There is widespread support for antitrust reform, fueled mainly by concerns about major platforms like Google, Facebook, and Amazon. Many believe that these companies have become too large and that they use their power in harmful ways. In the United States, some of the most aggressive reforms have recently been codified into two proposed bills, which focus largely on “self-preferencing” by online platforms and various restrictions imposed by mobile operating systems. This short article evaluates the proposed Big Tech reforms based on everything we’ve learned from antitrust’s successes and failures over the last fifty years. In some situations, the reforms could help to curb anticompetitive unilateral conduct that is extremely difficult to challenge under current law. Overall, however, the proposals are an ill-conceived, knee-jerk reaction to a set of complex issues requiring a more careful response. They do a very poor job of limiting antitrust scrutiny to cases that plausibly involve anticompetitive behavior. The self-preferencing proposals in particular offer no secure way to avoid scrutiny other than to stop introducing new products. For these and other reasons, these proposed reforms would have significant adverse effects on competition and innovation — the two things they are supposed to protect.
By Erik Hovenkamp[1]
I. INTRODUCTION
There is widespread support for antitrust reform, fueled mainly by concerns about major platforms like Google, Facebook, and Amazon. Many believe that these companies have become too large and that they use their power in harmful ways.
In the United States, some of the most aggressive suggested reforms have recently been codified into two proposed bills, both of which appear to have some bipartisan support. The American Innovation and Choice Online Act (“AICOA”) concerns unilateral conduct by large online platforms, focusing especially on “self-preferencing”— treating one’s own products more favorably than those of competitors.[2] The Open App Markets Act (“OAMA”) deals specifically with mobile app stores and operating systems.[3] It similarly prohibits certain forms of self-preferencing. But it would also eliminate the “walled garden” business model that requires all app transactions to run through a single app store.
There are indeed good reasons for thinking antitrust could benefit from pro-enforcement reform. Courts’ decisions are often guided by simplistic maxims — for example, that markets beset by anticompetitive conduct will conveniently “self-correct”— that reflect an idealized conception of competition. This has left the law increasingly out of synch with modern economics, which has a much deeper appreciation for the frictions and imperfections that leave real-world markets vulnerable to strategic behavior. As a result, courts tend to be unreasonably dismissive of antitrust claims in general.
This short article evaluates the most popular proposed reforms aimed at Big Tech, including those in AICOA and OAMA. It considers how they would likely affect competition and innovation based on everything we’ve learned from antitrust’s successes and failures over the last fifty years.
There is one area in which the proposed reforms would genuinely help to curb anticompetitive conduct that is extremely difficult to challenge under existing law. This involves cases where a dominant platform unilaterally refuses to deal with competing sellers or otherwise impairs their ability to make sales over the platform.
However, on the whole, the proposed reforms do not represent a judicious effort to fix the problems with current law. Rather, they are an ill-conceived, knee-jerk reaction to a set of complex issues requiring a more careful response. The core problem is that the proposals simply do not make a serious effort to limit antitrust scrutiny to cases involving anticompetitive conduct. Indeed, in many instances they do not impose any affirmative obligation on the plaintiff to prove anticompetitive effects.[4] This is exacerbated by the fact that the violations articulated by the proposals are generally vague and broad in scope. In the case of the self-preferencing proposals, the only clear way for a large platform to avoid antitrust scrutiny may be to abstain from introducing new products. Consequently, while the proposed reforms are intended to promote competition and innovation, they are much more likely to diminish both.
