Large platforms are often accused of refusing to serve (or discriminating against) competing sellers in adjacent product markets. Antitrust law labels such activity a unilateral “refusal to deal” (“RTD”) and evaluates it under a predation-like framework shaped by the two leading RTD cases, Aspen and Trinko. However, this framework is largely unhelpful in evaluating platform RTDs, because it does not ask the right economic questions. The relevant economic risk raised by platform RTDs is typically that the defendant might be leveraging its control of a dominant platform to foreclose rivals in an adjacent market. Therefore, as an economic matter, they tend to have much more in common with tying cases like Microsoft than they do with historical RTD cases like Aspen or Trinko. This suggests that courts ought to revise the legal standard used to evaluate platform RTDs to better capture economic realities. This would allow for meaningful antitrust oversight of platform conduct without opening the door to the freeriding or administrative concerns often associated with RTD doctrine. It would also likely end the Congressional push for an ill-conceived “self-preferencing” bill that might well do more harm than good.
By Erik Hovenkamp[1]
I. INTRODUCTION
Can a powerful firm really violate the antitrust laws by merely refusing to do business with a competitor? Officially, the answer has been yes for more than a century.[2] However, this “refusal-to-deal” (“RTD”) doctrine is highly controversial, and in recent decades the courts have narrowed it almost out of existence.[3] But the current debate on antitrust and digital platforms gives us reason to revisit the matter. This is not simply because platform conduct often raises RTD questions, but because it raises different economic issues than most historical cases in this area, and this is helpful in exposing the holes in current law.
Large digital platforms are often vertically integrated into adjacent product markets whose sellers rely on the platform to reach consumers. Such platforms are sometimes accused of refusing to serve (or discriminating against) rivals in adjacent markets. For example, Apple has been accused of removing competitors’ apps from its App Store.[4] And the FTC recently accused Facebook of analogous behavior.[5] And many platforms are accused of discriminating against rivals in ways that may impair their ability to make sales over the platform.[6]
Courts evaluate such conduct under the RTD rubric. The most significant RTD precedent is the Supreme Court’s 2004 Trinko decision.[7] Although the decision declined to overturn the Court’s prior decision in Aspen, which upheld liability for an RTD,[8] its overall effect was to narrow RTD doctrine substantially. In justifying this result, the Court emphasized that a widespread “duty to deal” with rivals would create freeriding problems and administrative challenges.
Trinko and its progeny were correct to emphasize the dangers of an overbroad “duty to deal.” And many of the complaints that have historically arisen under this doctrine were indeed poor candidates for antitrust intervention. Nevertheless, the prevailing Aspen/Trinko framework is largely unhelpful in evaluating platform RTDs. It developed from historical cases that shed little light on the economic concerns raised by platform RTDs. And, for related reasons, antitrust scrutiny in these cases generally does not create the same freeriding and administrative risks emphasized in Trinko.
Instead, the famous Microsoft case is a much more helpful model for courts to follow when evaluating platform RTDs.[9] Microsoft was a vertically integrated firm in a network industry. It controlled both a dominant platform (the Windows operating system) and a popular product in an adjacent market (its Internet Explorer web browser). Microsoft was convicted of an antitrust violation for leveraging its control of the platform to foreclose Netscape and other competitors in the adjacent market.
Of course, Microsoft involved a tying arrangement, not an RTD.[10] However, some RTD cases raise essentially the same economic concerns as tying: namely, that a defendant may be exploiting a dominant position in one market to impair competition in another. Indeed, this aligns with most RTD claims involving dominant platforms.
