Showing 9 of 58 PublicationsScholarship

A Choice-of-Law Alternative to Federal Preemption of State Privacy Law

Scholarship Executive Summary A prominent theme in debates about US national privacy legislation is whether federal law should preempt state law. A federal statute could create . . .

Executive Summary

A prominent theme in debates about US national privacy legislation is whether federal law should preempt state law. A federal statute could create one standard for markets that are obviously national in scope. Another approach is to allow states to be “laboratories of democracy” that adopt different laws so they can discover the best ones.

We propose a federal statute requiring states to recognize contractual choice-of-law provisions, so companies and consumers can choose what state privacy law to adopt. Privacy would continue to be regulated at the state level. However, the federal government would provide for jurisdictional competition among states, such that companies operating nationally could comply with the privacy laws of any one state.

Our proposed approach would foster a double competition aimed at discerning and delivering on consumers’ true privacy interests: market competition to deliver privacy policies that consumers prefer and competition among states to develop the best privacy laws.

Unlike a single federal privacy law, this approach would provide 50 competing privacy regimes for national firms. The choice-of-law approach can trigger competition and innovation in privacy practices while preserving a role for meaningful state privacy regulation.

Introduction

The question of preemption of state law by the federal government has bedeviled debates about privacy regulation in the United States. A prominent theme is to propose a national privacy policy that largely preempts state policies to create one standard for markets that are obviously national. Another approach is to allow states to be “laboratories of democracy” that adopt different laws, with the hope that they will adopt the best rules over time. Both approaches have substantial costs and weaknesses.

The alternative approach we propose would foster a double competition aimed at discerning and delivering on consumers’ true privacy interests: market competition to deliver privacy policies that consumers prefer and competition among states to develop the best privacy laws. Indeed, our proposal aims to obtain the best features—and avoid the worst features—of both a federal regime and a multistate privacy law regime by allowing firms and consumers to agree on compliance with the single regime of their choosing.

Thus, we propose a federal statute requiring states to recognize contractual choice-of-law provisions, so companies and consumers can choose what state privacy law to adopt. Privacy would continue to be regulated at the state level. However, the federal government would provide for jurisdictional competition among states, and companies operating nationally could comply with the privacy laws of any one state.

Unlike a single federal privacy law, this approach would provide 50 competing privacy regimes for national firms. Protecting choice of law can trigger competition and innovation in privacy practices while preserving a role for meaningful state privacy regulation.

The Emerging Patchwork of State Privacy Statutes Is a Problem for National Businesses

A strong impetus for federal privacy legislation is the opportunity national and multinational businesses see to alleviate the expense and liability of having a patchwork of privacy statutes with which they must comply in the United States. Absent preemptive legislation, they could conceivably operate under 50 different state regimes, which would increase costs and balkanize their services and policies without coordinate gains for consumers. Along with whether a federal statute should have a private cause of action, preempting state law is a top issue when policymakers roll up their sleeves and discuss federal privacy legislation.

But while the patchwork argument is real, it may be overstated. There are unlikely ever to be 50 distinct state regimes; rather, a small number of state legislation types is likely, as jurisdictions follow each other’s leads and group together, including by promulgating model state statutes.[1] States don’t follow the worst examples from their brethren, as the lack of biometric statutes modeled on Illinois’s legislation illustrates.[2]

Along with fewer “patches,” the patchwork’s costs will tend to diminish over time as states land on relatively stable policies, allowing compliance to be somewhat routinized.

Nonetheless, the patchwork is far from ideal. It is costly to firms doing business nationally. It costs small firms more per unit of revenue, raising the bar to new entry and competition. And it may confuse consumers about what their protections are (though consumers don’t generally assess privacy policies carefully anyway).

But a Federal Privacy Statute Is Far from Ideal as Well

Federal preemption has many weaknesses and costs as well. Foremost, it may not deliver meaningful privacy to consumers. This is partially because “privacy” is a congeries of interests and values that defy capture.[3] Different people prioritize different privacy issues differently. In particular, the elites driving and influencing legislation may prioritize certain privacy values differently from consumers, so legislation may not serve most consumers’ actual interests.[4]

Those in the privacy-regulation community sometimes assume that passing privacy legislation ipso facto protects privacy, but that is not a foregone conclusion. The privacy regulations issued under the Gramm-Leach-Bliley Act (concerning financial services)[5] and the Health Insurance Portability and Accountability Act (concerning health care)[6] did not usher in eras of consumer confidence about privacy in their respective fields.

The short-term benefits of preempting state law may come with greater long-term costs. One cost is the likely drop in competition among firms around privacy. Today, as some have noted, “Privacy is actually a commercial advantage. . . . It can be a competitive advantage for you and build trust for your users.”[7] But federal privacy regulation seems almost certain to induce firms to treat compliance as the full measure of privacy to offer consumers. Efforts to outperform or ace out one another will likely diminish.[8]

Another long-term cost of preempting state law is the drop in competition among states to provide well-tuned privacy and consumer-protection legislation. Our federal system’s practical genius, which Justice Louis Brandeis articulated 90 years ago in New State Ice v. Liebmann, is that state variation allows natural experiments in what best serves society—business and consumer interests alike.[9] Because variations are allowed, states can amend their laws individually, learn from one another, adapt, and converge on good policy.

The economic theory of federalism draws heavily from the Tiebout model.[10] Charles Tiebout argued that competing local governments could, under certain conditions, produce public goods more efficiently than the national government could. Local governments act as firms in a marketplace for taxes and public goods, and consumer-citizens match their preferences to the providers. Efficient allocation requires mobile people and resources, enough jurisdictions with the freedom to set their own laws, and limited spillovers among jurisdictions (effects of one jurisdiction’s policies on others).

A related body of literature on “market-preserving federalism” argues that strong and self-reinforcing limits on national and local power can preserve markets and incentivize economic growth and development.[11] The upshot of this literature is that when local jurisdictions can compete on law, not only do they better match citizens’ policy preferences, but the rules tend toward greater economic efficiency.

In contrast to the economic gains from decentralization, moving authority over privacy from states to the federal government may have large political costs. It may deepen Americans’ growing dissatisfaction with their democracy. Experience belies the ideal of responsive national government when consumers, acting as citizens, want to learn about or influence the legislation and regulation that governs more and more areas of their lives. The “rejectionist” strain in American politics that Donald Trump’s insurgency and presidency epitomized may illustrate deep dissatisfaction with American democracy that has been growing for decades. Managing a highly personal and cultural

issue like privacy through negotiation between large businesses and anonymous federal regulators would deepen trends that probably undermine the government’s legitimacy.

To put a constitutional point on it, preempting states on privacy contradicts the original design of our system, which assigned limited powers to the federal government.[12] The federal government’s enumerated powers generally consist of national public goods—particularly defense. The interstate commerce clause, inspired by state parochialism under the Articles of Confederation, exists to make commerce among states (and with tribes) regular; it is not rightly a font of power to regulate the terms and conditions of commerce generally.[13]

Preempting state law does not necessarily lead to regulatory certainty, as is often imagined. Section 230 of the Communications Decency Act may defeat once and for all the idea that federal legislation creates certainty.[14] More than a quarter century after its passage, it is hotly debated in Congress and threatened in the courts.[15]

The Fair Credit Reporting Act (FCRA) provides a similar example.[16] Passed in 1970, it comprehensively regulated credit reporting. Since then, Congress has amended it dozens of times, and regulators have made countless alterations through interpretation and enforcement.[17] The Consumer Financial Protection Bureau recently announced a new inquiry into data brokering under the FCRA.[18] That is fine, but it illustrates that the FCRA did not solve problems and stabilize the law. It just moved the jurisdiction to Washington, DC.

Meanwhile, as regulatory theory predicts, credit reporting has become a three-horse race.[19] A few slow-to-innovate firms have captured and maintained dominance thanks partially to the costs and barriers to entry that uniform regulation creates.

Legal certainty may be a chimera while business practices and social values are in flux. Certainty develops over time as industries settle into familiar behaviors and roles.

An Alternative to Preemption: Business and Consumer Choice

One way to deal with this highly complex issue is to promote competition for laws. The late, great Larry Ribstein, with several coauthors over the years, proposed one such legal mechanism: a law market empowered by choice-of-law statutes.[20] Drawing on the notion of market competition as a discovery process,[21] Ribstein and Henry Butler explained:

In order to solve the knowledge problem and to create efficient legal technologies, the legal system can use the same competitive process that encourages innovation in the private sector—that is, competition among suppliers of law. As we will see, this entails enforcing contracts among the parties regarding the applicable law. The greater the knowledge problem the more necessary it is to unleash markets for law to solve the problem.[22]

The proposal set forth below promotes just such competition and solves the privacy-law patchwork problem without the costs of federal preemption. It does this through a simple procedural regulation requiring states to enforce choice-of-law terms in privacy contracts, rather than through a heavy-handed, substantive federal law. Inspired by Butler and Ribstein’s proposal for pluralist insurance regulation,[23] the idea is to make the choice of legal regime a locus of privacy competition.

Modeled on the US system of state incorporation law, our proposed legislation would leave firms generally free to select the state privacy law under which they do business nationally. Firms would inform consumers, as they must to form a contract, that a given state’s laws govern their policies. Federal law would ensure that states respect those choice-of-law provisions, which would be enforced like any other contract term.

This would strengthen and deepen competition around privacy. If firms believed privacy was a consumer interest, they could select highly protective state laws and advertise that choice, currying consumer favor. If their competitors chose relatively lax state law, they could advertise to the public the privacy threats behind that choice. The process would help hunt out consumers’ true interests through an ongoing argument before consumers. Businesses’ and consumers’ ongoing choices— rather than a single choice by Congress followed by blunt, episodic amendments—would shape the privacy landscape.

The way consumers choose in the modern marketplace is a broad and important topic that deserves further study and elucidation. It nevertheless seems clear—and it is rather pat to observe—that consumers do not carefully read privacy policies and balance their implications. Rather, a hive mind of actors including competitors, advocates, journalists, regulators, and politicians pore over company policies and practices. Consumers take in branding and advertising, reputation, news, personal recommendations, rumors, and trends to decide on the services they use and how they use them.

That detail should not be overlooked: Consumers may use services differently based on the trust they place in them to protect privacy and related values. Using an information-intensive service is not a proposition to share everything or nothing. Consumers can and do shade their use and withhold information from platforms and services depending on their perceptions of whether the privacy protections offered meet their needs.

There is reason to be dissatisfied with the modern marketplace, in which terms of service and privacy policies are offered to the individual consumer on a “take it or leave it” basis. There is a different kind of negotiation, described above, between the hive mind and large businesses. But when the hive mind and business have settled on terms, individuals cannot negotiate bespoke policies reflecting their particular wants and needs. This collective decision-making may be why some advocates regard market processes as coercive. They do not offer custom choices to all but force individual consumers into channels cut by all.

The solution that orthodox privacy advocates offer does not respond well to this problem, because they would replace “take it or leave it” policies crafted in the crucible of the marketplace with “take it or leave it” policies crafted in a political and regulatory crucible. Their prescriptions are sometimes to require artificial notice and “choice,” such as whether to accept cookies when one visits websites. This, as experience shows, does not reach consumers when they are interested in choosing.

Choice of law in privacy competition is meant to preserve manifold choices when and where consumers make their choices, such as at the decision to transact, and then let consumers choose how they use the services they have decided to adopt. Let new entrants choose variegated privacy-law regimes, and consumers will choose among them. That does not fix the whole problem, but at least it doesn’t replace consumer choice with an “expert” one-size-fits-all choice.

In parallel to business competition around privacy choice of law, states would compete with one another to provide the most felicitous environment for consumers and businesses. Some states would choose more protection, seeking the rules businesses would choose to please privacy-conscious consumers. Others might choose less protection, betting that consumers prefer goods other than information control, such as free, convenient, highly interactive, and custom services.

Importantly, this mechanism would allow companies to opt in to various privacy regimes based on the type of service they offer, enabling a degree of fine-tuning appropriate for different industries and different activities that no alternative would likely offer. This would not only result in the experimentation and competition of federalism but also enable multiple overlapping privacy-regulation regimes, avoiding the “one-size-doesn’t-fit-all” problem.

While experimentation continued, state policies would probably rationalize and converge over time. There are institutions dedicated to this, such as the Uniform Law Commission, which is at its best when it harmonizes existing laws based on states’ experience.[24]

It is well within the federal commerce power to regulate state enforcement of choice-of-law provisions, because states may use them to limit interjurisdictional competition. Controlling that is precisely what the commerce power is for. Utah’s recent Social Media Regulation Act[25] barred enforcement of choice-of-law provisions, an effort to regulate nationally from a state capital. Federally backing contractual choice-of-law selections would curtail this growing problem.

At the same time, what our proposed protections for choice-of-law rules do is not much different from what contracts already routinely do and courts enforce in many industries. Contracting parties often specify the governing state’s law and negotiate for the law that best suits their collective needs.

Indeed, sophisticated business contracts increasingly include choice-of-law clauses that state the law that the parties wish to govern their relationship. In addition to settling uncertainty, these clauses might enable the contracting parties to circumvent those states’ laws they deem to be undesirable.[26]

This practice is not only business-to-business. Consumers regularly enter into contracts that include choice-of-law clauses—including regarding privacy law. Credit card agreements, stock and mutual fund investment terms, consumer-product warranties, and insurance contracts, among many other legal agreements, routinely specify the relevant state law that will govern.