II. PLATFORM REFUSALS TO DEAL
If challenged under existing law, most of the unilateral platform conduct at issue in the proposed reforms would be evaluated as a refusal to deal.[5] This doctrine — which applies in cases where a dominant refuses to do business with rivals or otherwise discriminates against them — has long been controversial.[6] In its 2004 Trinko decision,[7] the Supreme Court came close to killing it off altogether. Although the Court did not provide much specific guidance to lower courts, its sharp dicta made it clear that the scope of liability was to be extremely narrow.[8] Circuit courts have shaped their standards accordingly.[9]
A partial success of the proposed reforms is that they would do away with the modern refusal-to-deal framework in cases involving dominant platforms. That’s a sensible thing to do, because Trinko and its progeny are unreasonably dismissive of unilateral refusal claims.[10] Additionally, the legal standards courts presently apply in such cases are not attuned to the appropriate theory of harm, and consequently they do a remarkably poor job of identifying those cases in which intervention would make sense.[11]
As I have argued elsewhere,[12] meritorious refusal-to-deal cases generally involve two distinct product markets and raise essentially the same theory of harm as anticompetitive tying: namely, that the defendant is exploiting its monopoly over a primary product to foreclose rivals in a secondary market.[13] As in a tie, this is accomplished by making it hard or impossible for secondary market rivals to make sales to users of the primary product. But the defendant achieves this not through vertical agreements with consumers, but rather by refusing to let rivals access the primary product. This can generate foreclosure in the secondary market if many consumers rely on the primary product to access the secondary one.
Consider an illustration. In Microsoft,[14] the government prevailed in its Section 2 challenge against Microsoft’s tie of its Windows operating system and its Internet Explorer browser. The tie foreclosed competing browsers like Netscape. This was not a unilateral conduct case, because the tie was effectuated in part through vertical agreements with equipment manufacturers. However, consider a slight change in the facts. Suppose Windows contained a proprietary app store and that all software programs had to be obtained through it. In that case, Microsoft could have achieved the same result as its tie by simply refusing to let competing browsers like Netscape onto its app store. As with the tie, this would restrain consumers’ ability to use competing browsers on a Windows machine.
Although this change in facts would not alter the relevant competitive effects, it would transform Microsoft into a refusal-to-deal case, making it virtually impossible to win under current law. This shows that the modern approach to unilateral refusals leads to nonsensical results through arbitrary line-drawing. If a unilateral refusal acts just like an anticompetitive tie, why should it be saved by the absence of a vertical agreement?[15]
A logical position to take in Section 2 cases is that if unilateral conduct emulates a vertical restraint then courts should evaluate it as such. That is, they should consider the same theory of harm, and they should look for the same markers of anticompetitive effects (namely foreclosure). In fact, this is more or less the approach courts already take toward other types of unilateral conduct.[16]
A good example is Lorain Journal,[17] which involved a conditional refusal to deal with customers.[18] The defendant ran the local newspaper, which gave it a dominant position in the market for advertising services. A new radio station threatened the defendant’s market position, however. Local businesses began doing some of their advertising through the radio station. The defendant responded by refusing to provide advertising services to any businesses who ran ads over the radio station.
It was clear that the defendant’s goal was to force most customers to make it their exclusive provider of advertising services. This foreclosed the radio station and threatened to maintain the defendant’s dominant position in advertising. Although the Court did not say so explicitly, this was a case of de-facto exclusive dealing. And that is essentially how the Court treated it. In other cases, a defendant’s unilateral conduct is instead treated as a form of de-facto tying.[19]
The point is that courts already recognize that unilateral conduct may emulate a vertical restraint, and in such cases, they usually evaluate the conduct as such. But, for whatever reason, they do not yet acknowledge that the same approach may be appropriate in some cases involving a unilateral refusal to deal with competitors.[20]
The proposed Big Tech antitrust reforms would allow for antitrust intervention in some meritorious cases involving unilateral refusals by dominant platforms. That’s a good thing. Unfortunately, as the remaining sections explain, they do a very poor job of limiting antitrust scrutiny to those cases that plausibly threaten competition.
III. SELF-PREFERENCING
An outright refusal to deal is the most severe way that an integrated platform can discriminate against rivals in an adjacent market. By contrast, the term “self-preferencing” generally refers to very mild forms of discrimination. This typically involves the defendant giving its own product the most visible or prominent placement on its platform — such as by listing it first in the search results — without otherwise degrading consumer access to competing products.