I have recently argued that courts should distinguish RTD cases that raise tying-like concerns of leverage and foreclosure by evaluating them under a revised standard of liability that better captures their economic characteristics.[11] In a nutshell, this would treat certain RTDs (including most of those involving platforms) as similar to tying. By contrast, all other RTDs would continue to be evaluated under a conservative standard verging on per se legality. This two-track approach would allow for meaningful antitrust scrutiny in meritorious cases without creating freeriding or administrability problems.[12]
Given that Section 2 of the Sherman Act is usually viewed as less formalistic than other antitrust statutes, one might think that courts would already allow plaintiffs to challenge some RTDs as de–facto vertical restraints. Indeed, courts have taken an analogous approach when evaluating other types of unilateral conduct.[13] But, to date, courts have consistently rejected such efforts in RTD cases, instead adhering to the Aspen/Trinko framework.[14] (The FTC’s case against Facebook was a prominent recent example of this.[15]) As detailed below, this creates a huge mismatch between the questions courts ask to determine liability and the relevant economic concerns raised by the defendant’s conduct.
As this reflects, current law’s treatment of RTDs is highly formalistic and mechanical. This prevents it from seriously considering the possibility that some RTDs might raise very different economic questions than historical cases like Aspen or Trinko. What explains this obstinance? In the decades preceding Trinko, RTD doctrine acquired a (largely deserved) reputation for being unprincipled and potentially harmful. Courts worry that any expansion of the doctrine will necessarily reintroduce the freeriding and administrability problems that Trinko sought to eliminate. This fear is unwarranted, however. The approach advocated here would expand antitrust scrutiny only in cases that raise essentially the same concerns as tying and related vertical restraints — practices that already receive meaningful scrutiny under Section 2 (at least compared to RTDs[16]).
Questions about antitrust reform and platform RTDs are especially timely given that Congress is showing interest in the subject. A bipartisan coalition of lawmakers recently proposed a bill that would target platform RTDs, including relatively mild forms of “self-preferencing.”[17] This project has been lauded by the popular press and progressive political commentators. However, for reasons I have discussed elsewhere, the bill is poorly designed, and could cause significant collateral damage to competition and consumers.[18]
In fact, there is no need for complex new legislation to address platform RTDs. The economic and legal foundations needed to evaluate them already exist, at least in broad outline. They predominate in cases like Microsoft. An effective policy toward platform RTDs merely requires courts to start building upon those familiar foundations.
The balance of this article begins by discussing why RTD law became so controversial and how the law and has evolved in response. I then discuss how RTDs sometimes emulate the economic effects of tying arrangements (particularly in platform cases); how current law fails to evaluate them in a sensible way; and how courts could do better. The final section offers some thoughts on how courts might be persuaded to adopt this approach before concluding.
II. THE TROUBLED HISTORY OF RTDS
There are good reasons why RTD law became controversial. For a long time, it was highly amorphous and open-ended, lacking concrete rules for courts to apply. Absent sensible limitations on liability, RTD doctrine can be abused as a ticket to freeride. The result is that private plaintiffs sometimes claimed to be the victims of exclusionary conduct merely because a larger competitor declined to grant them unfettered access to its core technology or other resources.[19] These plaintiffs were usually small firms that were unable to develop a viable competing product on their own. Fortunately, the most egregious freeriding efforts were rarely successful. Nevertheless, jurists and scholars were justifiably frustrated by RTD doctrine’s lack of clear, sensible limiting principles, which would have made such cases easier to purge.[20]
Such freeriding efforts could chill investment in valuable technologies, as a duty to share new technology with one’s rivals may substantially reduce the private value of developing it. This is the most important policy concern surrounding RTD doctrine.[21] The other significant concern surrounds administrability. Courts may not be equipped to determine the price or other terms of compulsory dealing — a task normally assigned to technocratic regulators, not generalist judges.[22]
These policy concerns boiled over in Trinko, which sharply narrowed RTD doctrine, as noted above. Following Trinko, lower courts have defined the scope of liability narrowly around the specific facts of Aspen.[23] Under this regime, RTDs are viewed as a form of predation, analogous to predatory pricing. Consequently, in most circuits the primary focus is on whether the defendant’s RTD caused it to sacrifice short-run profits, and, if so, whether this indicates an anticompetitive purpose.
