The House Report assembles a large amount of information about competition in digital markets and will be a valuable resource for those who want to understand the leading digital platforms. It falls far short, however, of its stated objectives of examining “whether dominant firms are engaging in anticompetitive conduct” and whether existing antitrust laws “are adequate to address those issues.” As to the former, it does not say what it means by anticompetitive conduct, often fails to examine possible exculpatory explanations of the conduct and documents it discusses, and does not address issues of efficiency and economic welfare that are central to antitrust analysis. As to the latter, it does not discuss whether antitrust law is or might become adequate to address the issues it describes or the costs of the policy changes it recommends.
By A. Douglas Melamed1
I. INTRODUCTION
The so-called “House Report,” the lengthy report by the majority staff of the Antitrust Subcommittee of the House Judiciary Committee, assembles a large amount of interesting and often important information about competition in digital markets. It will be a valuable resource for those who want to understand the leading digital platforms. It is, however, a missed opportunity that falls far short of its stated objectives of examining “whether dominant firms are engaging in anticompetitive conduct” and whether existing antitrust laws “are adequate to address those issues.”
The Report explains in some detail how many digital markets function, the role of network effects, the several entry barriers, and the tendency of those markets to tip so that one firm accounts for most or all of the revenues. While its analysis of market definition and market power is not rigorous, it certainly suffices to persuade at least this reader that the major platforms – Amazon, Facebook, and Google – very likely have market power in their core businesses – online commerce, social networks, and search.
The analysis of alleged anticompetitive conduct by the dominant firms and of the antitrust and policy issues raised by that conduct is less persuasive. One, perhaps understandable reason is that the staff chose breadth – discussion of a wide variety of behaviors involving numerous lines of commerce – over deep analysis of any of them. Less forgivable is the failure of the Report to define core concepts, including even the term “anticompetitive conduct.” The Report seems to rest on the unstated assumption that aggressive conduct by a dominant firm to gain market share and revenues is anticompetitive when it harms rivals. That premise is inconsistent with existing law, and the Report nowhere addresses the difficult policy issues that need to be addressed in order to determine whether existing law is adequate or some other law would be better. Nor does the Report address the important issue of economic efficiency, which is central to antitrust law and policy and to economic welfare.
In the following paragraphs, I illustrate these shortcomings by examining, first, some conduct issues and, second, some related policy issues addressed in the Report.
II. ANTICOMPETITIVE CONDUCT
The Report discusses a vast range of behaviors by the dominant platforms – acquisitions, use of data, pricing, vertical integration, and alleged self-preferencing, among others. But, while it quotes provocative language from company statements and documents and uses manipulative language – misappropriation, predatory, and the like – it does not even attempt to identify the line between permissible and impermissible conduct. As a result, the reader cannot determine whether any particular conduct really was anticompetitive. Three among many possible examples, which happen to be three of the most common complaints about the platforms, illustrate the problem.
A. Facebook’s Acquisition of Instagram
The basic facts are familiar. Facebook bought Instagram when the latter was much smaller than it is today and has overseen its growth into a major social network platform. The fundamental factual question that needs to be addressed to determine whether the acquisition, even with the benefit of hindsight, can be said to have been anticompetitive is whether the acquisition (i) strengthened Instagram and enabled it to be more successful and more valuable to users and advertisers than it would otherwise have been or (ii) prevented Instagram from growing into a more fundamental competitive threat to Facebook, or maybe even establishing a new social network paradigm. The Report does not directly address those issues.
Instead, the Report tries to assess the desirability of the acquisition on the basis of contemporaneous statements that appear to reflect at least some of Facebook’s perceptions at the time. The Report notes that “Mr. Zuckerberg wrote that acquiring Instagram would allow Facebook to integrate the product to improve its service. But, he added, that ‘in reality we already know these companies’ social dynamics and will integrate them over the next 12-24 months anyway.’ He explained: . . . one way of looking at this is that what we’re really buying is time. . . . If we incorporate the social mechanics they were using,” potential new competitors are less likely to succeed. The quoted language and the discussion in the Report surely suggest that the acquisition might have been about nipping a competitive threat in the bud. They do not, however, come close to proving that conclusion because they are equally consistent with a conclusion that the acquisition enabled Facebook to “incorporate the social mechanics” Instagram was using more quickly and at less cost than it could have done on its own, and thus enabled Facebook to provide a more efficient and valued social network sooner.
The Report does not stop there. It notes that “Mr. Zuckerberg suggested [to Instagram] that refusing to enter into a partnership with Facebook, including an acquisition, would have consequences for Instagram, referencing the product Facebook was developing at the time.” One implication might be that Zuckerberg threatened Instagram, which was otherwise confident of its path to commercial success, with predatory retaliation in order to induce Instagram to sell to Facebook. But another possibility, not discussed in the Report, is that Facebook knew that it would eventually drive Instagram from the market because it could combine Instagram’s “social mechanics” with its existing economies of scale and scope but that that route would be more costly and time-consuming than an acquisition. In substance, Zuckerberg might have presented Instagram with a choice of being driven from the market by lawful competition or sharing in the time and cost-saving efficiencies that an acquisition would create.
