abstract. The antitrust duty to deal is perhaps the most confounding and controversial form of antitrust intervention. It is sought in situations where a monopolist controls a critical input (or “essential facility”) and unilaterally refuses to sell access to rivals. Courts have substantially narrowed the doctrine in recent decades. However, the rise of dominant platforms like Google, Facebook, and Amazon has provoked intense debate over whether the antitrust duty to deal needs a revival. Many such platforms are accused of refusing to deal with (or discriminating against) rivals in adjacent markets.

At present, all unilateral refusals are evaluated under a common standard, which is virtually impossible to satisfy. This paper identifies an important economic distinction between two types of refusal cases; it argues that they raise very different theories of harm and demand different standards of liability. In one line of cases, the defendant’s conduct raises essentially the same theory of harm as tying or related vertical restraints. However, formalistic doctrine prevents courts from evaluating them as such. As a result, these cases do not receive meaningful scrutiny. This is problematic, because a large majority of meritorious refusal-to-deal cases fall into this category, as do almost all cases involving dominant platforms.

In a separate line of cases, intervention is much harder to justify on economic policy grounds, as it risks chilling investment in valuable new technologies. Courts often acknowledge this investment concern in dicta, but the liability standard they apply—which focuses myopically on exclusion—ignores it. This has led to a major internal contradiction: courts are emphatic that a duty to deal is almost never warranted, but simple economic arguments show that the refusals in these cases are routinely exclusionary in precisely the sense that the law purports to condemn. This contradiction has spurred courts to erect suffocating evidentiary requirements, which now do most of the heavy lifting in practice.

These rules bear little logical connection to exclusion, but they excel at reining in liability. The problem is, they also kill off all the meritorious cases.

This Article argues that any effective reform must begin by disentangling these distinct lines of cases. Subjecting them to different liability standards would help to address many of the key concerns raised on both sides of the debate. The justifications for this approach are manifold. First, it protects investment incentives without needlessly stifling enforcement in meritorious cases. Second, it naturally limits antitrust scrutiny to cases in which intervention is most likely to be administrable. Third, it is exactly analogous to the way antitrust already treats other forms of unilateral conduct. Finally, this approach would allow for meaningful antitrust scrutiny of unilateral conduct by dominant platforms—an objective that has recently received bipartisan support in Congress—while remaining faithful to core antitrust principles.

author. Assistant Professor, USC Gould School of Law. I am grateful to Marina Lao; Christopher Leslie; Jonathan Masur; Doug Melamed; Mark Niefer; John Newman; Eric Posner; Steve Salop; and Howard Shelanski for providing helpful comments. I am also grateful to seminar participants at the U.S. Department of Justice, Antitrust Division; the University of Chicago Law School; UC Irvine School of Law; and Northwestern University Pritzker School of Law.

Introduction

As a general proposition, firms are free to choose the parties with whom they do business.1 Thus, the default rule is that a firm can lawfully refuse to deal with rivals, so long as this choice is unilateral.2 However, this freedom is subject to certain limitations under antitrust law.3 Under section 2 of the Sherman Act,4 a dominant firm’s refusal to deal with rivals becomes unlawful when it operates to “monopolize” the market through exclusion—that is, by impairing rivals’ ability to compete effectively.5 Conventionally, this involves a monopolist who controls an important technology, often described as a “bottleneck input” or an “essential facility,”6 and refuses to let rivals buy access to it, or else grants them access only on discriminatory terms. In such cases, the desired relief is compulsory dealing: an antitrust duty to deal.

This area is one of the most confounding and controversial subjects in antitrust law.7 Proponents argue that a duty to deal is necessary to prevent dominant firms from exploiting their control over critical inputs to eliminate competition. Opponents argue that it threatens incentives for investment in valuable new technologies, as those technologies are much less lucrative when they must be shared with rivals.8 Opponents also question courts’ ability to administer a duty to deal, as its implementation may require more supervision or expertise than courts can provide.9

In recent decades, the opponents’ arguments have mostly won out in the courts. In its 2004 Trinko decision, the Supreme Court weakened the doctrine substantially, although it stopped short of eliminating it altogether.10 The Court was cryptic in characterizing the underlying standard of liability. But, through strongly worded dicta, it sent a clear message to lower courts that the scope of liability was to be exceedingly narrow.11 Most federal circuits have responded by erecting hyperstringent proof requirements that make it nearly impossible for plaintiffs to win.12

The increased prevalence of dominant platforms has provoked intense debate over whether the antitrust duty to deal needs a revival. There is widespread concern that dominant digital platforms like Google, Facebook, and Amazon exploit their market power to distort competition.13 The practices that have inspired the most controversy tend to raise thorny refusal-to-deal questions. Most examples involve a platform that vertically integrates into some product market and then allegedly discriminates against competing producers of that product by making it harder for them to reach consumers over the platform (or perhaps by excluding them from the platform altogether).