Specifically, most circuits now require a showing that: (1) the defendant and its rival had a history of prior voluntary dealing before the defendant refused to continue such dealings; (2) this refusal caused the defendant to sacrifice profits in the short run; and (3) the only conceivable rationale for this sacrifice was to reap monopoly profits in the long run by excluding the rival.[24]
There are a few things worth noting about this framework. First, the standard is no longer vague and open-ended; it is now quite specific and narrow. Second, it is extremely hard to satisfy; indeed, no plaintiff has won a final judgment since Trinko.[25] Third, the doctrine focuses largely on the defendant’s intent or purpose, with comparatively little focus on actual competitive effects.[26] This is in contrast to how most types of conduct are evaluated in antitrust. Typically, antitrust analysis focuses primarily on competitive effects, and there are no special tests required to establish anticompetitive intent.[27] Finally, note that this standard bears no resemblance to Section 2 tying analysis. It does not contemplate the existence of separate products or an effort to exploit power in one market to foreclose rivals in another.
III. RTDS THAT RESEMBLE TYING
The following two principles are fundamental in antitrust. First, a firm does not violate the antitrust laws by earning a monopoly through competition on the merits. Second, even if a monopoly is earned on the merits, it is unlawful to exploit that monopoly to foreclose competition in a separate market. The latter is the basic premise of tying claims.
The problematic freeriding cases discussed in the previous section arose because RTD doctrine historically did a poor job of upholding the first principle mentioned above. It left open the possibility that a firm might be required essentially to give away a monopoly it earned on the merits through compulsory sharing of its core technology with rivals.
Contrast this situation with RTDs that resemble tying in terms of their economic effects. In these cases, the defendant has a monopoly over some “primary” product (e.g. a platform), but the plaintiff-rival does not compete with that product. Rather, the rival operates in some adjacent “secondary” product market (e.g. a good sold over the platform). The rival does not seek the defendant’s help to develop a viable product — it has already done so on its own. However, sellers in the secondary market rely on the primary product to some extent to reach consumers.[28] The antitrust claim alleges that the defendant is exploiting its control of the primary product to distort competition in the secondary market (e.g. by excluding them from the platform). The desired relief is an injunction requiring the defendant to let rivals access the primary product on the same terms as other customers.[29]
In such a case, antitrust scrutiny is manifestly consistent with the second principle mentioned above. It also upholds the first principle, because the desired remedy would not take away the defendant’s monopoly over the primary product; it would just prevent the defendant from using that monopoly to foreclose competitors in the secondary market. Finally, this relief would not be difficult to administer, as the court need not come up with the terms of trade on its own.
To underscore the close connection between these RTD cases and tying arrangements, consider a hypo involving a slight change in the facts of Microsoft.[30] Suppose that Windows came with an app store, and that users had to obtain all third-party software programs through this store. In that case, Microsoft could have excluded competing browsers like Netscape by simply refusing to let them onto its app store. This is an RTD, but it generates the same competitive effects a literal tie of Windows and Internet Explorer.[31] And the desired remedy in this hypo (an order to let the browsers onto the app store) would act just like an injunction of that tie. Finally, there is no meaningful sense in which Netscape might be called a “freerider” in this hypo. It developed a popular competing browser entirely on its own.
As this demonstrates, an RTD may be largely equivalent to a tying arrangement in terms of its economic effects. However, the formalistic nature of RTD doctrine means that the law treats the two completely differently. This leads to absurd results.
Continuing with the hypo, the operative change in facts means that the case would not turn on whether Microsoft was using its OS monopoly to foreclose rival browsers—the main focus of the actual Microsoft case. Instead, the case would turn on the Aspen-inspired predation elements discussed in the previous section.
For example, if Microsoft had begun excluding rival browsers from the moment it introduced its app store, then the antitrust claim would necessarily fail. This would negate the “prior voluntary dealing” element of the Aspen/Trinko framework. It is hard to overstate how nonsensical this would be. It would suggest that foreclosing rivals in an adjacent market is automatically lawful so long as you initiate it at the earliest opportunity.