In these situations, the insurance company, manufacturer, or mutual fund has effectively chosen the law. The consumer participates in this choice only to the same extent that she participates in any choices related to mass-produced products and services, that is, by deciding whether to buy the product or service.[27]

Allowing contracting parties to create their own legal certainty by contract would likely rankle states. Indeed, “we might expect governments to respond with hostility to the enforcement of choice-of-law clauses. In fact, however, the courts usually do enforce choice-of-law clauses.”[28] With some states trying to regulate nationally and some effectively doing so, the choice the states collectively face is having a role in privacy regulation or no role at all. Competition is better for them than exclusion from the field or minimization of their role through federal preemption of state privacy law. This proposal thus advocates simple federal legislation that preserves firms’ ability to make binding choice-of-law decisions and states’ ability to retain a say in the country’s privacy-governance regime.

Avoiding a Race to the Bottom

Some privacy advocates may object that state laws will not sufficiently protect consumers.[29] Indeed, there is literature arguing that federalism will produce a race to the bottom (i.e., competition leading every state to effectively adopt the weakest law possible), for example, when states offer incorporation laws that are the least burdensome to business interests in a way that arguably diverges from public or consumer interests.[30]

The race-to-the-bottom framing slants the issues and obscures ever-present trade-offs, however. Rules that give consumers high levels of privacy come at a cost in social interaction, price, and the quality of the goods they buy and services they receive. It is not inherently “down” or bad to prefer cheap or free goods and plentiful, social, commercial interaction. It is not inherently “up” or good to opt for greater privacy.

The question is what consumers want. The answers to that question—yes, plural—are the subject of constant research through market mechanisms when markets are free to experiment and are functioning well. Consumers’ demands can change over time through various mechanisms, including experience with new technologies and business models. We argue for privacy on the terms consumers want. The goal is maximizing consumer welfare, which sometimes means privacy and sometimes means sharing personal information in the interest of other goods. There is no race to the bottom in trading one good for another.

Yet the notion of a race to the bottom persists—although not without controversy. In the case of Delaware’s incorporation statutes, the issue is highly contested. Many scholars argue that the state’s rules are the most efficient—that “far from exploiting shareholders, . . . these rules actually benefit shareholders by increasing the wealth of corporations chartered in states with these rules.”[31]

As always, there are trade-offs, and the race-to-the-bottom hypothesis requires some unlikely assumptions. Principally, as Jonathan Macey and Geoffrey Miller discuss, the assumption that state legislators are beholden to the interests of corporations over other constituencies vying for influence. As Macey and Miller explain, the presence of a powerful lobby of specialized and well-positioned corporate lawyers (whose interests are not the same as those of corporate managers) transforms the analysis and explains the persistence and quality of Delaware corporate law.[32]

In much the same vein, there are several reasons to think competition for privacy rules would not succumb to a race to the bottom.

First, if privacy advocates are correct, consumers put substantial pressure on companies to adopt stricter privacy policies. Simply opting in to the weakest state regime would not, as with corporate law, be a matter of substantial indifference to consumers but would (according to advocates) run contrary to their interests. If advocates are correct, firms avoiding stronger privacy laws would pay substantial costs. As a result, the impetus for states to offer weaker laws would be diminished. And, consistent with Macey and Miller’s “interest-group theory” of corporate law,[33] advocates themselves would be important constituencies vying to influence state privacy laws. Satisfying these advocates may benefit state legislators more than satisfying corporate constituencies does.

Second, “weaker” and “stronger” would not be the only dimensions on which states would compete for firms to adopt their privacy regimes. Rather, as mentioned above, privacy law is not one-size-fits-all. Different industries and services entail different implications for consumer interests. States could compete to specialize in offering privacy regimes attractive to distinct industries based on interest groups with particular importance to their economies. Minnesota (home of the Mayo Clinic) and Ohio (home of the Cleveland Clinic), for example, may specialize in health care and medical privacy, while California specializes in social media privacy.

Third, insurance companies are unlikely to be indifferent to the law that the companies they cover choose. Indeed, to the extent that insurers require covered firms to adopt specific privacy practices to control risk, those insurers would likely relish the prospect of outsourcing the oversight of these activities to state law enforcers. States could thus compete to mimic large insurers’ privacy preferences—which would by no means map onto “weaker” policies—to induce insurers to require covered firms to adopt their laws.

If a race to the bottom is truly a concern, the federal government could offer a 51st privacy alternative (that is, an optional federal regime as an alternative to the states’ various privacy laws). Assuming federal privacy regulation would be stricter (an assumption inherent in the race-to-the-bottom objection to state competition), such an approach would ensure that at least one sufficiently strong opt-in privacy regime would always be available. Among other things, this would preclude firms from claiming that no option offers a privacy regime stronger than those of the states trapped in the (alleged) race to the bottom.

Choice of law exists to a degree in the European Union, a trading bloc commonly regarded as uniformly regulated (and commonly regarded as superior on privacy because of a bias toward privacy over other goods). The General Data Protection Regulation (GDPR) gives EU member states broad authority to derogate from its provisions and create state-level exemptions. Article 23 of the GDPR allows states to exempt themselves from EU-wide law to safeguard nine listed broad governmental and public interests.[34] And Articles 85 through 91 provide for derogations, exemptions, and powers to impose additional requirements relative to the GDPR for a number of “specific data processing situations.”[35]

Finally, Article 56 establishes a “lead supervisory authority” for each business.[36] In the political, negotiated processes under the GDPR, this effectively allows companies to shade their regulatory obligations and enforcement outlook through their choices of location. For the United States’ sharper rule-of-law environment, we argue that the choice of law should be articulate and clear.

Refining the Privacy Choice-of-Law Proposal

The precise contours of a federal statute protecting choice-of-law terms in contracts will determine whether it successfully promotes interfirm and interstate competition. Language will also determine its political salability.

Questions include: What kind of notice, if any, should be required to make consumers aware that they are dealing with a firm under a law regime not their own? Consumers are notoriously unwilling to investigate privacy terms—or any other contract terms—in advance, and when considering the choice of law, they would probably not articulate it to themselves. But the competitive dynamics described earlier would probably communicate relevant information to consumers even without any required notice. As always, competitors will have an incentive to ensure consumers are appropriately well-informed when they can diminish their rivals or elevate themselves in comparison by doing so.[37]

Would there be limits on which state’s laws a firm could choose? For example, could a company choose the law of a state where neither the company nor the consumer is domiciled? States would certainly argue that a company should not be able to opt out of the law of the state where it is domiciled. The federal legislation we propose would allow unlimited choice. Such a choice is important if the true benefits of jurisdictional competition are to be realized.

A federal statute requiring states to enforce choice-of-law terms should not override state law denying enforcement of choice-of-law terms that are oppressive, unfair, or improperly bargained for. In cases such as Carnival Cruise Lines v. Shute[38] and The Bremen v. Zapata Off-Shore Co.,[39] the Supreme Court has considered whether forum-selection clauses in contracts might be invalid. The Court has generally upheld such clauses, but they can be oppressive if they require plaintiffs in Maine to litigate in Hawaii, for example, without a substantial reason why Hawaii courts are the appropriate forum. Choice-of-law terms do not impose the cost of travel to remote locations, but they could be used not to establish the law governing the parties but rather to create a strategic advantage unrelated to the law in litigation. Deception built into a contract’s choice-of-law terms should remain grounds for invalidating the contract under state law, even if the state is precluded from barring choice-of-law terms by statute.

The race-to-the-bottom argument raises the question of whether impeding states from overriding contractual choice-of-law provisions would be harmful to state interests, especially since privacy law concerns consumer rights. However, there are reasons to believe race-to-the-bottom incentives would be tempered by greater legal specialization and certainty and by state courts’ ability to refuse to enforce choice-of-law clauses in certain limited circumstances. As Erin O’Hara and Ribstein put it:

Choice-of law clauses reduce uncertainty about the parties’ legal rights and obligations and enable firms to operate in many places without being subject to multiple states’ laws. These reduced costs may increase the number of profitable transactions and thereby increase social wealth. Also, the clauses may not change the results of many cases because courts in states that prohibit a contract term might apply the more lenient law of a state that has close connections with the parties even without a choice-of-law clause.[40]

Determining when, exactly, a state court can refuse to enforce a firm’s choice of privacy law because of excessive leniency is tricky, but the federal statute could set out a framework for when a court could apply its own state’s law. Much like the independent federal alternative discussed above, specific minimum requirements in the federal law could ensure that any race to the bottom that does occur can go only so far. Of course, it would be essential that any such substantive federal requirements be strictly limited, or else the benefits of jurisdictional competition would be lost.

The converse to the problem of a race to the bottom resulting from state competition is the “California effect”—the prospect of states adopting onerous laws from which no company (or consumer) can opt out. States can regulate nationally through one small tendril of authority: the power to prevent businesses and consumers from agreeing on the law that governs their relationships. If a state regulates in a way that it thinks will be disfavored, it will bar choice-of-law provisions in contracts so consumers and businesses cannot exercise their preference.

Utah’s Social Media Regulation Act, for example, includes mandatory age verification for all social media users,[41] because companies must collect proof that consumers are either of age or not in Utah. To prevent consumers and businesses from avoiding this onerous requirement, Utah bars waivers of the law’s requirements “notwithstanding any contract or choice-of-law provision in a contract.”[42] If parties could choose their law, that would render Utah’s law irrelevant, so Utah cuts off that avenue. This demonstrates the value of a proposal like the one contemplated here.

Proposed Legislation

Creating a federal policy to stop national regulation coming from state capitols, while still preserving competition among states and firms, is unique. Congress usually creates its own policy and preempts states in that area to varying degrees. There is a well-developed law around this type of preemption, which is sometimes implied and sometimes expressed in statute.[43] Our proposal does not operate that way. It merely withdraws state authority to prevent parties from freely contracting about the law that applies to them.

A second minor challenge exists regarding the subject matter about which states may not regulate choice of law. Barring states from regulating choice of law entirely is an option, but if the focus is on privacy only, the preemption must be couched to allow regulation of choice of law in other areas. Thus, the scope of “privacy” must be in the language.

Finally, the withdrawal of state authority should probably be limited to positive enactments, such as statutes and regulations, leaving intact common-law practice related to choice-of-law provisions.[44] “Statute,” “enactment,” and “provision” are preferable in preemptive language to “law,” which is ambiguous.

These challenges, and possibly more, are tentatively addressed in the following first crack at statutory language, inspired by several preemptive federal statutes, including the Employee Retirement Income Security Act of 1974,[45] the Airline Deregulation Act,[46] the Federal Aviation Administration Authorization Act of 1994,[47] and the Federal Railroad Safety Act.[48]

A state, political subdivision of a state, or political authority of at least two states may not enact or enforce any statute, regulation, or other provision barring the adoption or application of any contractual choice-of-law provision to the extent it affects contract terms governing commercial collection, processing, security, or use of personal information.

Conclusion

This report introduces a statutory privacy framework centered on individual states and consistent with the United States’ constitutional design. But it safeguards companies from the challenge created by the intersection of that design and the development of modern commerce and communication, which may require them to navigate the complexities and inefficiencies of serving multiple regulators. It fosters an environment conducive to jurisdictional competition and experimentation.

We believe giving states the chance to compete under this approach should be explored in lieu of consolidating privacy law in the hands of one central federal regulator. Competition among states to provide optimal legislation and among businesses to provide optimal privacy policies will help discover and deliver on consumers’ interests, including privacy, of course, but also interactivity, convenience, low costs, and more.

Consumers’ diverse interests are not known now, and they cannot be predicted reliably for the undoubtedly interesting technological future. Thus, it is important to have a system for discovering consumers’ interests in privacy and the regulatory environments that best help businesses serve consumers. It is unlikely that a federal regulatory regime can do these things. The federal government could offer a 51st option in such a system, of course, so advocates for federal involvement could see their approach tested alongside the states’ approaches.

[1] See Uniform Law Commission, “What Is a Model Act?,” https://www.uniformlaws.org/acts/overview/modelacts.

[2] 740 Ill. Comp. Stat. 14/15 (2008).

[3] See Jim Harper, Privacy and the Four Categories of Information Technology, American Enterprise Institute, May 26, 2020, https://www.aei.org/research-products/report/privacy-and-the-four-categories-of-information-technology.

[4] See Jim Harper, “What Do People Mean by ‘Privacy,’ and How Do They Prioritize Among Privacy Values? Preliminary Results,” American Enterprise Institute, March 18, 2022, https://www.aei.org/research-products/report/what-do-people-mean-by-privacy-and-how-do-they-prioritize-among-privacy-values-preliminary-results.

[5] Gramm-Leach-Bliley Act, 15 U.S.C. 6801, § 501 et seq.

[6] Health Insurance Portability and Accountability Act of 1996, Pub. L. No. 104-191, § 264.

[7] Estelle Masse, quoted in Ashleigh Hollowell, “Is Privacy Only for the Elite? Why Apple’s Approach Is a Marketing Advantage,” VentureBeat, October 18, 2022, https://venturebeat.com/security/is-privacy-only-for-the-elite-why-apples-approach-is-a-marketing-advantage.

[8] Competition among firms regarding privacy is common, particularly in digital markets. Notably, Apple has implemented stronger privacy protections than most of its competitors have, particularly with its App Tracking Transparency framework in 2021. See, for example, Brain X. Chen, “To Be Tracked or Not? Apple Is Now Giving Us the Choice,” New York Times, April 26, 2021, https://www.nytimes.com/2021/04/26/technology/personaltech/apple-app-tracking-transparency.html. For Apple, this approach is built into the design of its products and offers what it considers a competitive advantage: “Because Apple designs both the iPhone and processors that offer heavy-duty processing power at low energy usage, it’s best poised to offer an alternative vision to Android developer Google which has essentially built its business around internet services.” Kif Leswing, “Apple Is Turning Privacy into a Business Advantage, Not Just a Marketing Slogan,” CNBC, June 8, 2021, https://www.cnbc.com/2021/06/07/apple-is-turning-privacy-into-a-business-advantage.html. Apple has built a substantial marketing campaign around these privacy differentiators, including its ubiquitous “Privacy. That’s Apple.” slogan. See Apple, “Privacy,” https://www.apple.com/privacy. Similarly, “Some of the world’s biggest brands (including Unilever, AB InBev, Diageo, Ferrero, Ikea, L’Oréal, Mars, Mastercard, P&G, Shell, Unilever and Visa) are focusing on taking an ethical and privacy-centered approach to data, particularly in the digital marketing and advertising context.” Rachel Dulberg, “Why the World’s Biggest Brands Care About Privacy,” Medium, September 14, 2021, https://uxdesign.cc/who-cares-about-privacy-ed6d832156dd.