The drafters of the Report cannot be faulted for not knowing which explanation is correct. They did not have the resources of the enforcement agencies or a court. But they can be faulted for just assuming the correctness of one possible explanation and ignoring the others.
B. Agreements Making Google’s Search Engine the Default
These are the subject of the Justice Department’s suit against Google, and the undisputed facts are at the very least problematic. Google is the default search engine in its browser (Chrome) and in others that together account for 87 percent of browsers. Microsoft’s Edge (4 percent) sets Microsoft’s Bing as the default search engine. On its face, it seems clear that at least Google and Microsoft regard being the default search engine as valuable. Google very likely has monopoly power in search, and contracts that enable Google to be the default on a total of 87 percent of the principal means of distribution might well be a form of unlawful monopoly maintenance under the Microsoft case.
The Microsoft case, however, found to be unlawful only those agreements that served no efficiency purpose, so the question whether the default arrangements were efficient should be considered in order to understand whether they would be regarded as anticompetitive, at least under current law. Again, the Report focuses on facts from which inferences about effects and motives might be drawn. It notes that “Google executives closely tracked search defaults on Microsoft’s Internet Explorer and expressed concern that non-Google defaults could impede Google Search. . . . Google recommended that users be given an initial opportunity to select a search engine and that browsers minimize the steps needed to change the default provider. These discussions, as well as the steep sums Google pays Apple and various browsers for default search placement, further highlight the competitive significance of default positions.”
A complete investigation would inquire into how many, and how quickly, users of Microsoft Edge change the default browser. (The Justice Department complaint repeatedly alleges that very few users switch away from Google when it is the default search. That might reflect, not the difficulty of switching, but the superiority of Google search.) It is at least possible that switching is easy and common and that being the default benefits Google only because it gets to default status sooner and is thus able to acquire monetizable data sooner. In principle, the value of that time saving could be assessed and compared to the sums paid by Google for exclusive default status. If the sums paid exceed that value, the default agreements look like purchases of exclusion. If not, they might be efficient means of acquiring data. None of this is discussed in the Report.
C. Amazon’s Acquisition and Use of Data
The Facebook and Google examples discussed above are mostly about conduct that allegedly preserved the dominant positions of those firms in their core businesses. The allegations about Amazon’s acquisition and use of data are mostly about using Amazon’s dominance to aid its expansion into complementary businesses. The Report describes the problem as follows:
“Amazon leverages its access to third-party sellers’ data to identify and replicate popular and profitable products from among the hundreds of millions of listings on its marketplace. Armed with this information, it appears that Amazon would (1) copy the product to create a competing private-label product or (2) identify and source the product directly from the manufacturer to free ride off the seller’s efforts, and then cut that seller out of the equation. *** In one case, Amazon employees reportedly used non-public sales data.”
The manipulative terminology – e.g. leverage, cut out, free ride – suggests something nefarious. The passage as a whole, however, describes conduct that, at least in a static sense, is unambiguously efficient and welfare-enhancing. Amazon is able to take advantage of its economies of scale (including in data) and scope to create and market products that are either more valuable to consumers or less costly to produce and distribute than their predecessors. As the Report notes, “[d]ata allows companies to target advertising with scalpel-like precision, improve services and products through a better understanding of user engagement and preferences, and more quickly identify and exploit new business opportunities.”
It is not clear what the Report means by “cut that seller out of the equation.” The language suggests that Amazon stopped distributing the seller’s product after it developed its own. The Report does not address whether doing that reflected an efficient strategy of platform inventory management or entailed, instead, reduction of the value of the platform in order to aid Amazon’s foray into the complementary product space.
It is conceivable that Amazon acquired the data unlawfully or, in the case of the alleged use of non-public data, used it in violation of some legal or contractual constraint. But it is at least equally likely – perhaps more likely, given the silence of the Report on the issue – that the data were either not confidential or, in the case of the non-public information, were acquired by an agreement with the seller that did not prohibit the referenced use by Amazon. In that event, there was nothing improper under existing law, except possibly a threatened unlawful refusal to deal used to coerce an unwitting firm to permit Amazon to use its proprietary data. Emulating other firms and trying to build a better mousetrap, which the Report characterizes as “free rid[ing],” are central to the kind of competition the antitrust laws are intended to promote. The Report discusses none of this.