Even if the prior dealing element were satisfied, this would merely shift the court’s attention to the profit sacrifice element. This too likely would have resulted in the complaint failing, as there was not strong evidence that Microsoft had sacrificed profits.[32] That would be problematic, since antitrust economists have long recognized that profit sacrifice tests are generally unhelpful.[33] Fortunately, the question of whether Microsoft sacrificed profits did not play a major role in the actual Microsoft decision.
Ultimately, what matters is the actual competitive effect of the defendant’s conduct, not tenuous proxies for anticompetitive intent. Intent evidence can sometimes be useful when evidence on competitive effects is ambiguous; but it should not be the primary concern. This is consistent with how courts evaluate tying and other contractual restraints. They ought to take the same approach toward RTDs that raise substantially the same competition concerns as tying.
What would this approach look like in practice? The first step would be to verify that the defendant’s conduct does indeed raise the kind of intermarket leverage and foreclosure concerns that underly tying claims.[34] As in a typical Section 2 tying case, this requires showing that there exist separate product markets: a primary one in which the defendant is dominant and an adjacent secondary market in which it competes with the plaintiff-rival. Additionally, for administrative reasons, it may be appropriate to require that the defendant voluntarily offers the primary product to third party customers (i.e. that it is selectively refusing to give rivals the same access).[35]
This would establish that a tying-like analysis is a good fit, and it would screen out the sort of freeriding cases considered above. To win the case, the plaintiff would have to establish anticompetitive harm. This requires assessing the risk of appreciable foreclosure in the secondary market and considering any relevant defenses. I discuss these and other practical issues in a separate article.[36]
IV. FINAL THOUGHTS
The baggage surrounding RTD claims is weighty, and calling for the doctrine’s expansion is usually a good way to elicit eyerolling within antitrust circles. But the idea that some RTDs should be treated as similar to tying is not only sensible as an economic matter, but it is also consistent with how antitrust already treats other forms of unilateral conduct. For example, cases alleging an exclusionary product design are often treated essentially as tying cases.[37] And so-called “conditional refusals” to deal (with customers) are often treated as de-facto vertical restraints under Section 2.[38] The logic that unifies such examples is that if unilateral conduct behaves like a vertical restraint then it makes sense to treat it as such. That courts do not employ this approach in RTD cases is the exception, not the rule.
The most pressing reason to consider this approach is that it provides a natural way to address the widespread concerns surrounding platform RTDs. The public can see that existing antitrust law does not meaningfully constrain such conduct and wants to see something done about it. That is a perfectly reasonable reform project. But the ham-fisted solution proposed by Congress is not much more appetizing than the status quo.
It would be better for the courts to develop a workable framework for addressing platform RTDs. But that is easier said than done. It would require the courts to acknowledge that existing law needs revision, which they rarely do. However, the courts have always been adamant that it should be economic considerations and not legalistic formalities that shape the substantive content of antitrust law. And, for the reasons outlined above, this principle cuts in favor of revision.
However, I suspect that such reform will not occur if plaintiffs (particularly the antitrust agencies) continue to play within the existing rules — to argue for liability within the existing RTD framework (or to argue that an RTD is not actually an RTD, as the FTC recently attempted[39]). Instead, we need a plaintiff to ask the courts directly to revisit the Aspen/Trinko framework as it applies to platform RTDs, and to consider the possibility that decisions like Microsoft might have much more to offer us in these cases.
[1] Associate Professor, USC Gould School of Law.
[2] United States v. Colgate & Co., 250 U.S. 300, 307 (1919) (an RTD may be illegal when it serves a “purpose to create or maintain a monopoly”).
[3] In this article, I focus on RTD doctrine and not its sibling, the “essential facilities” doctrine. The two are very similar, and the Supreme Court has refused to acknowledge the latter as an independent doctrine.