For example, when a consumer looks up a restaurant on Google Search, the results page displays a Google “OneBox” at the top of the page — a box of relevant information about the restaurant, such as its address, phone number, and reviews. The reviews come from Google, not from Yelp. (However, the restaurant’s Yelp page will typically be among the first organic search results on the page.) Yelp has alleged that this is anticompetitive because it prioritizes Google’s content over that of Yelp, whose content is superior (according to Yelp, at least).[21] Amazon has similarly been accused of promoting its own products above those of competitors on its storefront.[22] And Apple has been accused prioritizing its own apps over competing apps within the iOS App Store.[23]
To begin, it is worth noting that mild self-preferencing is in fact ubiquitous, and not only in concentrated markets. Whenever a firm offers two complementary goods or services, it will have an incentive to encourage customers of one to consider using the other as well. A grocery store might give the most visible shelf space to its own private label goods. A cable network might use the most desirable advertising slots to run ads for its own programming. And a hotel brochure might list the hotel’s own restaurant above those of local competitors. But nobody seems to view examples like this as problematic. If nothing else, this shows that self-preferencing is not inherently anticompetitive.
Proponents of the proposed reforms will point out that these examples are different because they do not involve dominant platforms that consumers depend upon to find products. A dominant platform may indeed have the ability to materially undermine competition in some of the markets that rely on them, as discussed in the previous section. However, mild self-preferencing does not implicate those concerns, because it does not materially impair consumers’ ability to pick rivals’ products over those of the platform.
Mild self-preferencing allows a platform to capture a larger share of those consumers who are indifferent among the competing alternatives — those who are inclined to click on the top result no matter what it is. To be sure, capturing these consumers could be very lucrative. But that does not make it anticompetitive exclusion. The latter involves a dominant firm capturing many customers who strictly preferred to buy from the firm’s competitor instead. That requires a meaningful restraint — something that prevents consumers from choosing their preferred brand. Merely listing the platform’s own product first does not do this.
It is also worth noting that it is not usually possible to make all competing alternatives equally visible on a platform’s storefront. For example, in a list of search results, someone has to be listed first. Why shouldn’t it be the platform? One possible response was hinted at in the Yelp example mentioned above. A plaintiff might argue that it’s unfair for a platform to give its own product the most visible placement if there exist superior competing alternatives. In other words, perhaps the highest quality seller deserves to be listed first. However, this raises an obvious practical difficulty: who’s to say which product is best? There could be many firms with plausible rationales for ranking their products first, including the platform itself. This is a largely subjective inquiry that courts are not equipped to resolve in any satisfying way.[24]
Such practical challenges are a reminder of why antitrust generally takes a passive approach, condemning specific anticompetitive acts rather than attempting to micromanage markets to maximize efficiency. The latter approach is simply infeasible in most situations. But it is also important to remember that the antitrust system is not costless even when it works as intended. Antitrust cases take years and cost many millions of dollars to litigate. This calls into question the wisdom of using antitrust to police conduct that has such a limited impact on competition.
The high costs of antitrust litigation — and the uncertain scope of antitrust liability under the proposed reforms — would threaten incentives for competition and innovation. They may discourage platforms from introducing desirable new products based on the fear that competing sellers could then sue for any perceived slight. Indeed, no matter what an integrated platform does, there will almost always be some rival who can claim to be a victim of unequal treatment. For example, even if Amazon lists its own batteries 4th in the search results, it will still have to worry about potential complaints by rivals listed 5th or below. It might decide that remaining in the battery market would be more trouble than it’s worth. And that withdrawal of a significant competitor likely has a far greater impact on competition than Amazon’s position in the search results. In this way, overbroad restrictions on self-preferencing may have the perverse effect of diminishing both competition and innovation.
If Congress is determined to police self-preferencing, it would be helpful to include some clear and simple safe harbors. One possibility would be to grant the platform safe harbor if it applies a visual marker to its own product listings, similar to the “sponsored” label that attaches to paid listings appearing in internet search results. The marker would convey to consumers that the product’s placement on the platform may not be organic. Such a safe harbor would reduce uncertainty about potential liability and would not strongly interfere with trade.