[9] New State Ice Co. v. Liebmann, 285 US 262, 311 (1932) (Brandeis, J., dissenting) (“To stay experimentation in things social and economic is a grave responsibility. Denial of the right to experiment may be fraught with serious consequences to the Nation. It is one of the happy incidents of the federal system that a single courageous State may, if its citizens choose, serve as a laboratory; and try novel social and economic experiments without risk to the rest of the country.”).

[10] See Charles M. Tiebout, “A Pure Theory of Local Expenditures,” Journal of Political Economy 64, no. 5 (1956): 416–24, https://www.jstor.org/stable/1826343.

[11] See, for example, Barry R. Weingast, “The Economic Role of Political Institutions: Market-Preserving Federalism and Economic Development,” Journal of Law, Economics, & Organization 11, no. 1 (April 1995): 1 31, https://www.jstor.org/stable/765068; Yingyi Qian and Barry R. Weingast, “Federalism as a Commitment to Preserving Market Incentives,” Journal of Economic Perspectives 11, no. 4 (Fall 1997): 83–92, https://www.jstor.org/stable/2138464; and Rui J. P. de Figueiredo Jr. and Barry R. Weingast, “Self-Enforcing Federalism,” Journal of Law, Economics, & Organization 21, no. 1 (April 2005): 103–35, https://www.jstor.org/stable/3554986.

[12] See US Const. art. I, § 8 (enumerating the powers of the federal Congress).

[13] See generally Randy E. Barnett, Restoring the Lost Constitution: The Presumption of Liberty (Princeton, NJ: Princeton University Press, 2014), 274–318.

[14] Protection for Private Blocking and Screening of Offensive Material, 47 U.S.C. 230.

[15] See Geoffrey A. Manne, Ben Sperry, and Kristian Stout, “Who Moderates the Moderators? A Law & Economics Approach to Holding Online Platforms Accountable Without Destroying the Internet,” Rutgers Computer & Technology Law Journal 49, no. 1 (2022): 39–53, https://laweconcenter.org/wp-content/uploads/2021/11/Stout-Article-Final.pdf (detailing some of the history of how Section 230 immunity expanded and differs from First Amendment protections); Meghan Anand et al., “All the Ways Congress Wants to Change Section 230,” Slate, August 30, 2023, https://slate.com/technology/2021/03/section-230 reform-legislative-tracker.html (tracking every proposal to amend or repeal Section 230); and Technology & Marketing Law Blog, website, https://blog.ericgoldman.org (tracking all Section 230 cases with commentary).

[16] Fair Credit Reporting Act, 15 U.S.C. § 1681 et seq.

[17] See US Federal Trade Commission, Fair Credit Reporting Act: 15 U.S.C. § 1681, May 2023, https://www.ftc.gov/system/files/ftc_gov/pdf/fcra-may2023-508.pdf (detailing changes to the Fair Credit Reporting Act and its regulations over time).

[18] US Federal Reserve System, Consumer Financial Protection Bureau, “CFPB Launches Inquiry into the Business Practices of Data Brokers,” press release, May 15, 2023, https://www.consumerfinance.gov/about-us/newsroom/cfpb-launches-inquiry-into-the-business-practices-of-data-brokers.

[19] US Federal Reserve System, Consumer Financial Protection Bureau, List of Consumer Reporting Companies, 2021, 8, https://files.consumerfinance.gov/f/documents/cfpb_consumer-reporting-companies-list_03-2021.pdf (noting there are “three big nationwide providers of consumer reports”).

[20] See, for example, Erin A. O’Hara and Larry E. Ribstein, The Law Market (Oxford, UK: Oxford University Press, 2009); Erin A. O’Hara O’Connor and Larry E. Ribstein, “Conflict of Laws and Choice of Law,” in Procedural Law and Economics, ed. Chris William Sanchirico (Northampton, MA: Edward Elgar Publishing, 2012), in Encyclopedia of Law and Economics, 2nd ed., ed. Gerrit De Geest (Northampton, MA: Edward Elgar Publishing, 2009); and Bruce H. Kobayashi and Larry E. Ribstein, eds., Economics of Federalism (Northampton, MA: Edward Elgar Publishing, 2007).

[21] See F. A. Hayek, “The Use of Knowledge in Society,” American Economic Review 35, no. 4 (September 1945): 519–30, https://www.jstor.org/stable/1809376?seq=12.

[22] Henry N. Butler and Larry E. Ribstein, “Legal Process for Fostering Innovation” (working paper, George Mason University, Antonin Scalia Law School, Fairfax, VA), 2, https://masonlec.org/site/rte_uploads/files/Butler-Ribstein-Entrepreneurship-LER.pdf.

[23] See Henry N. Butler and Larry E. Ribstein, “The Single-License Solution,” Regulation 31, no. 4 (Winter 2008–09): 36–42, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1345900.

[24] See Uniform Law Commission, “Acts Overview,” https://www.uniformlaws.org/acts/overview.

[25] Utah Code Ann. § 13-63-101 et seq. (2023).

[26] O’Hara and Ribstein, The Law Market, 5.

[27] O’Hara and Ribstein, The Law Market, 5.

[28] O’Hara and Ribstein, The Law Market, 5.

[29] See Christiano Lima-Strong, “The U.S.’s Sixth State Privacy Law Is Too ‘Weak,’ Advocates Say,” Washington Post, March 30, 2023, https://www.washingtonpost.com/politics/2023/03/30/uss-sixth-state-privacy-law-is-too-weak-advocates-say.

[30] See, for example, William L. Cary, “Federalism and Corporate Law: Reflections upon Delaware,” Yale Law Journal 83, no. 4 (March 1974): 663–705, https://openyls.law.yale.edu/bitstream/handle/20.500.13051/15589/33_83YaleLJ663_1973_1974_.pdf (arguing Delaware could export the costs of inefficiently lax regulation through the dominance of its incorporation statute).

[31] Jonathan R. Macey and Geoffrey P. Miller, “Toward an Interest-Group Theory of Delaware Corporate Law,” Texas Law Review 65, no. 3 (February 1987): 470, https://openyls.law.yale.edu/bitstream/handle/20.500.13051/1029/Toward_An_Interest_Group_Theory_of_Delaware_Corporate_Law.pdf. See also Daniel R. Fischel, “The ‘Race to the Bottom’ Revisited: Reflections on Recent Developments in Delaware’s Corporation Law,” Northwestern University Law Review 76, no. 6 (1982): 913–45, https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=2409&context=journal_articles.

[32] Macey and Miller, “Toward an Interest-Group Theory of Delaware Corporate Law.”

[33] Macey and Miller, “Toward an Interest-Group Theory of Delaware Corporate Law.”

[34] Commission Regulation 2016/679, General Data Protection Regulation art. 23.

[35] Commission Regulation 2016/679, General Data Protection Regulation art. 85–91.

[36] Commission Regulation 2016/679, General Data Protection Regulation art. 56.

[37] See the discussion in endnote 8.

[38] Carnival Cruise Lines v. Shute, 499 US 585 (1991).

[39] The Bremen v. Zapata, 407 US 1 (1972).

[40] O’Hara and Ribstein, The Law Market, 8.

[41] See Jim Harper, “Perspective: Utah’s Social Media Legislation May Fail, but It’s Still Good for America,” Deseret News, April 6, 2023, https://www.aei.org/op-eds/utahs-social-media-legislation-may-fail-but-its-still-good-for-america.

[42] Utah Code Ann. § 13-63-401 (2023).

[43] See Bryan L. Adkins, Alexander H. Pepper, and Jay B. Sykes, Federal Preemption: A Legal Primer, Congressional Research Service, May 18, 2023, https://sgp.fas.org/crs/misc/R45825.pdf.

[44] Congress should not interfere with interpretation of choice-of-law provisions. These issues are discussed in Tanya J. Monestier, “The Scope of Generic Choice of Law Clauses,” UC Davis Law Review 56, no. 3 (February 2023): 959–1018, https://digitalcommons.law.buffalo.edu/cgi/viewcontent.cgi?article=2148&context=journal_articles.

[45] Employee Retirement Income Security Act of 1974, 29 U.S.C. § 1144(a).

[46] Airline Deregulation Act, 49 U.S.C. § 41713(b).

[47] Federal Aviation Administration Authorization Act of 1994, 49 U.S.C. § 14501.

[48] Federal Railroad Safety Act, 49 U.S.C. § 20106.

Continue reading
Data Security & Privacy

From Data Myths to Data Reality: What Generative AI Can Tell Us About Competition Policy (and Vice Versa)

Scholarship I. Introduction It was once (and frequently) said that Google’s “data monopoly” was unassailable: “If ‘big data’ is the oil of the information economy, Google . . .

I. Introduction

It was once (and frequently) said that Google’s “data monopoly” was unassailable: “If ‘big data’ is the oil of the information economy, Google has Standard Oil-like monopoly dominance — and uses that control to maintain its dominant position.”[1] Similar epithets have been hurled at virtually all large online platforms, including Facebook (Meta), Amazon, and Uber.[2]

While some of these claims continue even today (for example, “big data” is a key component of the U.S. Justice Department’s (“DOJ”) Google Search and AdTech antitrust suits),[3] a shiny new data target has emerged in the form of generative artificial intelligence. The launch of ChatGPT in November 2022, as well as the advent of AI image-generation services like Midjourney and Dall-E, have dramatically expanded people’s conception of what is, and what might be, possible to achieve with generative AI technologies built on massive data sets.

While these services remain in the early stages of mainstream adoption and are in the throes of rapid, unpredictable technological evolution, they nevertheless already appear on the radar of competition policymakers around the world. Several antitrust enforcers appear to believe that, by acting now, they can avoid the “mistakes” that were purportedly made during the formative years of Web 2.0.[4] These mistakes, critics assert, include failing to appreciate the centrality of data in online markets, as well as letting mergers go unchecked and allowing early movers to entrench their market positions.[5] As Lina Khan, Chair of the FTC, put it: “we are still reeling from the concentration that resulted from Web 2.0, and we don’t want to repeat the mis-steps of the past with AI”.[6]

In that sense, the response from the competition-policy world is deeply troubling. Instead of engaging in critical self-assessment and adopting an appropriately restrained stance, the enforcement community appears to be chomping at the bit. Rather than assessing their prior assumptions based on the current technological moment, enforcers’ top priority appears to be figuring out how to deploy existing competition tools rapidly and almost reflexively to address the presumed competitive failures presented by generative AI.[7]

It is increasingly common for competition enforcers to argue that so-called “data network effects” serve not only to entrench incumbents in the markets where that data is collected, but also confer similar, self-reinforcing benefits in adjacent markets. Several enforcers have, for example, prevented large online platforms from acquiring smaller firms in adjacent markets, citing the risk that they could use their vast access to data to extend their dominance into these new markets.[8] They have also launched consultations to ascertain the role that data plays in AI competition. For instance, in an ongoing consultation, the European Commission asks: “What is the role of data and what are its relevant characteristics for the provision of generative AI systems and/or components, including AI models?”[9] Unsurprisingly, the U.S. Federal Trade Commission (“FTC”) has been bullish about the risks posed by incumbents’ access to data. In comments submitted to the U.S. Copyright Office, for example, the FTC argued that:

The rapid development and deployment of AI also poses potential risks to competition. The rising importance of AI to the economy may further lock in the market dominance of large incumbent technology firms. These powerful, vertically integrated incumbents control many of the inputs necessary for the effective development and deployment of AI tools, including cloud-based or local computing power and access to large stores of training data. These dominant technology companies may have the incentive to use their control over these inputs to unlawfully entrench their market positions in AI and related markets, including digital content markets.[10]

Against this backdrop, it stands to reason that the largest online platforms—including Alphabet, Meta, Apple, and Amazon — should have a meaningful advantage in the burgeoning markets for generative AI services. After all, it is widely recognized that data is an essential input for generative AI.[11] This competitive advantage should be all the more significant given that these firms have been at the forefront of AI technology for more than a decade. Over this period, Google’s DeepMind and AlphaGo and Meta’s have routinely made headlines.[12] Apple and Amazon also have vast experience with AI assistants, and all of these firms use AI technology throughout their platforms.[13]

Contrary to what one might expect, however, the tech giants have, to date, been unable to leverage their vast data troves to outcompete startups like OpenAI and Midjourney. At the time of writing, OpenAI’s ChatGPT appears to be, by far, the most successful chatbot[14], despite the fact that large tech platforms arguably have access to far more (and more up-to-date) data.

This article suggests there are important lessons to be learned from the current technological moment, if only enforcers would stop to reflect. The meteoric rise of consumer-facing AI services should offer competition enforcers and policymakers an opportunity for introspection. As we explain, the rapid emergence of generative AI technology may undercut many core assumptions of today’s competition-policy debates — the rueful after-effects of the purported failure of 20th-century antitrust to address the allegedly manifest harms of 21st-century technology. These include the notions that data advantages constitute barriers to entry and can be leveraged to project dominance into adjacent markets; that scale itself is a market failure to be addressed by enforcers; and that the use of consumer data is inherently harmful to those consumers.

