest this means that “changes in market-wide output are the best indicator of competitive effects in two-sided markets.”
To begin, the Amex Court’s discussion of the role of output evidence for the determination of competitive harm was both problematic and misguided. The Court cites 30% growth in credit card transactions between 2008 and 2018 as evidence against the proposition that under the alleged conduct, “output was restricted or prices were above a competitive level.” As a matter of logic as well as economics, that conclusion stands as a non sequitur. As Justice Breyer explained:
the relevant restriction of output is as compared with a hypothetical world in which the restraint was not present and prices were lower. The fact that credit-card use in general has grown over the last decade, as the majority says, see ante, at 17–18, . . . says nothing about whether such use would have grown more or less without the nondiscrimination provisions. And because the relevant question is a comparison between reality and a hypothetical state of affairs, to require actual proof of reduced output is often to require the impossible—tantamount to saying that the Sherman Act does not apply at all.
Importantly, the Amex decision does not require that its one-sided analysis of output be adopted in any other case because the relevant language clearly represents a case-specific reading of the record.
Even were it possible reliably to measure counterfactual output levels, higher output is not necessarily indicative of greater welfare, and particularly not in the presence of contractual restrictions like the anti-steering terms at the heart of the Amex challenge. Indeed, those restrictions are intended to prevent consumers from responding to information about the cost of Amex credit card transactions by precluding the provision of that information by merchants. Michael Katz and Doug Melamed note that “in a two-sided market, changes in transaction volumes and changes in user welfare can diverge because the interests of the users on the two sides are not aligned, and a platform may be able to exploit this fact to increase its profits in ways that increase output but harm competition and the platform’s users.” As Steven Salop and his collaborators point out, “higher merchant fees caused by these parallel antisteering rules placed consumers into a prisoners’ dilemma game, which led inevitably to increased use of credit cards above the efficient, competitive level, making the volume of card transactions a poor proxy for welfare effects.”
Rule of Reason: Chesler contends that Amex created a new rule for application of the rule of reason under which “[a] shorthand review in the form of per se rule or ‘quick’ look . . . is inappropriate” for vertical agreements between a platform and its customers. But there is no basis for that view in the text of Amex, nor in Justice Breyer’s dissent (a conclusion that would likely have registered with him given his earlier writings on the “quick look” standard), nor even in the unclear dicta about the analysis of vertical restraints generally. The test of whether a vertical agreement is subject to per se or “quick look” treatment should not turn on whether the defendant is a “transaction” platform, it should turn, as it traditionally has, on whether the conduct at issue is so likely to be anti-competitive as to be conclusively illegal or whether, under the quick look test, “an observer with even a rudimentary understanding of economics could conclude that the arrangements in question would have an anticompetitive effect on customers and market.” The attempt to derive a general rule based on the status of the defendant rather than focusing on the specific conduct at issue is part and parcel of a dangerous, and unjustified, trend of assuming the irrelevance of facts pointing to competitive harm, which needlessly burdens antitrust enforcement.
So long as it remains in place, the decision in Ohio v. American Express must be recognized as precedent that, by its own terms, is very narrow and fact-specific. Neither the decision itself nor the arguments pressing for its expansion justify adding to the burdens on antitrust enforcement through novel applications of bad law, erroneous economics, or any combination thereof.
As we have emphasized, a prime risk of the over-application of Amex is that by too broadly combining distinct user groups, the role of market definition can be distorted to serve as a means of cloaking, not identifying, market power. That can happen where harm can be inflicted to the users on one side without harming users on another, such as where geographic markets differ or in the case of labor markets or more generally where companies compete on one “side” of a platform but not on the other. Moreover, “[p]utting production complements into the same market simply because making a deal requires both introduces economic nonsense into the law and economics of market power.” We argue on the side of clarity, not economic nonsense, in advancing the cause of effective antitrust enforcement.