For example, Apple has been accused of manipulating the iOS App Store to discriminate against third-party apps that compete with Apple’s own apps.14 Similarly, critics accuse Facebook of barring third parties from its platform for trying to compete with it.15 Amazon allegedly discriminates against competitors who rely on its platform to sell their products.16 And in ongoing litigation between Epic Games (creator of the popular game Fortnite) and Apple, one relevant question is whether Apple violated the antitrust laws by refusing to let rival-made app stores on the iPhone.17 The common thread in such cases is a powerful platform defendant accused of refusing to deal with rivals or otherwise discriminating against them.18

These and similar practices by many large tech firms have spurred an outpouring of recent scholarship on antitrust and platforms, much of which focuses on platform refusals to deal and related conduct.19 Additionally, congressional representatives have recently proposed a number of antitrust bills aimed at platform conduct, including refusals to deal.20 Although these particular bills may not pass, the issue’s growing salience suggests that any eventual legislative reform will include operative changes focused on dominant firms who refuse to deal with rivals.

This Article offers a critical reexamination of antitrust policy toward unilateral refusals to deal. It proposes a framework that would address concerns about investment and administrability while still allowing for procompetitive enforcement in meritorious cases. It places particular emphasis on cases involving digital platforms, although the analysis applies to monopolists of all stripes.

For decades, much of the academic debate surrounding the duty to deal has simply focused on what the operative standard of liability should be.21 However, I argue that no amount of tweaking the existing standard will provide desirable results until a more fundamental problem is addressed: the longstanding but mistaken assumption that all refusal-to-deal cases raise the same theory of harm and should thus be evaluated under the same legal rules.22 Based on that assumption, existing law applies a common rubric to all cases accusing a defendant of (1) unilaterally23 and (2) unconditionally24 (3) refusing to deal (4) with an actual or potential competitor. That is, all cases satisfying these conditions are evaluated under the same liability standard,25 which is almost impossible to satisfy.26

In this Article, I identify an important economic distinction between two categories of unilateral refusals. Although they both satisfy conditions (1)-(4), they raise very different economic considerations and have no business being evaluated under a common standard. I refer to these as “primary” and “secondary” refusals to deal.

An example helps to illustrate the distinction.27 Suppose a defendant develops a network of satellites capable of beaming high-speed internet access to consumers. Assume that there are no competing satellite internet service providers and that the satellites were very expensive to develop. The defendant’s control over the satellites thus gives it monopoly power in rural areas where residents lack access to conventional high-speed internet service.

In a primary-refusal case, the alleged exclusion occurs in the defendant’s primary product market, which in this example is the internet-service market. Suppose prospective rivals want to introduce competing satellite-internet companies, but cannot afford the high costs of constructing their own satellites. Suppose further that they could use the defendant’s satellites to deliver their own internet service without inhibiting the defendant’s ability to continue providing its own service. But the defendant refuses to sell the rivals the right to use its satellites. This forestalls competition in the satellite-internet market, preserving the defendant’s monopoly power.

By contrast, in a secondary-refusal case, the alleged exclusion occurs in an adjacent (“secondary”) market into which the defendant has vertically integrated. Suppose that, in addition to satellite internet, the defendant offers an internet-based telephone service, known as a “voice over internet protocol” (VoIP). There are numerous competing VoIP services that have already been developed by rivals. However, the defendant blocks them from being transmitted over its satellites—that is, it refuses to let rival VoIP services run on its network. Consequently, the defendant’s internet subscribers are unable to buy VoIP service from anyone other than the defendant. This gives the defendant a VoIP monopoly in rural areas, whereas there would otherwise be vigorous competition among many VoIP service providers.

In fact, the second hypo raises substantially the same theory of harm as a tying arrangement28—namely, that a dominant firm is exploiting its preexisting monopoly in a primary market to foreclose competitors in a secondary market.29 In tying cases, the plaintiff usually does not dispute that the defendant obtained its primary monopoly on the merits; and its desired remedy (an injunction of the tie) would leave that monopoly intact. It would only prevent the defendant from using its primary-market monopoly to distort competition elsewhere. Similarly, in the secondary-refusal example, a duty to deal would not diminish the defendant’s monopoly over satellite internet. It would only prevent the defendant from exploiting it to foreclose competitors in the VoIP market.

By contrast, a primary refusal is not analogous to any traditional type of antitrust violation. Rather, it is most similar to a hypothetical violation known as “no-fault monopolization.” This doctrine would subject a firm to antitrust liability just for being a monopolist, even if it did not engage in any anticompetitive misconduct. However, antitrust law rejects this theory of liability,30 mainly because it would harm incentives for investment.31 Antitrust intervention in primary-refusal cases raises the same concern. The defendant’s satellites were expensive to develop, and it may well have decided not to do so if it anticipated an obligation to share them with rivals. After all, the ensuing competition will substantially diminish the profits it earns from its internet service.

It therefore makes sense that the law would strongly disfavor intervention in primary-refusal cases. But secondary refusals do not raise the same concerns.32 Ideally, secondary cases would be evaluated under a standard that resembles those we apply to tying arrangements and similar restraints. But courts do not permit this.33 This is grounded not in any principles of antitrust economics, but rather in an unjustified assumption that all types of unilateral refusals are equally deserving of judicial hostility.