III. RECOMMENDATIONS
The Report recommends increased funding for the enforcement agencies. That seems clearly correct. There is ample reason to think that the antitrust laws should be more aggressively enforced. Industry concentration is increasing, as is market concentration in at least some sectors. The large digital platforms are growing, seemingly inexorably, and appear to be beneficiaries of boundless economies of scale and scope. Merger retrospectives suggest that too many permitted mergers have resulted in price increases or other indicia of increased market power and that expected merger efficiencies are usually not realized. The case law appears to have tilted too far toward avoiding false positives, and a number of important recent cases (including American Express and Qualcomm) found anticompetitive conduct to be lawful.
The Report’s other recommendations, however, are less persuasive. One reason is that the Report does not carefully assess existing law and thus does not make clear what problems the recommendations are intended to address. The Report could have used its study of the conduct of major digital platforms as a context in which to assess the antitrust laws. It could, for example, have identified the antitrust issues raised by the facts it assembled regarding acquisitions by the platforms or alleged self-preferencing by the platforms. These issues include how to evaluate mergers before one can know whether a nascent competitor will succeed absent the merger, how to police self-preferencing, how to make a non-discrimination requirement meaningful with respect to sales transactions when a sale to a corporate affiliate just moves money from one corporate pocket to another, and so on. And, after identifying the issues, it could have asked whether existing antitrust law is up to the task and, if not, how it might be improved.
The Report does none of that. Nor does it discuss the costs and benefits of its recommendations. The recommendations seem to rest on the notion that the dominant digital platforms are big, aggressive and growing, and that laws intended to reign them in are therefore desirable. The issues are more complicated. This can be seen by considering some of the recommendations that appear intended to address the kinds of conduct discussed above.
A. Merger Restrictions
The Report recommends prohibiting all “acquisitions of potential rivals and nascent competitors.” The Report states that Facebook, Google, and Amazon have acquired more than 423 companies. The vast majority would probably qualify as potential or nascent competitors and were probably firms that would not have become important competitors absent the acquisition, and whose acquisition gave the acquiring firm valuable human and technical resources that improved their products. It is likely that some of the acquisitions prevented the acquired assets, whether separately owned or after acquisition by a third party, from growing into an important competitive force whose presence would have increased economic welfare.
The policy issue is whether existing merger law is capable of identifying the problematic mergers and, if not, what changes would on balance improve the law. A flat prohibition on all acquisitions of potential and nascent competitors would be likely ex post to prevent a multitude of procompetitive transactions and ex ante, by cutting off a valuable exit strategy for venture capital investments, to deter some valuable investments and resulting innovations. The Report makes no effort to assess the costs and benefits of its proposed prohibition.
Nor does it discuss narrower alternatives. One would lower the pre-merger reporting threshold, so that more acquisitions of start-ups would be subject to pre-merger review by the enforcement agencies. As the Instagram story makes clear, however, pre-merger review does not solve the very difficult problem of determining whether a merger of a small and growing firm will turn out to be good or bad.
One aspect of the Report’s recommendation might itself make sense as a narrower alternative. The recommendation would prohibit mergers without proof “that the potential or nascent competitor would have been a successful entrant in a but-for world.” It is generally thought that current law prohibits only those mergers that are more likely than not to harm competition. One implication is that acquisition of a nascent competitor will be lawful if the acquired firm would have been unlikely to grow into an important competitive force absent the merger, even if the acquisition offered insubstantial efficiency benefits and would extinguish a small but realistic possibility that the acquired firm would otherwise develop into the Next Big Thing. Perhaps merger law should be revised or understood to prohibit mergers whose expected value – taking into account magnitudes and likelihoods – is negative even if a positive outcome is more likely than not. Regrettably, the Report did not address this possibility.
B. Monopolization Law
The Report does not make a specific recommendation about agreements to obtain default placement of digital applications, but it does recommend a number of changes intended to “rehabilitate monopolization law.” Current law rather clearly holds that conduct that tends to create or maintain monopoly power and has no efficiency benefits is illegal and that conduct that does not tend to create or maintain monopoly power is not illegal. The hard question arises when the conduct does both, which might be the case with the agreements that make Google’s search engine the default on the Safari (Apple) and Firefox browsers. Courts usually avoid that question by finding either no harm or no benefit or by finding a less restrictive alternative, but those solutions will not always be available.
There are three basic alternatives for dealing with conduct that both excludes rivals and creates efficiencies. One, implicit in the Report, is to ignore the efficiencies and find the conduct to be illegal. That alternative could be very costly for welfare and might deter important innovations. Another, which seems implicit in some cases, is to ignore the harm to competition and find the conduct to be legal. That alternative risks permitting substantial harm to competition for what might be trivial efficiencies.