II. Data Network Effects Theory and Enforcement

Proponents of tougher interventions by competition enforcers into digital markets often cite data network effects as a source of competitive advantage and barrier to entry (though terms like “economies of scale and scope” may offer more precision).[15] The crux of the argument is that “the collection and use of data creates a feedback loop of more data, which ultimately insulates incumbent platforms from entrants who, but for their data disadvantage, might offer a better product.”[16] This self-reinforcing cycle purportedly leads to market domination by a single firm. Thus, for Google, for example, it is argued that its “ever-expanding control of user personal data, and that data’s critical value to online advertisers, creates an insurmountable barrier to entry for new competition.”[17]

Right off the bat, it is important to note the conceptual problem of these claims. Because data is used to improve the quality of products and/or to subsidize their use, the idea of data as an entry barrier suggests that any product improvement or price reduction made by an incumbent could be a problematic entry barrier to any new entrant. This is tantamount to an argument that competition itself is a cognizable barrier to entry. Of course, it would be a curious approach to antitrust if this were treated as a problem, as it would imply that firms should under-compete — should forego consumer-welfare enhancements—in order to bring about a greater number of firms in a given market simply for its own sake.[18]

Meanwhile, actual economic studies of data network effects are few and far between, with scant empirical evidence to support the theory.[19] Andrei Hagiu and Julian Wright’s theoretical paper offers perhaps the most comprehensive treatment of the topic.[20] The authors ultimately conclude that data network effects can be of different magnitudes and have varying effects on firms’ incumbency advantage.[21] They cite Grammarly (an AI writing-assistance tool) as a potential example: “As users make corrections to the suggestions offered by Grammarly, its language experts and artificial intelligence can use this feedback to continue to improve its future recommendations for all users.”[22]

This is echoed by other economists who contend that “[t]he algorithmic analysis of user data and information might increase incumbency advantages, creating lock-in effects among users and making them more reluctant to join an entrant platform.”[23]

Crucially, some scholars take this logic a step further, arguing that platforms may use data from their “origin markets” in order to enter and dominate adjacent ones:

First, as we already mentioned, data collected in the origin market can be used, once the enveloper has entered the target market, to provide products more efficiently in the target market. Second, data collected in the origin market can be used to reduce the asymmetric information to which an entrant is typically subject when deciding to invest (for example, in R&D) to enter a new market. For instance, a search engine could be able to predict new trends from consumer searches and therefore face less uncertainty in product design.[24]

This possibility is also implicit in the paper by Hagiu and Wright.[25] Indeed, the authors’ theoretical model rests on an important distinction between within-user data advantages (that is, having access to more data about a given user) and across-user data advantages (information gleaned from having access to a wider user base). In both cases, there is an implicit assumption that platforms may use data from one service to gain an advantage in another market (because what matters is information about aggregate or individual user preferences, regardless of its origin).

Our review of the economic evidence suggests that several scholars have, with varying degrees of certainty, raised the possibility that incumbents may leverage data advantages to stifle competitors in their primary market or adjacent ones (be it via merger or organic growth). As we explain below, however, there is ultimately little evidence to support such claims.

Policymakers, however, have largely been receptive to these limited theoretical findings, basing multiple decisions on these theories, often with little consideration of the caveats that accompany them.[26] Indeed, it is remarkable that, in the Furman Report’s section on “[t]he data advantage for incumbents,” only two empirical economic studies are cited, and they offer directly contradictory conclusions with respect to the question of the strength of data advantages.[27] Nevertheless, the Furman Report concludes that data “may confer a form of unmatchable advantage on the incumbent business, making successful rivalry less likely,”[28] and adopts without reservation “convincing” evidence from non-economists with apparently no empirical basis.[29]

In the Google/Fitbit merger proceedings, the European Commission found that the combination of data from Google services with that of Fitbit devices would reduce competition in advertising markets:

Giving [sic] the large amount of data already used for advertising purposes that Google holds, the increase in Google’s data collection capabilities, which goes beyond the mere number of active users for which Fitbit has been collecting data so far, the Transaction is likely to have a negative impact on the development of an unfettered competition in the markets for online advertising.[30]

As a result, the Commission cleared the merger on the condition that Google refrain from using data from Fitbit devices for its advertising platform.[31] The Commission will likely focus on similar issues during its ongoing investigation into Microsoft’s investment into OpenAI.[32]

Along similar lines, the FTC’s complaint to enjoin Meta’s purchase of a virtual-reality (VR) fitness app called “Within” relied, among other things, on the fact that Meta could leverage its data about VR-user behavior to inform its decisions and potentially outcompete rival VR-fitness apps: “Meta’s control over the Quest platform also gives it unique access to VR user data, which it uses to inform strategic decisions.”[33]

The U.S. Department of Justice’s twin cases against Google also raise data leveraging and data barriers to entry. The agency’s AdTech complaint that “Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”[34] Similarly, in its Search complaint, the agency argues that:

Google’s anticompetitive practices are especially pernicious because they deny rivals scale to compete effectively. General search services, search advertising, and general search text advertising require complex algorithms that are constantly learning which organic results and ads best respond to user queries; the volume, variety, and velocity of data accelerates the automated learning of search and search advertising algorithms.[35]

Finally, the merger guidelines published by several competition enforcers cite the acquisition of data as a potential source of competitive concerns. For instance, the FTC and DOJ’s newly published guidelines state that “acquiring data that helps facilitate matching, sorting, or prediction services may enable the platform to weaken rival platforms by denying them that data.”[36] Likewise, the UK Competition and Markets Authority (“CMA”) warns against incumbents acquiring firms in order to obtain their data and foreclose other rivals:

Incentive to foreclose rivals…

7.19(e) Particularly in complex and dynamic markets, firms may not focus on short term margins but may pursue other objectives to maximise their long-run profitability, which the CMA may consider. This may include… obtaining access to customer data….[37]

In short, competition authorities around the globe are taking an aggressive stance on data network effects. Among the ways this has manifested is in basing enforcement decisions on fears that data collected by one platform might confer a decisive competitive advantage in adjacent markets. Unfortunately, these concerns rest on little to no empirical evidence, either in the economic literature or the underlying case records.

III. Data Incumbency Advantages in Generative AI Markets

Given the assertions canvassed in the previous section, it seems reasonable to assume that firms such as Google, Meta, and Amazon would be in pole position to dominate the burgeoning market for generative AI. After all, these firms have not only been at the forefront of the field for the better part of a decade, but they also have access to vast troves of data, the likes of which their rivals could only dream when they launched their own services. Thus the authors of the Furman Report caution that “to the degree that the next technological revolution centres around artificial intelligence and machine learning, then the companies most able to take advantage of it may well be the existing large companies because of the importance of data for the successful use of these tools.[38]

At the time of writing, however, this is not how things have unfolded — although it bears noting these markets remain in flux and the competitive landscape is susceptible to change. The first significantly successful generative AI service was arguably not from either Meta—which had been working on chatbots for years and had access to, arguably, the world’s largest database of actual chats—or Google. Instead, the breakthrough came from a previously unknown firm called OpenAI.

OpenAI’s ChatGPT service currently holds an estimated 60% of the market (though reliable numbers are somewhat elusive).[39] It broke the record for the fastest online service to reach 100 million users (in only a couple of months), more than four times faster than the previous record holder, TikTok.[40] Based on Google Trends data, ChatGPT is nine times more popular than Google’s own Bard service worldwide, and 14 times more popular in the U.S.[41] In April 2023, ChatGPT reportedly registered 206.7 million unique visitors, compared to 19.5 million for Google’s Bard.[42] In short, at the time of writing, ChatGPT appears to be the most popular chatbot. And, so far, the entry of large players such as Google Bard or Meta AI appear to have had little effect on its market position.[43]

The picture is similar in the field of AI image generation. As of August 2023, Midjourney, Dall-E, and Stable Diffusion appear to be the three market leaders in terms of user visits.[44] This is despite competition from the likes of Google and Meta, who arguably have access to unparalleled image and video databases by virtue of their primary platform activities.[45]

This raises several crucial questions: how have these AI upstarts managed to be so successful, and is their success just a flash in the pan before Web 2.0 giants catch up and overthrow them? While we cannot answer either of these questions dispositively, some observations concerning the role and value of data in digital markets would appear to be relevant.

A first important observation is that empirical studies suggest data exhibits diminishing marginal returns. In other words, past a certain point, acquiring more data does not confer a meaningful edge to the acquiring firm. As Catherine Tucker puts it, following a review of the literature: “Empirically there is little evidence of economies of scale and scope in digital data in the instances where one would expect to find them.”[46]

Likewise, following a survey of the empirical literature on this topic, Geoffrey Manne & Dirk Auer conclude that:

Available evidence suggests that claims of “extreme” returns to scale in the tech sector are greatly overblown. Not only are the largest expenditures of digital platforms unlikely to become proportionally less important as output increases, but empirical research strongly suggests that even data does not give rise to increasing returns to scale, despite routinely being cited as the source of this effect.[47]

In other words, being the firm with the most data appears to be far less important than having enough data, and this lower bar may be accessible to far more firms than one might initially think possible.

And obtaining enough data could become even easier — that is, the volume of required data could become even smaller — with technological progress. For instance, synthetic data may provide an adequate substitute to real-world data[48] — or may even outperform real-world data.[49] As Thibault Schrepel and Alex Pentland point out, “advances in computer science and analytics are making the amount of data less relevant every day. In recent months, important technological advances have allowed companies with small data sets to compete with larger ones.”[50]

Indeed, past a certain threshold, acquiring more data might not meaningfully improve a service, where other improvements (such as better training methods or data curation) could have a large effect. In fact, there is some evidence that excessive data impedes a service’s ability to generate results appropriate for a given query: “[S]uperior model performance can often be achieved with smaller, high-quality datasets than massive, uncurated ones. Data curation ensures that training datasets are devoid of noise, irrelevant instances, and duplications, thus maximizing the efficiency of every training iteration.”[51]

Consider, for instance, a user who wants to generate an image of a basketball. Using a model trained on an indiscriminate range and number of public photos in which a basketball appears, but is surrounded by copious other image data, the user may end up with an inordinately noisy result. By contrast, a model trained with a better method on fewer, more-carefully selected images, could readily yield far superior results.[52] In one important example,

[t]he model’s performance is particularly remarkable, given its small size. “This is not a large language model trained on the whole Internet; this is a relatively small transformer trained for these tasks,” says Armando Solar-Lezama, a computer scientist at the Massachusetts Institute of Technology, who was not involved in the new study…. The finding implies that instead of just shoving ever more training data into machine-learning models, a complementary strategy might be to offer AI algorithms the equivalent of a focused linguistics or algebra class.[53]

Current efforts are thus focused on improving the mathematical and logical reasoning of large language models (“LLMs”), rather than maximizing training datasets.[54] Two points stand out. The first is that firms like OpenAI rely largely on publicly available datasets — such as GSM8K — to train their LLMs.[55] Second, the real challenge to create cutting-edge AI is not so much in collecting data, but rather in creating innovative AI training processes and architectures:

[B]uilding a truly general reasoning engine will require a more fundamental architectural innovation. What’s needed is a way for language models to learn new abstractions that go beyond their training data and have these evolving abstractions influence the model’s choices as it explores the space of possible solutions.

We know this is possible because the human brain does it. But it might be a while before OpenAI, DeepMind, or anyone else figures out how to do it in silicon.[56]

Furthermore, it is worth noting that the data most relevant to startups operating in a given market may not be those data held by large incumbent platforms in other markets, but rather data specific to the market in which the startup is active or, even better, to the given problem it is attempting to solve:

As Andres Lerner has argued, if you wanted to start a travel business, the data from Kayak or Priceline would be far more relevant. Or if you wanted to start a ride-sharing business, data from cab companies would be more useful than the broad, market-cross-cutting profiles Google and Facebook have. Consider companies like Uber, Lyft and Sidecar that had no customer data when they began to challenge established cab companies that did possess such data. If data were really so significant, they could never have competed successfully. But Uber, Lyft and Sidecar have been able to effectively compete because they built products that users wanted to use — they came up with an idea for a better mousetrap. The data they have accrued came after they innovated, entered the market and mounted their successful challenges — not before.[57]

The bottom line is that data is not the be-all and end-all that many in competition circles rather casually make it out to be.[58] While data may often confer marginal benefits, there is little sense these are ultimately decisive.[59] As a result, incumbent platforms’ access to vast numbers of users and data in their primary markets might only marginally affect their AI competitiveness.

A related observation is that firms’ capabilities and other features of their products arguably play a more important role than the data they own.[60] Examples of this abound in digital markets. Google overthrew Yahoo, despite initially having access to far fewer users and far less data; Google and Apple overcame Microsoft in the smartphone OS market despite having comparatively tiny ecosystems (at the time) to leverage; and TikTok rose to prominence despite intense competition from incumbents like Instagram, which had much larger user bases. In each of these cases, important product-design decisions (such as the PageRank algorithm, recognizing the specific needs of mobile users,[61] and TikTok’s clever algorithm) appear to have played a far greater role than initial user and data endowments (or lack thereof).

All of this suggests that the early success of OpenAI likely has more to do with its engineering decisions than the data it did (or did not) own. And going forward, OpenAI and its rivals’ ability to offer and monetize compelling stores offering custom versions of their generative AI technology will arguably play a much larger role than (and contribute to) their ownership of data.[62] In other words, the ultimate challenge is arguably to create a valuable platform, of which data ownership is a consequence, but not a cause.

It is also important to note that, in those instances where it is valuable, data does not just fall from the sky. Instead, it is through smart business and engineering decisions that firms can generate valuable information (which does not necessarily correlate with owing more data).

For instance, OpenAI’s success with ChatGPT is often attributed to its more efficient algorithms and training models, which arguably have enabled the service to improve more rapidly than its rivals.[63] Likewise, the ability of firms like Meta and Google to generate valuable data for advertising arguably depends more on design decisions that elicit the right data from users, rather than the raw number of users in their networks.