IV. ANTICOMPETITIVE DISCRIMINATION AND FORECLOSURE
Mild self-preferencing does not materially impair rivals’ ability to make sales, but more severe forms of discrimination could do so, even if they fall short of a full-blown refusal to deal. For example, suppose a rival’s product initially appears at the top of the first page of Google Search results but Google demotes it to the bottom of the 10th page. Very few consumers will click through to the 10th page, so this is practically the same as excluding the rival from the search results altogether. If the rival were instead demoted to, say, the 3rd page, there could still be some anticompetitive effects, but they might be less acute.
The competitive impact of platform discrimination is thus a matter of degree. This raises the question of what standard or test antitrust should use to distinguish those cases in which liability is appropriate. The most natural answer is to limit antitrust scrutiny to those cases in which the discriminatory conduct could plausibly generate appreciable foreclosure in the adjacent product market.[25]
This is consistent with how antitrust already evaluates other exclusionary practices that resemble tying. A good example is bundled discounting.[26] If the discount is sufficiently large, there is no practical difference between this arrangement and a literal tie. In that case, foreclosure is certainly plausible if the defendant is dominant. But if the discount is tiny, then the arrangement will not disturb competition at all. To draw the line between harmful cases and benign ones, courts require some showing that the discount is large enough to create a risk of significant foreclosure.
Lawmakers could improve the proposed reforms substantially by requiring plaintiffs to show a likelihood of appreciable foreclosure in order to carry their initial burden. This would not necessarily have to require the same amount of foreclosure that courts demand in other types of exclusion cases.[27] But it is not reasonable to let plaintiffs carry their initial burden without having to show any foreclosure at all.
V. PLATFORM VERTICAL INTEGRATION
The most extreme proposed reform would prohibit large platforms from vertically integrating into adjacent product markets. For example, one proposed bill, the Ending Platform Monopolies Act,[28] would prohibit any sufficiently large platform from owning or controlling any “line of business other than the covered platform” that creates a “conflict of interest” in the sense that it could create an incentive to use the platform to exclude or discriminate against rivals in the same line of business.
Under such proposals, Amazon could not introduce products that compete with any other goods sold on Amazon. Microsoft could not sell software that competes with other Widows-compatible software. And Apple and Google could not offer apps that compete with any other apps sold through their app stores.
It is hard to overstate how misguided this proposal is. Vertical integration offers significant efficiency benefits, such as reducing transaction costs or eliminating a double markup problem. Moreover, when integration is achieved through internal expansion rather than merger, this is form of competitive entry. The proposal would block such entry by large platforms. In addition to diminishing competition, this would eliminate platforms’ incentive to develop innovative new products.
Clearly the better approach is to attack anticompetitive conduct by integrated platforms directly whenever it occurs (and thus to deter it from occurring in the first place). That would allow antitrust to purge the relevant harms without creating widespread collateral damage to competition and innovation. This point has been well-understood in antitrust for many decades.
The hostility toward platform vertical integration likely derives not from traditional antitrust concerns, but rather a desire to protect small businesses from the rigors of competition. A large platform will often benefit from economies of scale or scope that allow it to offer better deals than smaller rivals can afford to match. There is a common populist sentiment that this is unfair. But antitrust’s objective is to protect the competitive meritocracy, not to predetermine who should win and who should lose. That determination is properly left to consumers.
VI. WALLED GARDENS
On Apple and Android mobile devices, a consumer can obtain apps only through the official app store associated with her operating system. These app stores take a cut (usually 30 percent) of app transactions, including purchases of in-app content. One significant proposed reform, which is codified in OAMA, is to prohibit this “walled garden” business model. Specifically, it would prevent an app store from requiring app makers to use its own in-app payment system. And it would require the companies who control the most popular mobile operating systems (most notably Apple and Google) to let users “sideload” apps — to acquire them by means other than the companies’ official app stores.