[4] See, e.g. Sarah Perez, Apple CEO Tim Cook questioned over App Store’s removal of rival screen time apps in antitrust hearing, TechCrunch (July 29, 2020) (discussing allegations that Apple removed third-party screen-time monitoring apps shortly after introducing its own screen-time app).
[5] This portion of the FTC’s complaint was dismissed, however. Fed. Trade Comm’n v. Facebook, Inc., No. 1:20-cv-03590 (JEB), 2021 WL 2643627 (D.D.C. June 28, 2021).
[6] Committee on the Judiciary of the House of Representatives, Investigation of Competition in Digital Markets, Part I (117th Congress, 2022), https://www.govinfo.gov/content/pkg/CPRT-117HPRT47832/pdf/CPRT-117HPRT47832.pdf.
[7] Verizon Commc’ns Inc. v. Law Offs. of Curtis V. Trinko, LLP, 540 U.S. 398 (2004).
[8] Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985).
[9] United States v. Microsoft Corp., 253 F.3d 34, 58 (D.C. Cir. 2001).
[10] Microsoft used a combination of vertical agreements and software design choices to tie its Internet Explorer web browser to its Windows operating system. This was found to exclude competing browsers, such as Netscape. Id.
[11] Erik Hovenkamp, The Antitrust Duty to Deal in the Age of Big Tech, 131 Yale L.J. 1483 (2022), https://papers.ssrn.com/abstract=3889774 [hereinafter Hovenkamp, Duty to Deal]; Erik Hovenkamp, Platform Discrimination Against Rivals: An Economic Framework for Antitrust Enforcement, J. Corp. L. (forthcoming, 2023), https://papers.ssrn.com/abstract=4323207 [hereinafter Hovenkamp, Platform Discrimination].
[12] Hovenkamp, Duty to Deal, supra note 11.
[13] See notes 35-37, infra, and accompanying text.
[14] Some cases (e.g. Kodak) involve both an RTD and some tie-like conditional dealing with customers. In these cases, courts are usually willing to treat the conduct as a tie. But absent any such conditional dealing with customers, courts generally apply the standard RTD framework. See id. at 1519-1523.
[15] The FTC attempted to portray Facebook’s RTD as a kind of “conditional dealing” (an attempted analogy to the famous Lorain Journal case). But the court rejected this, instead applying Aspen/Trinko and dismissing the RTD claim. Fed. Trade Comm’n v. Facebook, Inc., No. 1:20-cv-03590 (JEB), 2021 WL 2643627 (D.D.C. June 28, 2021). Similar allegations in a revised pleading were again dismissed, although the FTC’s merger claim survived. Fed. Trade Comm’n v. Facebook, Inc., 581 F. Supp. 3d 34, 57-60 (D.D.C. 2022).
[16] All exclusion cases (including tying cases) are hard to win. But, as a practical matter, most circuits treat RTDs as essentially legal per se. The same cannot be said of Section 2 tying cases, as evidenced by decisions like Microsoft.
[17] S.2992 (The American Innovation and Choice Online Act), https://www.congress.gov/bill/117th-congress/senate-bill/2992/text.
[18] Erik Hovenkamp, Proposed Reforms in Big Tech: What Do They Imply for Competition and Innovation?, CPI Antitrust Chronicle (2022), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4127334.
[19] For example, in Morris, the defendant PGA Tour had just developed a technology for tracking its golf tournament scores in real time. The plaintiff, a small media company, sued the PGA for refusing to let it publish the scores (for free) before the PGA publicly released them. In effect, the plaintiff sought to be the primary beneficiary of the PGA’s own score-tracking technology. Fortunately, the plaintiff’s claim was rejected. Morris Commc’ns Corp. v. PGA Tour, Inc., 364 F.3d 1288, 1294 (11th Cir. 2004).