Put differently, setting up a business so as to generate the right information is more important than simply owning vast troves of data.[64] Even in those instances where high-quality data is an essential parameter of competition, it does not follow that having vaster databases or more users on a platform necessarily leads to better information for the platform.

Given what precedes, it seems clear that OpenAI and other generative AI startups’ early success, as well as their chances of prevailing in the future, hinge on a far broader range of factors than the mere ownership of data. Indeed, if data ownership consistently conferred a significant competitive advantage, these new firms would not be where they are today. This does not mean that data is worthless, of course. Rather, it means that competition authorities should not assume that merely possessing data is a dispositive competitive advantage, absent compelling empirical evidence to support such a finding. In this light, the current wave of decisions and competition-policy pronouncements that rely on data-related theories of harm are premature.

IV. Five Key Takeaways: Reconceptualizing the Role of Data in Generative AI Competition

As we explain above, data (network effects) are not the source of barriers to entry that they are sometimes made out to be; rather, the picture is far more nuanced. Indeed, as economist Andres Lerner demonstrated almost a decade ago (and the assessment is only truer today):

Although the collection of user data is generally valuable for online providers, the conclusion that such benefits of user data lead to significant returns to scale and to the entrenchment of dominant online platforms is based on unsupported assumptions. Although, in theory, control of an “essential” input can lead to the exclusion of rivals, a careful analysis of real-world evidence indicates that such concerns are unwarranted for many online businesses that have been the focus of the “big data” debate.[65]

While data can be an important part of the competitive landscape, incumbent data advantages are far less pronounced than today’s policymakers commonly assume. In that respect, five main lessons emerge:

  1. Data can be (very) valuable, but past a certain threshold, the benefits tend to diminish. In other words, having the most data is less important than having enough;
  2. The ability to generate valuable information does not depend on the number of users or the amount of data a platform has previously acquired;
  3. The most important datasets are not always proprietary;
  4. Technological advances and platforms’ engineering decisions affect their ability to generate valuable information, and this effect swamps the effect of the amount of data they own; and
  5. How platforms use data is arguably more important than what data or how much data they own.

These lessons have important ramifications for competition-policy debates over the competitive implications of data in technologically evolving areas.

First, it is not surprising that startups, rather than incumbents, have taken an early lead in generative AI (and in Web 2.0 before it). After all, if data-incumbency advantages are small or even nonexistent, then smaller and more nimble players may have an edge over established tech platforms. This is all the more likely given that, despite significant efforts, the biggest tech platforms were unable to offer compelling generative AI chatbots and image-generation services before the emergence of ChatGPT, Dall-E, Midjourney, etc. This failure suggests that, in a process akin to Christensen’s Innovator’s Dilemma,[66] something about their existing services and capabilities was holding them back in those markets. Of course, this does not necessarily mean that those same services/capabilities could not become an advantage when the generative AI market starts addressing issues of monetization and scale.[67] But it does mean that assumptions of a firm’s market power based on its possession of data are off the mark.

Another important implication is that, paradoxically, policymakers’ efforts to prevent Web 2.0 platforms from competing freely in generative AI markets may ultimately backfire and lead to less, not more, competition. Indeed, OpenAI is currently acquiring a sizeable lead in generative AI. While competition authorities might like to think that other startups will emerge and thrive in this space, it is important not to confuse desires with reality. For, while there is a vibrant AI-startup ecosystem, there is at least a case to be made that the most significant competition for today’s AI leaders will come from incumbent Web 2.0 platforms — although nothing is certain at this stage. Policymakers should beware not to stifle that competition on the misguided assumption that competitive pressure from large incumbents is somehow less valuable to consumers than that which originates from smaller firms.

Finally, even if there were a competition-related market failure to be addressed (which is anything but clear) in the field of generative AI, it is unclear that contemplated remedies would do more good than harm. Some of the solutions that have been put forward have highly ambiguous effects on consumer welfare. Scholars have shown that mandated data sharing — a solution championed by EU policymakers, among others — may sometimes dampen competition in generative AI markets.[68] This is also true of legislation like the GDPR that make it harder for firms to acquire more data about consumers — assuming such data is, indeed, useful to generative AI services.[69]

In sum, it is a flawed understanding of the economics and practical consequences of large agglomerations of data that lead competition authorities to believe that data-incumbency advantages are likely to harm competition in generative AI markets — or even in the data-intensive Web 2.0 markets that preceded them. Indeed, competition or regulatory intervention to “correct” data barriers and data network and scale effects is liable to do more harm than good.

[1] Nathan Newman, Taking on Google’s Monopoly Means Regulating Its Control of User Data, Huffington Post (Sep. 24, 2013), http://www.huffingtonpost.com/nathan-newman/taking-on-googlesmonopol_b_3980799.html.

[2] See e.g. Lina Khan & K. Sabeel Rahman, Restoring Competition in the U.S. Economy, in Untamed: How to Check Corporate, Financial, and Monopoly Power (Nell Abernathy, Mike Konczal, & Kathryn Milani, eds., 2016), at 23 (“From Amazon to Google to Uber, there is a new form of economic power on display, distinct from conventional monopolies and oligopolies…, leverag[ing] data, algorithms, and internet-based technologies… in ways that could operate invisibly and anticompetitively.”); Mark Weinstein, I Changed My Mind — Facebook Is a Monopoly, Wall St. J. (Oct. 1, 2021), https://www.wsj.com/articles/facebook-is-monopoly-metaverse-users-advertising-platforms-competition-mewe-big-tech-11633104247 (“[T]he glue that holds it all together is Facebook’s monopoly over data…. Facebook’s data troves give it unrivaled knowledge about people, governments — and its competitors.”).

[3] See generally Abigail Slater, Why “Big Data” Is a Big Deal, The Reg. Rev. (Nov. 6, 2023), https://www.theregreview.org/2023/11/06/slater-why-big-data-is-a-big-deal/; Amended Complaint at ¶36, United States v. Google, 1:20-cv-03010- (D.D.C. 2020); Complaint at ¶37, United States v. Google, 1:23-cv-00108 (E.D. Va. 2023), https://www.justice.gov/opa/pr/justice-department-sues-google-monopolizing-digital-advertising-technologies (“Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”).

[4] See e.g. Press Release, European Commission, Commission Launches Calls for Contributions on Competition in Virtual Worlds and Generative AI (Jan. 9, 2024), https://ec.europa.eu/commission/presscorner/detail/en/IP_24_85; Krysten Crawford, FTC’s Lina Khan warns Big Tech over AI, SIEPR (Nov. 3, 2020), https://siepr.stanford.edu/news/ftcs-lina-khan-warns-big-tech-over-ai (“Federal Trade Commission Chair Lina Khan delivered a sharp warning to the technology industry in a speech at Stanford on Thursday: Antitrust enforcers are watching what you do in the race to profit from artificial intelligence.”) (emphasis added).

[5] See e.g. John M. Newman, Antitrust in Digital Markets, 72 Vand. L. Rev. 1497, 1501 (2019) (“[T]he status quo has frequently failed in this vital area, and it continues to do so with alarming regularity. The laissez-faire approach advocated for by scholars and adopted by courts and enforcers has allowed potentially massive harms to go unchecked.”);
Bertin Martins, Are New EU Data Market Regulations Coherent and Efficient?, Bruegel Working Paper 21/23 (2023), available at https://www.bruegel.org/working-paper/are-new-eu-data-market-regulations-coherent-and-efficient (“Technical restrictions on access to and re-use of data may result in failures in data markets and data-driven services markets.”); Valéria Faure-Muntian, Competitive Dysfunction: Why Competition Law Is Failing in a Digital World, The Forum Network (Feb. 24, 2021), https://www.oecd-forum.org/posts/competitive-dysfunction-why-competition-law-is-failing-in-a-digital-world.

[6] Rana Foroohar, The Great US-Europe Antitrust Divide, FT (Feb. 5, 2024), https://www.ft.com/content/065a2f93-dc1e-410c-ba9d-73c930cedc14.

[7] See e.g. Press Release, European Commission, supra note 5.

[8] See infra, Section II. Commentators have also made similar claims. See, e.g., Ganesh Sitaram & Tejas N. Narechania, It’s Time for the Government to Regulate AI. Here’s How, Politico (Jan. 15, 2024) (“All that cloud computing power is used to train foundation models by having them “learn” from incomprehensibly huge quantities of data. Unsurprisingly, the entities that own these massive computing resources are also the companies that dominate model development. Google has Bard, Meta has LLaMa. Amazon recently invested $4 billion into one of OpenAI’s leading competitors, Anthropic. And Microsoft has a 49 percent ownership stake in OpenAI — giving it extraordinary influence, as the recent board struggles over Sam Altman’s role as CEO showed.”).

[9] Press Release, European Commission, supra note 5.

[10] Comment of U.S. Federal Trade Commission to the U.S. Copyright Office, Artificial Intelligence and Copyright, Docket No. 2023-6 (Oct. 30, 2023) at 4, available at https://www.ftc.gov/legal-library/browse/advocacy-filings/comment-federal-trade-commission-artificial-intelligence-copyright (emphasis added).

[11] See, e.g. Joe Caserta, Holger Harreis, Kayvaun Rowshankish, Nikhil Srinidhi, and Asin Tavakoli, The data dividend: Fueling generative AI, McKinsey Digital (Sept. 15, 2023), https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-data-dividend-fueling-generative-ai (“Your data and its underlying foundations are the determining factors to what’s possible with generative AI.”).

[12] See e.g. Tim Keary, Google DeepMind’s Achievements and Breakthroughs in AI Research, Techopedia (Aug. 11, 2023), https://www.techopedia.com/google-deepminds-achievements-and-breakthroughs-in-ai-research; See e.g. Will Douglas Heaven, Google DeepMind used a large language model to solve an unsolved math problem, MIT Technology Review (Dec. 14, 2023), https://www.technologyreview.com/2023/12/14/1085318/google-deepmind-large-language-model-solve-unsolvable-math-problem-cap-set/; See also A Decade of Advancing the State-of-the-Art in AI Through Open Research, Meta (Nov. 30, 2023), https://about.fb.com/news/2023/11/decade-of-advancing-ai-through-open-research/; See also 200 languages within a single AI model: A breakthrough in high-quality machine translation, Meta, https://ai.meta.com/blog/nllb-200-high-quality-machine-translation/ (last visited Jan. 18, 2023).

[13] See e.g. Jennifer Allen, 10 years of Siri: the history of Apple’s voice assistant, Tech Radar (Oct. 4, 2021), https://www.techradar.com/news/siri-10-year-anniversary; see also Evan Selleck, How Apple is already using machine learning and AI in iOS, Apple Insider (Nov. 20, 2023), https://appleinsider.com/articles/23/09/02/how-apple-is-already-using-machine-learning-and-ai-in-ios; see also Kathleen Walch, The Twenty Year History Of AI At Amazon, Forbes (July 19, 2019), https://www.forbes.com/sites/cognitiveworld/2019/07/19/the-twenty-year-history-of-ai-at-amazon/?sh=1734bcb268d0.

[14] See infra Section III.

[15] See e.g. Cédric Argenton & Jens Prüfer, Search Engine Competition with Network Externalities, 8 J. Comp. L. & Econ. 73, 74 (2012); Mark A. Lemley & Matthew Wansley, Coopting Disruption (February 1, 2024), https://ssrn.com/abstract=4713845.

[16] John M. Yun, The Role of Big Data in Antitrust, in The Global Antitrust Institute Report on the Digital Economy (Joshua D. Wright & Douglas H. Ginsburg, eds., Nov. 11, 2020) at 233, available at https://gaidigitalreport.com/2020/08/25/big-data-and-barriers-to-entry/#_ftnref50. See also e.g. Robert Wayne Gregory, Ola Henfridsson, Evgeny Kaganer, & Harris Kyriakou, The Role of Artificial Intelligence and Data Network Effects for Creating User Value, 46 Acad. of Mgmt. Rev. 534 (2020), final pre-print version at 4, available at http://wrap.warwick.ac.uk/134220) (“A platform exhibits data network effects if, the more that the platform learns from the data it collects on users, the more valuable the platform becomes to each user.”). See also Karl Schmedders, José Parra-Moyano & Michael Wade, Why Data Aggregation Laws Could be the Answer to Big Tech Dominance, Silicon Republic (Feb. 6, 2024), https://www.siliconrepublic.com/enterprise/data-ai-aggregation-laws-regulation-big-tech-dominance-competition-antitrust-imd.

[17] Nathan Newman, Search, Antitrust, and the Economics of the Control of User Data, 31 Yale J. Reg. 401, 409 (2014) (emphasis added). See also id. at 420 & 423 (“While there are a number of network effects that come into play with Google, [“its intimate knowledge of its users contained in its vast databases of user personal data”] is likely the most important one in terms of entrenching the company’s monopoly in search advertising…. Google’s overwhelming control of user data… might make its dominance nearly unchallengeable.”).

[18] See also Yun, supra note 17 at 229 (“[I]nvestments in big data can create competitive distance between a firm and its rivals, including potential entrants, but this distance is the result of a competitive desire to improve one’s product.”).

[19] For a review of the literature on increasing returns to scale in data (this topic is broader than data network effects) see Geoffrey Manne & Dirk Auer, Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins, 28 Geo Mason L. Rev. 1281, 1344 (2021).

[20] Andrei Hagiu & Julian Wright, Data-Enabled Learning, Network Effects, and Competitive Advantage, 54 RAND J. Econ. 638 (2023) (final preprint available at https://andreihagiu.com/wp-content/uploads/2022/08/Data-enabled-learning-Final-RAND-Article.pdf).