One widespread concern with this proposal is that, by eliminating the major platforms’ ability to screen out malicious or invasive apps, it could undermine users’ security and privacy interests. This is a valid concern that Congress should investigate carefully before adopting the proposals into law.[29]
The more traditional antitrust question centers on the proposal’s likely economic effects. This includes its impact on prices and on the selection of apps available to consumers, among other things. The focus on the security concerns has distracted from these questions to some extent. But it would be a mistake not to consider them carefully, because it is not obvious that the economic effects of this proposal would be beneficial.
Eliminating the walled garden model is likely to have polarizing economic effects on consumers. It will likely benefit “high intensity” users who purchase a lot of apps and in-app content. But it will likely harm low intensity users who spend little on apps. This makes it hard to conclude that mobile users will collectively benefit from the elimination of walled gardens — particularly since most mobile device users are in the low-intensity group.
The key point is that, while eliminating walled gardens will likely reduce prices of apps and in-app content, it will also likely raise prices at the device level. A company that controls an operating system faces a two-part pricing problem, because there are two distinct ways it can charge prices.[30] This leads to two alternative pricing models. First, the OS maker might decline to take a cut from app transactions and instead charge a one-time lump sum fee for access to its OS. This has long been the model employed by Microsoft Windows. Microsoft charges equipment manufacturers a license fee for installing windows on their computers; much of these fees will then be passed through to consumers in the form of higher computer prices. But Microsoft does not subsequently take a cut out of software transactions.
Alternatively, the firm could make the OS cheap or even free, and instead collect fees by taking a percentage cut of each software purchase. This model, which economists sometimes call “metering,” is the one currently employed on smartphones. For example, Google does not charge fees for its Android OS, but it takes a cut of most app transactions in its app store.
Metering is very common. Its purpose is to keep the product affordable to the large majority of consumers while allowing the firm to extract more fees from those consumers who use it more intensively. The latter is achieved through aftermarket transactions that gauge (or “meter”) the intensity of a consumer’s use. For example, a video game console might be sold at or below cost, and the seller will instead earn its profits by taking fees from game sales. Similarly, many printer manufacturers charge low prices for the printer and high prices for the ink cartridges.
Metering is often essential in network industries, where success depends on building a large network of users. In these cases, it is important not to price low-intensity users out of the market by charging high upfront access fees. Instead, the firm will often give everyone basic access for free, but then charge fees for premium aftermarket content that only appeals to high-intensity users. This is embodied in the “freemium” pricing model employed by many mobile games.
The hitch with metering is that the seller must be able to prevent customers from circumventing its aftermarket fees by transacting with third parties instead.[31] In the mobile context, this is where the walled garden comes in. It ensures that the OS maker can charge fees for all app transactions by requiring that all such transactions run through its own app store.
If walled gardens are outlawed, an OS maker will maximize profits by shifting to the Windows-style model under which it charges a high fee upfront but does not take a significant cut from most app transactions. In the case of a vertically integrated firm like Apple, which sells its own devices utilizing its own OS, this will take the form of a direct increase in the device price.[32] Either way, the result will be cheaper app transactions but more expensive mobile devices.
One could still attempt to justify the policy shift by arguing that the benefits of lower transaction fees will outweigh the higher prices of devices.[33] However, this is an empirical question, and to my knowledge policymakers did not consider it before proposing to eliminate the walled garden model.
Moreover, any justification Congress might give will have to confront the most unappetizing aspect of its proposal: it effectively asks the large majority of mobile device users (most of whom do not spend much on apps) to subsidize a relatively small group comprising mainly children and young adults who spend a lot on mobile games. Eliminating walled gardens will benefit the latter users (and perhaps their parents), but only at the expense of raising overall costs for everyone else.
Of course, a platform’s ability to control the flow of app transactions on its OS can lead to competition concerns other than just higher transaction fees. Most notably, the firm might exploit this control to exclude apps that compete with its own apps. This could affect not only app pricing but also the variety of apps available. As discussed above, that is an important concern that current law largely dismisses. But the most direct way to address this concern is to make sensible reforms to Trinko’s unreasonably restrictive refusal-to-deal framework.