The middle ground would try to take account of the competing effects. Ad hoc balancing is optimal in theory but often impossible in practice. How, for example, does one compare a static harm to competition against the kind of dynamic benefits from permitting exploitation of monopoly power that the Court emphasized in Trinko? Scholars have suggested variations, like enabling the decision maker to paint with a broader brush by condemning only conduct that causes harm that is “disproportionate” to the efficiency benefits. Google’s default agreements might be condemned under that approach on the ground that they tie up 87 percent of browser distribution and, unless switching is really trivial, all but ensure the perpetuation of Google’s search monopoly. Alternatively, the law might condemn only conduct whose efficiency benefits for the defendant are less than the cost of the conduct to the defendant and thus makes no business sense as a strategy for realizing efficiencies. This approach might find Google’s agreements to be lawful if the price Google paid for them was less than the anticipated value of the additional data they could reasonably have been expected to provide to Google. One can find support for each of these alternatives in the cases.
The Report could have made a valuable contribution if it had addressed the issues raised by conduct that both excludes rivals and creates efficiencies, either in general or with respect to the dominant digital platforms in particular. Instead, it ignored the efficiency part of the story and elided the critical antitrust question.
C. Structural Separations and Nondiscrimination
The Report recommends consideration of a number of reforms to address problems that can arise when a dominant platform competes in complementary businesses with other firms that need access to the platform. The reforms include “structural separations” to “prohibit a dominant intermediary [like Amazon] from operating in markets that place the intermediary in competition with the firms dependent on its infrastructure,” non-discrimination rules to prevent self-preferencing by the platforms, and prohibiting the platforms from extracting “greater money or data than users would be willing to provide in a competitive market.” All are intended to prevent the platforms from disadvantaging complementary businesses operated by third parties that need access to the platforms in order to benefit competing complementary businesses owned by the platforms.
The Report repeatedly says that dominant platforms “can” harm competition against its affiliated complementary business, but it does not assess the frequency of those harms. It seems to assume, without discussion, that harm of that sort is commonplace. Economic analysis suggests, however, that a dominant platform will have an incentive to engage in such handicapping only in some instances, most of which would entail a likely creation or preservation of market power in the complementary business. In other instances, a platform like Amazon will not have an incentive to handicap complementary businesses in order to aid its own complementary business because doing so would reduce the value of the platform itself.
Nor does the Report address the costs and benefits of the rules it proposes. Those rules could be beneficial to the extent that they preserve valuable competition and incentives for investment and innovation in lines of business that might be dependent on access to the platform. But they could also have large costs. These would include static costs, like preventing the efficiencies resulting from Amazon’s use of data to improve products, and dynamic costs, which could include preventing important innovations by the platform in both the complementary business and the platform itself. As to the latter, the Report recommends without explanation prohibiting any “design change that excludes competitors or otherwise undermines competition . . ., regardless of whether the design change can be justified as an improvement for consumers” (emphasis added).
The Report says that a simple separation requirement has the benefit, compared to rules that would permit vertical expansion and prohibit anticompetitive conduct by the platform directed at competitors, of being “easier to administer than conduct remedies, which can require close and continuous monitoring.” The benefit is, however, overstated because a separation requirement needs an accompanying non-discrimination rule to prevent the platform from engaging in the same anticompetitive conduct as a result of, or pursuant to, contractual arrangements with complementary businesses owned by third parties. A separation requirement also needs ongoing monitoring to address boundary issues that arise as industries and technologies evolve over time. The Report does not address these complexities.
IV. CONCLUSION
The Report’s first recommendation is that Congress consider making clear that the antitrust laws “are designed to protect not just consumers, but also workers, entrepreneurs, independent businesses, open markets, a fair economy, and democratic ideals.” Using the antitrust laws to address all these worthy objectives is problematic, in part because any law that has multiple and often conflicting objectives risks arbitrary and unpredictable decisions and is more susceptible to regulatory capture than a law whose enforcement and judicial decisions can be assessed by an unambiguous metric. These risks seem especially serious with antitrust law because it applies to almost all sectors of the economy, is enforced in a decentralized system by a multitude of potential plaintiffs, and is therefore applied to divers and often complex commercial problems.
The recommendation is problematic for another reason. All of the proposed objectives are distributional in nature. The recommendation implicitly sees the issues raised by the platforms, and perhaps antitrust law in general, as issues of the powerful harming the powerless and recommends a host of changes to redress that imbalance. There is in that recommendation, and indeed in the Report, almost no attention to economic welfare as a whole. For all practical purposes, the Report ignores considerations of economic efficiency.
The antitrust laws are presently focused solely on economic welfare. Adding additional objectives matters only to the extent that doing so would result in decisions different from those focused on economic welfare. Those decisions are likely to reduce economic welfare. The recommendation of adding objectives, and indeed antitrust law in general, cannot be assessed without discussing economic welfare and efficiency.
1 Professor of the Practice of Law, Stanford Law School.