[20] See, e.g. Phillip Areeda, Essential Facilities: An Epithet in Need of Limiting Principles, 58 Antitrust L.J. 841 (1990).
[21] See, e.g. Verizon Commc’ns Inc. v. Law Offs. of Curtis V. Trinko, LLP, 540 U.S. 398, 407-08 (2004) (“compelling such firms to share the source of their advantage … may lessen the incentive … to invest in those economically beneficial facilities.”).
[22] Id. (“Enforced sharing also requires antitrust courts to act as central planners, identifying the proper price, quantity and other terms of dealing – a role for which they are ill suited.”)
[23] For a detailed discussion of post-Trinko RTD doctrine, see Hovenkamp, Duty to Deal, supra note 11, at 1498-1502.
[24] See, e.g. Novell, Inc. v. Microsoft Corp., 731 F.3d 1064, 1074-75 (10th Cir. 2013) (Gorsuch, J.); FTC v. Qualcomm Inc., 969 F.3d 974, 994-95 (9th Cir. 2020).
[25] A plaintiff did recently win a final judgment in district court, but it was reversed on appeal. See id. at 982.
[26] This is because an RTD is not deemed unlawful unless it was undertaken for an anticompetitive purpose, regardless of its actual effects on competition. See, e.g. Simon & Simon, PC v. Align Tech., Inc., No. 1:19-cv-00506-LPS, 2019 WL 5191068, at *5-6 (D. Del. Oct. 15, 2019) (RTD liability requires an “anticompetitive purpose” and this renders it “underinclusive” in the sense that “some refusals … evade antitrust scrutiny even when they actually have anticompetitive effects”) (citing Novell, Inc. v. Microsoft Corp., 731 F.3d 1064, 1076 (10th Cir. 2013)).
[27] An exception is predatory pricing, which modern RTD doctrine is modeled after.
[28] For example, they may use the primary product for distribution or marketing services.
[29] This assumes that the defendant voluntarily sells the primary product to noncompeting third parties. For administrative reasons, it may be appropriate to limit antitrust scrutiny to such cases. See note 35, infra, and accompanying text.
[30] This hypo is borrowed from Hovenkamp, Platform Discrimination, supra note 11.
[31] Both strategies make it difficult or impossible for consumers to use a rival browser on a Windows machine.
[32] See, e.g. Andrew I. Gavil, Exclusionary Distribution Strategies by Dominant Firms, 72 Antitrust L.J. 3, 56-7 (2004); Jonathan M. Jacobson & Scott A. Sher, No Economic Sense Makes No Sense for Exclusive Dealing, 73 Antitrust L.J. 779, 794-95 (2006).
[33] In particular, most types of exclusionary conduct do not necessarily require a profit sacrifice, nor does a profit sacrifice necessarily indicate anticompetitive animus. See, e.g. Thomas G. Krattenmaker & Steven C. Salop, Anticompetitive Exclusion: Raising Rivals’ Costs to Achieve Power over Price, 96 Yale L.J. 209, 224 (1986).
[34] I discuss this in detail in Hovenkamp, Duty to Deal, supra note 11, at 1539-44.
[35] This avoids the need for the court to stipulate the price or other terms of compulsory dealing (as the court can just copy the terms offered to other customers), and it helps to prevent difficulties or ambiguities in establishing the “separate markets” requirement. See id. at 1530-31, 1542-43.
[36] Hovenkamp, Platform Discrimination, supra note 11.
[37] Typically the defendant is accused of redesigning its primary product to make it incompatible with competing versions of a secondary good (a complement). See, e.g. C.R. Bard, Inc. v. M3 Sys., Inc., 157 F.3d 1340, 1367 (Fed. Cir. 1998).
[38] A famous example is Lorain Journal, where a monopolist refused to serve any customers who transacted with its lone competitor. The Supreme Court treated this as de-facto exclusive dealing. Lorain Journal Co. v. United States, 342 U.S. 143 (1951).
[39] See note 15, supra.