[21] Id. at 2. The authors conclude that “Data-enabled learning would seem to give incumbent firms a competitive advantage. But how strong is this advantage and how does it differ from that obtained from more traditional mechanisms….”

[22] Id.

[23] Bruno Jullien & Wilfried Sand-Zantman, The Economics of Platforms: A Theory Guide for Competition Policy, 54 Info. Econ. & Pol’y 10080, 101031 (2021).

[24] Daniele Condorelli & Jorge Padilla, Harnessing Platform Envelopment in the Digital World, 16 J. Comp. L. & Pol’y 143, 167 (2020).

[25] See Hagiu & Wright, supra note 21.

[26] For a summary of these limitations, see generally Catherine Tucker, Network Effects and Market Power: What Have We Learned in the Last Decade?, Antitrust (Spring 2018) at 72, available at https://sites.bu.edu/tpri/files/2018/07/tucker-network-effects-antitrust2018.pdf. See also Manne & Auer, supra note 20, at 1330.

[27] See Jason Furman, Diane Coyle, Amelia Fletcher, Derek McAuley & Philip Marsden (Dig. Competition Expert Panel), Unlocking Digital Competition (2019) at 32-35 (“Furman Report”), available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/785547/unlocking_digital_competition_furman_review_web.pdf.

[28] Id. at 34.

[29] Id. at 35. To its credit, it should be noted, the Furman Report does counsel caution before mandating access to data as a remedy to promote competition. See id. at 75. That said, the Furman Report does maintain that such a remedy should certainly be on the table because “the evidence suggests that large data holdings are at the heart of the potential for some platform markets to be dominated by single players and for that dominance to be entrenched in a way that lessens the potential for competition for the market.” Id. In fact, the evidence does not show this.

[30] Case COMP/M.9660 — Google/Fitbit, Commission Decision (Dec. 17, 2020) (Summary at O.J. (C 194) 7), available at https://ec.europa.eu/competition/mergers/cases1/202120/m9660_3314_3.pdf at 455.

[31] Id. at 896.

[32] See Natasha Lomas, EU Checking if Microsoft’s OpenAI Investment Falls Under Merger Rules, TechCrunch (Jan. 9, 2024), https://techcrunch.com/2024/01/09/openai-microsoft-eu-merger-rules/.

[33] Amended Complaint at 11, Meta/Zuckerberg/Within, Fed. Trade Comm’n. (2022) (No. 605837), available at https://www.ftc.gov/system/files/ftc_gov/pdf/D09411%20-%20AMENDED%20COMPLAINT%20FILED%20BY%20COUNSEL%20SUPPORTING%20THE%20COMPLAINT%20-%20PUBLIC%20%281%29_0.pdf.

[34] Amended Complaint (D.D.C), supra note 4, at ¶37.

[35] Amended Complaint (E.D. Va), supra note 4, at ¶8.

[36] US Dep’t of Justice & Fed. Trade Comm’n, Merger Guidelines (2023) at 25, https://www.ftc.gov/system/files/ftc_gov/pdf/2023_merger_guidelines_final_12.18.2023.pdf.

[37] Competition and Mkts. Auth., Merger Assessment Guidelines (2021) at  ¶7.19(e), https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1051823/MAGs_for_publication_2021_–_.pdf.

[38] Furman Report, supra note 28, at ¶4.

[39] See e.g. Chris Westfall, New Research Shows ChatGPT Reigns Supreme in AI Tool Sector, Forbes (Nov. 16, 2023), https://www.forbes.com/sites/chriswestfall/2023/11/16/new-research-shows-chatgpt-reigns-supreme-in-ai-tool-sector/?sh=7de5de250e9c.

[40] See Krystal Hu, ChatGPT Sets Record for Fastest-Growing User Base, Reuters (Feb. 2, 2023), https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/; Google: The AI Race Is On, App Economy Insights (Feb. 7, 2023), https://www.appeconomyinsights.com/p/google-the-ai-race-is-on.

[41] See Google Trends, https://trends.google.com/trends/explore?date=today%205-y&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited, Jan. 12, 2024) and https://trends.google.com/trends/explore?date=today%205-y&geo=US&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited Jan. 12, 2024).

[42] See David F. Carr, As ChatGPT Growth Flattened in May, Google Bard Rose 187%, Similarweb Blog (June 5, 2023), https://www.similarweb.com/blog/insights/ai-news/chatgpt-bard/.

[43] See Press Release, Meta, Introducing New AI Experiences Across Our Family of Apps and Devices (Sept. 27, 2023), https://about.fb.com/news/2023/09/introducing-ai-powered-assistants-characters-and-creative-tools/; Sundar Pichai, An Important Next Step on Our AI Journey, Google Keyword Blog (Feb. 6, 2023), https://blog.google/technology/ai/bard-google-ai-search-updates/.

[44] See Ion Prodan, 14 Million Users: Midjourney’s Statistical Success, Yon (Aug. 19, 2023), https://yon.fun/midjourney-statistics/. See also Andrew Wilson, Midjourney Statistics: Users, Polls, & Growth [Oct 2023], ApproachableAI (Oct. 13, 2023), https://approachableai.com/midjourney-statistics/.

[45] See Hema Budaraju, New Ways to Get Inspired with Generative AI in Search, Google Keyword Blog (Oct. 12, 2023), https://blog.google/products/search/google-search-generative-ai-october-update/; Imagine with Meta AI, Meta (last visited Jan. 12, 2024), https://imagine.meta.com/.

[46] Catherine Tucker, Digital Data, Platforms and the Usual [Antitrust] Suspects: Network Effects, Switching Costs, Essential Facility, 54 Rev. Indus. Org. 683, 686 (2019).

[47] Manne & Auer, supra note 20, at 1345.

[48] See e.g. Stefanie Koperniak, Artificial Data Give the Same Results as Real Data—Without Compromising Privacy, MIT News (Mar. 3, 2017), https://news.mit.edu/2017/artificial-data-give-same-results-as-real-data-0303 (“[Authors] describe a machine learning system that automatically creates synthetic data—with the goal of enabling data science efforts that, due to a lack of access to real data, may have otherwise not left the ground. While the use of authentic data can cause significant privacy concerns, this synthetic data is completely different from that produced by real users—but can still be used to develop and test data science algorithms and models.”).

[49] See e.g. Rachel Gordon, Synthetic Imagery Sets New Bar in AI Training Efficiency, MIT News (Nov. 20, 2023), https://news.mit.edu/2023/synthetic-imagery-sets-new-bar-ai-training-efficiency-1120 (“By using synthetic images to train machine learning models, a team of scientists recently surpassed results obtained from traditional ‘real-image’ training methods.).

[50] Thibault Schrepel & Alex ‘Sandy’ Pentland, Competition Between AI Foundation Models: Dynamics and Policy Recommendations, MIT Connection Science Working Paper (Jun. 2023), at 8.

[51] Igor Susmelj, Optimizing Generative AI: The Role of Data Curation, Lightly (last visited Jan 15, 2024), https://www.lightly.ai/post/optimizing-generative-ai-the-role-of-data-curation.

[52] See e.g. Xiaoliang Dai, et al., Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack , ArXiv (Sep. 27, 2023) at 1, https://ar5iv.labs.arxiv.org/html/2309.15807 (“[S]upervised fine-tuning with a set of surprisingly small but extremely visually appealing images can significantly improve the generation quality.”). See also Hu Xu, et al., Demystifying CLIP Data, ArXiv (Sep. 28, 2023), https://arxiv.org/abs/2309.16671.

[53] Lauren Leffer, New Training Method Helps AI Generalize like People Do, Sci. Am. (Oct. 26, 2023), https://www.scientificamerican.com/article/new-training-method-helps-ai-generalize-like-people-do/ (discussing Brendan M. Lake & Marco Baroni, Human-Like Systematic Generalization Through a Meta-Learning Neural Network, 623 Nature 115 (2023)).

[54] Timothy B. Lee, The Real Research Behind the Wild Rumors about OpenAI’s Q* Project, Ars Technica (Dec. 8, 2023), https://arstechnica.com/ai/2023/12/the-real-research-behind-the-wild-rumors-about-openais-q-project/.

[55] Id. See also GSM8K, Papers with Code (last visited Jan. 18, 2023), available at https://paperswithcode.com/dataset/gsm8k; MATH Dataset, GitHub (last visited Jan. 18, 2024), available at https://github.com/hendrycks/math.

[56] Lee, supra note 55.

[57] Geoffrey Manne & Ben Sperry, Debunking the Myth of a Data Barrier to Entry for Online Services, Truth on the Market (Mar. 26, 2015), https://truthonthemarket.com/2015/03/26/debunking-the-myth-of-a-data-barrier-to-entry-for-online-services/ (citing Andres V. Lerner, The Role of ‘Big Data’ in Online Platform Competition (Aug. 26, 2014), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2482780.).

[58] See e.g., Lemley & Wansley, supra note 18, at 22 (“Incumbents have all that information. It would be difficult for a new entrant to acquire similar datasets independently….”).

[59] See Catherine Tucker, Digital Data as an Essential Facility: Control, CPI Antitrust Chron. (Feb. 2020) at 11 (“[U]ltimately the value of data is not the raw manifestation of the data itself, but the ability of a firm to use this data as an input to insight.”).

[60] Or, as John Yun puts it, data is only a small component of digital firms’ production function. See Yun, supra note 17, at 235 (“Second, while no one would seriously dispute that having more data is better than having less, the idea of a data-driven network effect is focused too narrowly on a single factor improving quality. As mentioned in supra Section I.A, there are a variety of factors that enter a firm’s production function to improve quality.”).

[61] Luxia Le, The Real Reason Windows Phone Failed Spectacularly, History–Computer (Aug. 8, 2023), https://history-computer.com/the-real-reason-windows-phone-failed-spectacularly/.

[62] Introducing the GPT Store, Open AI (Jan. 10, 2024), https://openai.com/blog/introducing-the-gpt-store.

[63] See Michael Schade, How ChatGPT and Our Language Models are Developed, OpenAI, https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-language-models-are-developed; Sreejani Bhattacharyya, Interesting innovations from OpenAI in 2021, AIM (Jan. 1, 2022), https://analyticsindiamag.com/interesting-innovations-from-openai-in-2021/; Danny Hernadez & Tom B. Brown, Measuring the Algorithmic Efficiency of Neural Networks, ArXiv (May 8, 2020), available at https://arxiv.org/abs/2005.04305.

[64] See Yun, supra note 17 at 235 (“Even if data is primarily responsible for a platform’s quality improvements, these improvements do not simply materialize with the presence of more data—which differentiates the idea of data-driven network effects from direct network effects. A firm needs to intentionally transform raw, collected data into something that provides analytical insights. This transformation involves costs including those associated with data storage, organization, and analytics, which moves the idea of collecting more data away from a strict network effect to more of a ‘data opportunity.’”).

[65] Lerner, supra note 58, at 4-5 (emphasis added).

[66] See Clayton M. Christensen, The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail (2013).

[67] See David J. Teece, Dynamic Capabilities and Strategic Management: Organizing for Innovation and Growth (2009).

[68] See Hagiu and Wright, supra note 21, at 4 (“We use our dynamic framework to explore how data sharing works: we find that it in-creases consumer surplus when one firm is sufficiently far ahead of the other by making the laggard more competitive, but it decreases consumer surplus when the firms are sufficiently evenly matched by making firms compete less aggressively, which in our model means subsidizing consumers less.”). See also Lerner, supra note 58.

[69] See e.g. Hagiu & Wright, id. (“We also use our model to highlight an unintended consequence of privacy policies. If such policies reduce the rate at which firms can extract useful data from consumers, they will tend to increase the incumbent’s competitive advantage, reflecting that the entrant has more scope for new learning and so is affected more by such a policy.”); Jian Jia, Ginger Zhe Jin & Liad Wagman, The Short-Run Effects of the General Data Protection Regulation on Technology Venture Investment, 40 Marketing Sci. 593 (2021) (finding GDPR reduced investment in new and emerging technology firms, particularly in data-related ventures); James Campbell, Avi Goldfarb, & Catherine Tucker, Privacy Regulation and Market Structure, 24 J. Econ. & Mgmt. Strat. 47 (2015) (“Consequently, rather than increasing competition, the nature of transaction costs implied by privacy regulation suggests that privacy regulation may be anti-competitive.”).

Continue reading
Antitrust & Consumer Protection

Gerrymandered Market Definitions in FTC v Amazon

ICLE Issue Brief Introduction Market definition is a critical component of any antitrust case. Not only does it narrow consideration to a limited range of relevant products or . . .

Introduction

Market definition is a critical component of any antitrust case. Not only does it narrow consideration to a limited range of relevant products or services but, perhaps more importantly, it specifies a domain of competition at issue in an antitrust case—that is, the nature of the competition between certain firms that might (or might not) be harmed by the conduct of the defendant. As Greg Werden has characterized it:

Alleging the relevant market in an antitrust case does not merely identify the portion of the economy most directly affected by the challenged conduct; it identifies the competitive process alleged to be harmed.[1]

Unsurprisingly, plaintiffs—not least, antitrust agencies—are often tempted to define artificially narrow markets in order to reinforce their cases (sometimes, downright ridiculously so[2]). The consequence is not merely to artificially inflate the market significance of the firm under scrutiny, although it does do that; it is also to misapprehend and misdescribe the true nature of competition relevant to the challenged conduct.

This unfortunate trend—allegations of harm to artificially constrained and gerrymandered markets—is exemplified in the Federal Trade Commission’s (FTC) recent proceedings against Amazon.