VII. PRESUMPTIONS AND PROOF BURDENS
In an exclusion case, a plaintiff typically must show market power along with some indicia of anticompetitive harm. It is one thing to argue that these proof requirements have become unreasonably hard to satisfy (a position I agree with). It is another to suggest that they should be eliminated altogether. But the proposed reforms come close to doing exactly that.
First, in most instances, the proposed reforms place no burden on the plaintiff to prove a likelihood of anticompetitive effects. Rather, the burden is typically on the defendant to disprove them. For example, AICOA enumerates ten violations, and in seven of them anticompetitive harm is simply presumed from the conduct itself. The defendant then carries the burden of rebutting that presumption. OAMA similarly presumes anticompetitive harm, but it recognizes only certain specific defenses relating to user protection, national security, and IP rights.
Second, the proposals eschew any analysis of market power. Instead, they effectively presume market power whenever the platform is sufficiently large, measured either in terms of its market capitalization or subscriber base. Relatedly, they presume that control of a large platform is sufficient to disrupt competition in any adjacent product markets.[34] This is problematic given that most large tech platforms operate in many different product markets that vary widely in the extent to which they rely upon the platform. For example, suppose Amazon sells its own coffee, but that 99 percent of all retail-level coffee is purchased in brick-and-mortar grocery stores. Then it is hard to see how Amazon could hope to undermine competition in the coffee market. As this illustrates, the fact that a platform is large does not imply that it has the power to thwart competition in every market that it operates in.
VIII. CONCLUSION
There are good reasons for seeking pro-enforcement antitrust reform, including in high-tech industries. But to be successful, such an undertaking must make a serious effort to distinguish anticompetitive practices from reasonable or benign ones. The proposed Big Tech reforms fall woefully short in this regard. They shoot first and ask questions later. As a result, they pose a major risk to competition and innovation. Policymakers should continue to pursue valuable antitrust reform, but they should not settle for ill-conceived proposals designed to achieve liability by any means necessary. We can do much better.
[1] Assistant Professor, USC Gould School of Law. I’m grateful to Steve Salop for providing helpful feedback.
[2] S.2992, https://www.congress.gov/bill/117th-congress/senate-bill/2992/text.
[3] S.2710, https://www.congress.gov/bill/117th-congress/senate-bill/2710/text.
[4] See Section VII, infra.
[5] Alternatively, it might be evaluated as a denial of access to an essential facility, which is a very similar type of claim.
[6] See e.g. Phillip Areeda, Essential Facilities: An Epithet in Need of Limiting Principles, 58 Antitrust L.J. 841 (1990) (criticizing the doctrine).
[7] Verizon Commc’ns Inc. v. Law Offs. of Curtis V. Trinko, LLP, 540 U.S. 398 (2004).
[8] Id. at 399 (describing the doctrine as “at or near the outer boundary of §2 liability”).
[9] To my knowledge, no plaintiff has won a final judgment since Trinko. In one recent case, the plaintiff prevailed in district court, but the judgment was reversed by the 9th Circuit on appeal. Fed. Trade Comm’n v. Qualcomm Inc., 969 F.3d 974, 994 (9th Cir. 2020).
[10] Of course, ideally the problems created by Trinko would be eliminated everywhere, not just in Big Tech.
[11] Erik Hovenkamp, The Antitrust Duty to Deal in the Age of Big Tech, 131 Yale L. J. 1483 (2022), papers.ssrn.com/sol3/papers.cfm?abstract_id=3889774.
[12] Id.
[13] In platform cases, the primary product is typically the platform itself, while the secondary good is something sold or marketed over the platform.
[14] United States v. Microsoft Corp., 253 F.3d 34, 58 (D.C. Cir. 2001).