Current doctrine’s most glaring problems become apparent in its application to primary-refusal cases. There is a profound contradiction between the liability standard courts apply and their oft-stated dictum that the duty to deal should be imposed only in very rare circumstances. Whatever its ambiguities, the prevailing standard of liability is plainly an exclusion-based standard: liability is supposed to hinge on whether or not the defendant’s refusal is exclusionary.34 Although this is the norm in section 2 cases, it leads to major problems when applied to primary refusals.

As I explain, simple economic arguments show that primary refusals by dominant firms are in fact routinely exclusionary in the very sense that the law purports to condemn.35 That is, they are routinely profitable only because they forestall competition by excluding rivals. As such, if an exclusion-based standard were faithfully applied, courts would impose liability quite regularly. But in practice the exclusion-based standard is not faithfully applied. Instead, courts have erected very demanding proof requirements that grossly exaggerate the difficulty of inferring an exclusionary motive, making it nearly impossible for plaintiffs to win.36 Courts often suggest that these proof requirements are necessary to avoid an undue risk of judicial error.37 But, on closer inspection, it is clear that these rules are pretextual, for they bear little logical connection to exclusion. To reinforce this conclusion, I show that the refusals most likely to generate significant exclusionary effects are also the least likely to satisfy existing proof requirements.38

Applying an exclusion-based standard to primary refusals is problematic because it necessarily ignores the main risk posed by intervention in such cases, which is that it would chill investment in valuable technologies.39 Indeed, a primary refusal to deal is highly analogous to a situation in which a patent monopolist (e.g., the seller of a patented drug) refuses to license would-be competitors. The obvious inference is that this refusal is intended to preserve the patent owner’s monopoly by excluding rivals.40 Patent law allows this in order to protect incentives for investment in innovation. But a purely exclusion-based standard does not permit this. By its very definition it rejects any “right to exclude.”41

Thus, in primary-refusal cases, courts are forced to find other ways of withholding liability, leading to the proof requirements mentioned above. But these rules end up preventing meaningful antitrust scrutiny of secondary-refusal cases as well, even though intervention in those cases does not raise the same normative concerns. On the contrary, an exclusion-based standard makes perfect sense in secondary-refusal cases—for the same reason that it makes sense in ordinary tying cases. This is a serious problem for enforcement against platform monopolists, since almost all such cases are secondary.42

This Article’s main proposal is simple: primary and secondary refusals should be subjected to different legal standards. First, primary refusals should be free from antitrust scrutiny unless an external source of law (most likely intellectual property (IP) or regulatory law) indicates that the defendant should be denied the right to exclude rivals from its input technology.43 This is because the propriety of intervention in these cases hinges on a complex IP-like balance between competition and investment, and the practical and institutional characteristics of antitrust law leave it ill equipped to implement such a balancing act.44 Instead, antitrust would defer to other areas of law that confront the competition/investment tradeoff more directly.

Second, plaintiffs should be able to challenge secondary refusals as de facto ties or similar vertical restraints. As with a tying arrangement, secondary refusals should invite liability only if the plaintiff can show appreciable foreclosure in the secondary
market.45 Thus, the courts would apply a familiar foreclosure-based standard of liability, not the far more stringent (and largely inapposite) standard presently applied in refusal-to-deal cases. Significantly, this aspect of the proposal arguably would not contravene any Supreme Court decisions, as all clearly conflicting precedents come from lower courts.

To be sure, secondary refusals differ from traditional tying arrangements in some significant ways. For example, some important defenses in conventional tying cases will not usually apply in secondary-refusal cases—and vice versa.46 And secondary-refusal cases are more likely to raise remedial difficulties than are traditional tying
cases.47 However, such difficulties are rarely prohibitive. On the whole, antitrust is well equipped to address secondary-refusal cases by making sensible modifications to existing standards on vertical restraints.

The benefits of this proposal are manifold. First, it provides a reasonable and practicable means of addressing investment concerns without unnecessarily stifling antitrust enforcement in meritorious cases.48 Second, courts are well positioned to administer this policy, as the most serious administrative difficulties are largely limited to primary cases (in which intervention would remain quite rare).49 Third, this policy is exactly analogous to the way antitrust already treats other forms of unilateral conduct.50 And finally, in contrast to the status quo, these changes would allow for meaningful antitrust scrutiny of most platform refusals, including antirival discrimination.

The Article is organized as follows. Part I provides a brief overview of how existing antitrust law treats unilateral refusals to deal with competitors. Part II discusses the distinction between primary and secondary refusals and provides many examples of each type. Part III discusses major problems and inconsistencies within current law. Part IV argues that the most serious of these problems could be addressed by subjecting primary and secondary refusals to different standards of liability. Part V discusses a number of considerations relating to the application of the framework proposed in Part IV. Finally, Part VI concludes.