The FTC’s complaint against Amazon describes two relevant markets in which anticompetitive harm has allegedly occurred: (1) the “online superstore market” and (2) the “online marketplace services market.”[3]

Unfortunately, both markets are excessively narrow, thereby grossly inflating Amazon’s apparent market share and minimizing the true extent of competition. Moreover, the FTC’s approach to market definition here—lumping together wildly different products and wildly different sellers into single “cluster markets”—grossly misapprehends the nature of competition relating to the challenged conduct.

First, the FTC’s complaint limits the online-superstore market to online stores only, and further limits it to stores that have an “extensive breadth and depth”[4] of products. The latter means online stores that carry virtually all categories of products (“such as sporting goods, kitchen goods, apparel, and consumer electronics”[5]) and that also have an extensive variety of brands within each category (such as Nike, Under Armor, Adidas, etc.).[6] In practice, this definition excludes leading brands’ private channels (such as Nike’s online store),[7] as well as online stores that focus on a particular category of goods (such as Wayfair’s focus on furniture).[8] It also excludes the brick-and-mortar stores that still account for the vast majority of retail transactions.[9] Firms with significant online and brick-and-mortar sales might count, but only their online sales would be considered part of the market.

Second, the online-marketplace-services market is limited to online platforms that provide access to a “significant base of shoppers”;[10] a search function to identify products; a means for the seller to set prices and present product information; and a method to display customer reviews. This implies that current Amazon sellers can’t reach consumers through mechanisms that don’t incorporate all these specific functions, even though consumers regularly use multiple services and third-party sites that accomplish the same thing (e.g., Google Shopping, Shopify, Instagram, etc.)[11] Moreover, it implies that these myriad alternative channels do not constrain Amazon’s pricing of its services.

Documents identified in the complaint do appear to demonstrate that Amazon pays substantial attention to competition from online superstores and online marketplaces. But cherry-picked business documents do not define economically relevant markets.[12] At trial, Amazon will doubtless produce a host of ordinary-course documents that show significant competition from a wide array of competitors on both sides of its retail platform. The scope of competition that the FTC sketches—based on a few documents from among tens of thousands—is a public-relations and litigation tactic, but not remotely the full story.

Third, the FTC’s casual use of “cluster markets,” which lump together distinct types of products and different types of sellers into single markets, may severely undermine the commission’s case. It’s one thing to group, say, all recorded music into a single market (despite the lack of substitutability between, say, death metal and choral Christmas music), but it’s another thing entirely to group batteries and bedroom furniture into a single “market,” just because Amazon happens to facilitate sales of both.

Fourth and finally, it is notable that the relevant markets alleged in the FTC’s complaint draw a distinct line between the seller and buyer sides of Amazon’s platform. Implicit in this characterization is the rejection of cross-market effects as a justification for Amazon’s business conduct. Some of the FTC’s specific concerns—e.g., the alleged obligation imposed on sellers to use Amazon’s fulfillment services to market their products under Amazon’s Prime label—have virtually opposite implications for the seller and buyer sides of the market. Arbitrarily cordoning off such conduct to one market or the other based on where it purportedly causes harm (and thus ignoring where it creates benefit) mangles the two-sided, platform nature of Amazon’s business and would almost certainly lead to its erroneous over-condemnation.[13]

Ultimately, what will determine the scope of the relevant markets will be economic analysis based on empirical data. But based on the FTC’s complaint, public data, and common sense (the best we have to go on, for now), it seems implausible that the FTC’s conception of distinct, and distinctly narrow, relevant markets will comport with reality.

An artificially narrow and gerrymandered market definition is a double-edged sword. If the court accepts it, it’s much easier to show market power. But the odder the construction, the more likely it is to strain the court’s credulity. The FTC has the burden of proving its market definition, as well as competitive harm. By defining these markets so narrowly, the FTC has ensured it will face an uphill battle before the courts.

I.           The Alleged ‘Online Superstore’ Market

A first weakness of the FTC’s suit pertains to the alleged “online superstore market.” This market definition excludes the following: (1) brick-and-mortar retailers, (2) brick-and-mortar sales by firms that do considerable business online and in-person, and (3) online retailers that don’t meet the definition of a “superstore.”[14] The FTC’s market definition also excludes sales of perishable grocery items.[15] The agency argues that consumers don’t consider these other types of retailers to be substitutes for online superstores.[16] This seems dubious, and the FTC’s complaint does little to dispel the doubt.

To see how the market definition tilts the balance, consider the FTC’s allegation that Amazon dominates the online-superstore market with approximately 82% market share.[17] That is, Amazon is reported to have approximately 82% market share (in gross merchandise value, or “GMV”), provided we exclude perishables, and consider the market to comprise solely U.S. online sales by Amazon, Walmart, Target, and eBay, but no other vendors. Note, for example, that Walmart, Target, and Costco all have both online and in-person sales at brick-and-mortar stores, but Costco’s online sales are excluded from the online-superstore category, presumably due to their relatively limited scale and scope. But counting both online and in-person sales, it turns out that twelve-month trailing revenue at Costco is reported to be more than double that of Target, which is included in the FTC’s online-superstore category.[18] Amazon’s share of overall online retail is substantial, but it’s much smaller (37.6%) than its share of a purported market that comprises Amazon, Walmart online, Target online, eBay, and nobody else.[19] Indeed, if one includes total retail sales, then Walmart leads Amazon, not vice versa.[20] And while e-commerce may be substantial and growing, it still represents only about 15% of U.S. retail.[21]

There are countless examples where consumers cross-shop online and offline—televisions and other electronics, clothing, and sporting goods (among many others) spring to mind. Indeed, most consumers would surely be hard-pressed to identify any product they’ve purchased from Amazon that they have not, at some point, also purchased from an offline or non-superstore retailer.

Defining a market with reference to a single retailer’s particular product offering—that is, by a single channel of distribution—is unlikely to “identif[y] the competitive process alleged to be harmed.”[22] In fact, for consumers, it doesn’t identify a product at all, and ends up excluding a host of competing sellers that offer economic substitutes for the products consumers actually buy.[23] By failing to do so, the FTC’s purported market definition is woefully deficient in describing the scope of competition: “Including economic substitutes ensures that the relevant product market encompasses ‘the group or groups of sellers or producers who have actual or potential ability to deprive each other of significant levels of business.’”[24]

A.        Brick-and-Mortar Competes with Amazon Because Shopping Is Not the Same Thing as Consuming

While it may be that some consumers do not consider offline vendors or non-superstores to be substitutes, it does not follow that such rivals don’t impose competitive constraints on online superstores.

If a hypothetical monopolist raises prices, some consumers—perhaps many, perhaps even most—may switch to a brick-and-mortar retailer. That may be enough to constrain the monopolist’s pricing. How many might switch, and the extent to which that constrains pricing, are empirical questions, but there is no question that some consumers might switch: retail multi-homing is common.

And the constraints on switching are far weaker than the FTC claims. The complaint observes that 1) brick-and-mortar retailers are less convenient because it takes time to go to a physical store, 2) stores are not open for shopping at all hours, and 3) consumers may have to visit multiple stores to buy the necessary items.[25]

Online shopping is almost certainly quicker than offline—at least, once one is sitting in front of a computer with Internet access. But the complaint seems to conflate shopping with consuming.

Even with Amazon’s impressive fulfillment and delivery network, if a consumer needs a product that very moment or even that day, a brick-and-mortar retailer may be preferable. The same may be true in circumstances in which a consumer wants to see a product in person, try on clothing, consult an experienced salesperson, etc. And while some consumers may enjoy shopping, they may or may not prefer the experience of online shopping.

More generally and more to the point, consumers purchase goods to use and consume them. Online stores may be “always open,” but shipping and delivery are not instantaneous. That one can shop online at all hours may be convenient, but it may do nothing to hasten the ability to consume the items purchased.

Meanwhile, brick-and-mortar retailers typically have websites that show their inventory and pricing online. Consumers can, accordingly, comparison shop across e-commerce and brick-and-mortar vendors, even when the brick-and-mortar retailers have closed for the evening.

B.         ‘Depth and Breadth’ Isn’t Solely Available from Superstores, and Consumers Buy Products, Not Store Types

Consumers within the “online superstore market” may be able to prevent a hypothetical monopolist from raising prices by switching to other online channels that don’t qualify as a “superstore,” as defined by the FTC.

For example, if a consumer is looking for sporting goods, she can shop at an online superstore, or she can shop at Dick’s online, REI online, or Bass Pro online, all of which have an exceptional “depth and breadth” of items.[26] Alternatively, if the consumer is shopping for a Columbia Sportswear jacket, in addition to the sporting-goods retailers listed, she can also shop on Columbia’s website[27] or at any other online-clothing retailer that carries Columbia jackets (e.g., Macy’s or Nordstrom[28]).

The complaint anticipates and responds to this concern by saying that non-superstore online retailers (as well as brick-and-mortar retailers) lack the depth and breadth of products sold by superstores.[29] But so what? For many consumers, Amazon purchases are made one (or a few) item(s) at a time. When consumers need a bolt cutter, they log in and order it, and when they need a pair of sneakers the next day, they log in and order that. They don’t wait to buy the bolt cutter until they are ready to buy sneakers (i.e., people don’t typically log in to Amazon with a shopping list and purchase multiple items at the same time, except perhaps for perishable groceries, which are excluded from the proposed market). Whether the consumer is buying one item or three or five, a purchase that bundles products across the broad scope of the online-superstore market is not at all the norm.

Indeed, part of the purported advantage of online shopping—when it’s an advantage—is that consumers don’t have to bundle purchases together to minimize the transaction costs of physically visiting a brick-and-mortar retailer. Meanwhile, another part of the advantage of online shopping is the ease of comparison shopping: consumers don’t even have to close an Amazon window on their computers to check alternatives, prices, and availability elsewhere. All of this undermines the claim that one-stop shopping is a defining characteristic of the alleged market.

Data are hard to come by (and the data will ultimately demonstrate whether and to what extent the complaint portrays reality), but public sources indicate that the average number of units per transaction is less than three (admittedly, this is worldwide, and for all online e-commerce, not just Amazon).[30] This does not suggest that shoppers demand extensive “depth and breadth” each time they shop online.

Meanwhile, important lacunae in Amazon’s offerings belie the notion that it offers a true “depth and breadth” that transcends competitive constraints from other retailers. The fact that Nike, on the seller side, doesn’t view Amazon as an essential marketplace[31]—in other words, it believes it has plenty of alternative, competing channels of distribution—has important consequences for the FTC’s market definition on the consumer side. It’s difficult to conceive of a retailer offering anything approaching a comprehensive “depth and breadth” of footwear without offering any Nike shoes. For consumers who buy shoes, Amazon is hardly a unique outlet, and finding even a minimally suitable range of options requires shopping elsewhere, either in combination with Amazon or in its stead.

But the implications are even greater. Because the FTC has grouped sales of all products together—not just footwear or even apparel—and defined the relevant market around that broad clustering of disparate products, can it really be said that Amazon is a “one-stop-shop” at all if it doesn’t offer Nike shoes?

The example may seem trivial, but it aptly illustrates the inherent error in defining the product market essentially by the offerings of a single entity. Necessarily, those offerings will be unique and affected by a host of seller/buyer interactions specific to that company. And in many cases, those specific inclusions and exclusions may be significantly more important than the simple number of SKUs on offer (which is essentially the basis for including Walmart and Target online, but excluding, say, Costco online from the FTC’s “superstores” market).

Further, despite its repeated reliance on “depth and breadth,” the complaint ignores e-commerce aggregators, which allow consumers to search products and pricing across an incredible variety of retailers. Google Shopping is, of course, the most notable example—and, for such a prominent example, curiously absent from the complaint. Through Google Shopping—among other sites—consumers can see extensive results in one place for almost any product, including across all categories and across many brands (the breadth-and-depth factors relied upon by the complaint). Indeed, while many product searches today begin at Amazon, a huge amount of online shopping takes place via Google.[32]

Moreover, online shoppers regularly use third-party sites to research (shop) for products, and these, too, aggregate information from across a huge range of sources. As Search Engine Land reports:

Reviews and ratings can make or break a sale more than any other factor, including product price, free shipping, free returns and exchanges, and more.

Overall, 77% of respondents said they specifically seek out websites with reviews—and this number was even higher for Gen Z (87%) and millennials (81%).[33]

While Amazon is where consumers most often read reviews (94%), other retail websites (91%), search engines (70%), brand websites (68%), and independent review sites (40%) are all significant.[34] And yet, despite their manifest importance in the competitive process of online retail, the FTC’s complaint entirely dismisses the significance of shopping aggregators and non-Amazon, product-review sources.

II.         The Alleged ‘Online Marketplace Services’ Market

The complaint is similarly flawed when it assesses the scope of competition from the point of view of sellers.

The complaint endeavors to distinguish and exclude from the market for online marketplace services all other methods by which a seller can market and sell its products to end consumers. For instance, the complaint distinguishes online marketplaces from online retailers where the seller functions as a vendor (i.e., it transfers title to the retailer) and those where sellers provide their own storefronts or sell directly through social media and other aggregators using “software-as-a-service” (“SaaS”) to market products (e.g., Shopify and BigCommerce).[35]

The complaint alleges that neither operating as a vendor nor utilizing SaaS is “reasonably interchangeable”[36] with online marketplace services—the key language from the Brown Shoe case.[37] But merely saying so does not make it true. Service markets can display differentiated competition, just as product markets do. Superficial—and even significant—differences among services do not, in themselves, establish that they are not competitors.

First, where sellers operate as vendors by transferring title to another party to sell the product (either online or at a brick-and-mortar retailer), they could very well constrain the costs that a hypothetical monopolist imposes on sellers. For example, if a hypothetical monopolist increased prices or decreased quality for selling a product, why would Nike not transfer its products away from the monopolist and toward Foot Locker, Macy’s, or any other number of retailers where Nike operates as a vendor? Or why not rely on Nike’s own website, selling directly to the consumer? In fact, Nike has already done this. In 2019, Nike stopped selling products to Amazon because it was dissatisfied with Amazon’s efforts to limit counterfeit products.[38] Instead, Nike opted to sell directly to its consumers or through its other retailers (both online and offline, of course).