[15] The only plausible response is that intervention in a unilateral refusal case could raise administrative challenges, while enjoining an agreement does not. However, no such difficulties arise if the platform already deals voluntarily with third parties, as would be expected in a case where a platform selectively excludes specific competing products. In that case, the court can simply order the defendant to deal with its rival on the same terms it offers to everyone else.
[16] See Hovenkamp, supra note 11, at 1535-38.
[17] Lorain Journal Co. v. United States, 342 U.S. 143 (1951).
[18] In a conditional refusal, the defendant stops dealing with customers who do not abide by its preferred conditions. This is often (but not always) treated as unilateral conduct. Unlike a traditional unilateral refusal case, there is no allegation that the defendant unlawfully refused to deal with competitors.
[19] Examples include cases of bundled discounting or exclusionary product design. The latter arise when a product is redesigned in a way that renders it incompatible with competing versions of a complementary good — a practice sometimes called “tech tying.”
[20] Not all unilateral refusal cases are properly evaluated in this way, however. Some do not raise a tie-like theory of harm, but rather represent an effort by prospective rivals to free ride on the defendant’s technology. Such cases are likely responsible for most of the judicial hostility toward unilateral refusal cases generally. See Hovenkamp, supra note 11.
[21] See e.g. Lauren Feiner, Yelp gives senators its list of grievances against Google in antitrust hearing, CNBC (March 10, 2020), www.cnbc.com/2020/03/10/yelp-testifies-against-google-in-antitrust-senate-hearing.html.
[22] See e.g. Sara Morrison, The true cost of Amazon’s low prices, Vox (Jan 13, 2022), www.vox.com/recode/22836368/amazon-antitrust-ftc-marketplace.
[23] See e.g. Jack Nicas & Keith Collins, How Apple’s Apps Topped Rivals in the App Store it Controls, NY Times (Sept. 9, 2019), www.nytimes.com/interactive/2019/09/09/technology/apple-app-store-competition.html.
[24] One might posit that courts could instead police platforms’ ranking algorithms to make sure they rank different products in a reasonable way. But this is likely to raise the same kinds of practical difficulties, as there may plausible justifications for numerous alternative ranking algorithms. And the algorithms themselves are likely to be difficult for courts to evaluate.
[25] See Hovenkamp, supra note 11, at 1544-46.
[26] This is where a defendant sells two complements separately while also offering a discount to consumers who buy a bundle of both goods.
[27] As an example, in a tying or exclusive dealing case, courts usually require about 40 percent foreclosure to establish a prima facie violation.
[28] H.R. 3825, https://www.congress.gov/bill/117th-congress/house-bill/3825/text.
[29] OAMA does offer safe harbors for certain conduct aimed at protecting user security or privacy, but there is disagreement as to whether they are sufficient to allay the concerns. See e.g. Caitlin Chin, Breaking Down the Arguments for and Against U.S. Antitrust Legislation, Center for Strategic and International Studies (April 22, 2022).
[30] On the economics of two-part pricing, see Richard Schmalensee, Monopolistic Two-Part Pricing Arrangements, 12 Bell J. Econ. 445 (1981).
[31] For example, in the case of printers and ink cartridges, metering will not work if consumers can easily buy compatible ink cartridges from third parties.
[32] To see why, note that, under metering, the average app transaction fees collected per device act like a subsidy on device sales. This induces the seller to cut the price of the device. Absent metering, that subsidy goes away and the firm will raise the price of the device.
[33] For example, it is possible that device level competition could prevent device prices from rising by enough to outweigh the benefits of lower transaction fees. In general, however, the net effect of eliminating metering is ambiguous. See, e.g., Erik Hovenkamp & Herbert Hovenkamp, Tying Arrangements, in Oxford Handbook of International Competition Policy, (D. Daniel Sokol & Roger Blair, eds.) (2014).
[34] AICOA does require that a platform is a “critical trading partner” in the sense that it has the power to “materially impede” a seller’s ability to reach consumers. However, under the bill this need be true for only one product sold over the platform. Thus, there is no need to show the platform has such power over the specific product market at issue in a given case – only that it has such power over some market.