The same can be said for sellers without well-known brands or those who opt to use SaaS to sell their products. Certainly, there are differences between SaaS and online-marketplace services, but that doesn’t mean that a seller can’t or won’t use SaaS in the face of increased prices or decreased quality from an online marketplace. Notably, Shopify claims to be the third-largest online retailer in the United States, with 820,000 merchants selling through the platform.[39] It’s remarkable that it is completely absent from the FTC’s market definition.

Also remarkable is that he FTC’s complaint alleges that SaaS providers are not in the relevant market because:

SaaS providers, unlike online marketplace service providers, do not provide access to an established U.S. customer base. Rather, merchants that use SaaS providers to establish direct-to-consumer online stores must invest in marketing and promotion to attract U.S. shoppers to their online stores.[40]

This is remarkable because a significant claim in the FTC’s complaint is that Amazon has “degraded” its service by introducing sponsored search results, “litter[ing] its storefront with pay-to-play advertisements,” and allegedly requiring (some would say enabling…) sellers to pay for marketing and promotion.[41] It’s unclear why the need to invest in marketing and promotion to attract shoppers to one’s online storefront is qualitatively different than the need to invest in marketing and promotion to attract shoppers to one’s products on Amazon’s platform.

Indeed, the notion that large platforms like Amazon simply “provide access” to consumers glosses over the immense work that such access entails. Amazon and similar platforms (including, of course, SaaS providers) make significant investment in designing and operating user interfaces, matching algorithms, marketing channels, and innumerable other functionalities to convert undifferentiated masses of consumers and sellers into a functional retail experience. Amazon’s value for sellers in providing access to customers must be balanced by the reality that, in doing so, large “superstores” like Amazon also necessarily put a large quantity of disparate sellers in the same unified space.

For obvious reasons, sellers don’t necessarily value selling their products in the same location as other sellers. They do, of course, want access to consumers, but the “marketplace” or “superstore” aspects of Amazon simultaneously impedes that access by congesting it with other sellers and products (and consumers seeking other products). A specialized outlet may, in fact, offer the optimal sales environment: all consumers seeking the seller’s category of goods (but somewhat fewer consumers), and fewer sellers impeding discovery and access (though more selling the same category of goods). A furniture seller may have dozens of online outlets (and, of course, many offline outlets, catalog sales, decorator sales, etc.), and there is little or no reason to think that, by virtue of also offering batteries, clothes, and bolt cutters, Amazon offers anything truly unique to a furniture seller that it can’t get by selling through another distribution channel with a different business model.

The complaint relies heavily on this notion that online-marketplace services deliver a large customer base that cannot be matched by selling as a vendor or using SaaS. (It is entirely unclear if the FTC considers single-category online marketplaces like Wayfair to be in the “online marketplace services” market, a topic to which I return below in the “cluster markets” discussion; it is clear the FTC doesn’t consider Wayfair part of the “online superstores market.”).[42] Again, in this context, the complaint ignores e-commerce aggregators and how they affect sellers’ ability to access customers. Through Google Shopping, consumers can see extensive results for almost any product, including across all categories and across many brands. And Google aggregates product listings without charging the seller.[43] Thus, through Google Shopping, a seller can access a large consumer base that may constrain a hypothetical monopolist in the online-marketplace-services market.

And Google Shopping is not alone. Selling through social media has boomed. According to one source, Instagram is an online-shopping juggernaut.[44] Among other things:

  • 130 million people engage with shoppable Instagram posts monthly;
  • 72% of users say they made a purchase based on something they saw on Instagram;
  • 70% of Instagram users open the app in order to shop; and
  • 81% of Instagram users research new products and services on the platform.[45]

Sellers on Instagram can use Meta’s “Checkout on Instagram”[46] service to process orders directly on Instagram, as well as logistics services like Shopify or ShipBob to manage their supply chains and fulfill sales,[47] replicating the core functionality of a vertically integrated storefront like Amazon.

The bottom line is that Amazon is not remotely the only (or, in many cases, even the best) place for sellers to find, market, and sell to consumers. Its superficial differences from other distribution channels are just that: superficial.

III.       Cluster Markets

One of the most important problems with the FTC’s alleged relevant markets is that they treat all products and all sellers the same. They effectively assume that consumers shop for bolt cutters the same way they shop for furniture, and that Adidas sells shoes the same way that drop-shippers sell toilet paper.

Courts have recognized that such an approach—using “cluster markets” to assess a group of disparate products or services in a single market—can be appropriate for the sake of “administrative[ ]convenience.” As the 6th U.S. Circuit Court of Appeals noted in Promedica Health v. FTC, “[t]his theory holds, in essence, that there is no need to perform separate antitrust analyses for separate product markets when competitive conditions are similar for each.”[48]

A second basis for clustering is the “transactional-complements” theory, relabeled by the 6th Circuit as the “‘package-deal’ theory.”[49] This approach clusters products together for relevant market analysis when “‘most customers would be willing to pay monopoly prices for the convenience’ of receiving certain products as a package.”[50]

For example, it may be appropriate to refer to a “market for recorded music” even though consumers of music by Taylor Swift probably exert little or no competitive pressure on the price or demand for recordings of, say, Cannibal Corpse. Thus, in the EU’s 2012 clearance (with conditions) of the Universal Music Group/EMI Music merger, the Commission determined that, although classical music may present somewhat different competitive dynamics, there was no basis for defining separate markets by artist or even by genre.[51]

Hospital mergers provide another classic example.[52] Labor and delivery services are not a substitute for open-heart surgery, but the FTC nonetheless frequently defines a market as “inpatient general acute care services” or something similar because of the similar relationship of each to a hospital’s organization and administration, as well as the fact that payers typical demand such services (and hospitals typically provide such services) in combination (even though patients, of course, do not consume them together).

The Supreme Court put its imprimatur on the notion of a cluster market in Philadelphia National Bank, accepting the lower court’s determination that “commercial banking” constituted a relevant market because of the distinctiveness, cost advantages, or consumer preferences of the constituent products.[53]

A.        Assessing Cluster Markets

Widespread use (and the occasional fairly serious analysis) of cluster markets notwithstanding, it is worth noting that the economic logic of such markets is, at best, poorly established.

In the UMG/EMI case, for example, the Commission rested on the following factors in concluding that markets should not be separated out by genre (let alone by artist):

The market investigation showed that, by and large, a segmentation of the recorded music market based on genre is not appropriate. First, the borders between genres are often blurred and artists and songs can fit within several genres at the same time. Second, several customers also underline that placing of a song or an album into a specific genre is entirely subjective. Third, a vast majority of customers indicated that they purchase and sell all genres of music.[54]

These facts may all be true, but they do little to permit the inference drawn. Indeed, the first two factors arguably refer only to administrability, not economic reality, and the third is woefully incomplete (e.g., it says little about a potential monopolist’s ability to raise prices if price increases can be passed on to end-consumers in some genres but not others). While the frailties of the market determination may not ultimately have mattered in that case (after all, the parties got their merger, and the Commission presumably brought the strongest case it could), such casual conclusions may well prove problematic elsewhere and do little to advance the logic of the cluster-markets concept.

Similar defects plague the Supreme Court’s endorsement of the theory in PNB. The Court suggests some reasons why, even in its own telling, “some commercial banking products or services”[55] may be insulated from competition, but that still leaves open the possibility that others aren’t, and that the relevant insulating characteristics could be eroded by simple product repositioning, different pricing strategies, or changes in reputation and brand allegiance.

In fact, the defendants in PNB argued before the district court that:

commercial banking in its entirety is not a product line. Rather, they submit it is a business which has two major subdivisions—the acceptance of deposits in which the bank is the debtor, and the making of loans in which the bank is the creditor. Both of these major divisions are further divided by distinct types of deposits and loans. As to many of these functions, there are different types of customers, different market areas, and, most importantly, different types of competitors and competition. With the possible exception of demand deposits, there is an identical or effective substitute for each one of the services which a commercial bank offers.[56]

The court, however, rejected these arguments with little more than a wave of the hand (a conclusion that was then simply accepted by the Supreme Court):

It seems quite apparent that both plaintiff’s and defendants’ positions have some merit. However, it is not the intention of this Court to subdivide a commercial bank into certain selected services and functions. An approach such as this, carried to the logical extreme, would result in many additional so-called lines of commerce. It is the conglomeration of all the various services and functions that sets the commercial bank off from other financial institutions. Each item is an integral part of the whole, almost every one of which is dependent upon and would not exist but for the other. The Court can perceive no useful purpose here in going any further than designating commercial banking a separate and distinct line of commerce within the meaning of the statute. It is undoubtedly true that some services of a commercial bank overlap, to some degree, with those of certain other institutions. Nevertheless, the Court feels quite confident in holding that commercial banking, viewed collectively, has sufficient peculiar characteristics which negate reasonable interchangeability.[57]

None of this response goes to the question of how users of commercial-banking services consume them. Instead, it essentially takes the superficial marketing distinction as economically dispositive, despite the acknowledgment that economic substitutes for the constituent products exist. It is, of course, possible that, in PNB, the error was not outcome determinative; perhaps none of the overlap between commercial banks and other providers of commercial lending is significant enough to change the analysis. But this is not a rigorous defense of the notion.

In a few cases, a more rigorous econometric analysis has been used to establish the viability of cluster markets. Consider, for example, the FTC’s successful challenge of the proposed Penn State Hershey Medical Center/Pinnacle Health System merger.[58] At issue there were the likely effects of a merger for certain services provided by general acute care (GAC) hospitals—that is, a range or “cluster” of services sold to commercial health plans in a defined geographic area covering roughly four counties in central Pennsylvania. Two small community hospitals offered some of the same acute care services, and various clinics and group practices provided some of the primary and secondary care services in the cluster.

At the same time, there was evidence that commercial health plans needed to negotiate for coverage over a range of GAC services that other providers could not offer, and that the merging parties competed on price in such negotiations with commercial health plans. Copious econometric evidence—analysis of price data and patient-draw data—substantiated the FTC’s market definition, bolstered by an amicus brief filed by more than three dozen experts in antitrust, competition, and health-care economics.[59]

All of this supported the FTC’s argument that the provision of GAC services constituted a single “cluster market”—and the 3rd U.S. Circuit Court of Appeals agreed, overturning a flawed geographic-market definition initially adopted by the district court.[60] That is, the agency didn’t merely waive its hands at an impression of ways that certain hospital services were similar to each other; rather, it provided detailed economic analysis of the price competition at issue for a specific range of GAC hospital services.

Notably, in that case, there were specific, identifiable consumers—commercial health plans—that were negotiating prices for a diverse “cluster” of GAC services. An individual patient will not, we hope, need to shop for oncology, cardio-thoracic surgery, a hip replacement, and ob-gyn services at the same time. But a health plan typically considers all of those and more. The same dynamic is not, of course, applicable in the Amazon case.

Perhaps the best example of the rigorous defense of cluster markets came in the first Staples/Office Depot merger matter, where ordinary-course documents played a role in the FTC’s review, but were by no means core to the staff’s analysis.[61] The FTC Bureau of Economics applied considerable econometric analysis of price data to establish that office superstore chains constrained each other’s pricing in a way that other vendors of office supplies did not.[62] That analysis of price effects (as evidence of likely merger effects and as evidence on behalf of the FTC’s market definition) is not apparent in the district court’s opinion enjoining the transaction.[63] But it figured heavily in the FTC’s presentation of the case and, presumably, in the commission’s internal decision to bring the case.

Two things are particularly notable about the cluster markets employed in Staples/Office Depot. First is that the exercise was undertaken at all. That is, it was assumed to be a crucial question whether other types of retailers (those with fewer products or catalog-only sales) constrained the pricing power of office-supply “superstores.” Second, the groupings of products analyzed were based on detailed analyses of pricing and price sensitivity over identified products, not superficial, subjective impressions of the market. The same was likewise the case in the Penn State Hershey hospital case mentioned above, and in other hospital-merger cases.

These types of evidence and analyses are simply not in evidence in the FTC’s case against Amazon—certainly not as they’ve presented it thus far.

B.         The Problem of Cluster Markets in the FTC’s Amazon Complaint

The FTC’s approach to market definition in Amazon appears in sharp contrast with prior cases involving what were, arguably, valid cluster markets and somewhat narrow market definitions.

Although the Amazon case is only at the complaint stage, of course, no factors or analysis similar to those adduced in the hospital and office-superstore cases discussed above are present in the FTC’s complaint against Amazon. Indeed, the complaint offers no evidence that the FTC considered the possibility that different products and different sellers would need to be considered separately (the FTC certainly saw no need to preemptively defend its clustering in the complaint). Instead—and consistent with the apparent assumption that Amazon and its particular characteristics are virtually unique—the complaint appears to assume that if Amazon offers a grouping of products, or if Amazon offers services to different types of sellers, this constitutes an economically rigorous “relevant market.” (Spoiler alert: It does not.)

Such an assumption would seem to need some defense. Certainly, a customer buying a bolt cutter will not consider buying a sneaker to be a reasonable alternative; it is clearly not on the basis of demand substitution that the FTC lumps these products together.[64] Instead, similar competitive conditions across products are implicit in the FTC’s alleged markets. But are competitive conditions sufficiently similar across products sold on Amazon to justify clustering them?

1.           Buyer-side clustering

Conditions vary considerably across the broad swath of products sold on Amazon. For some products sold at online superstores, brick-and-mortar retailers are a much closer substitute. Conceivably, consumers may prefer buying shoes at a brick-and-mortar retailer so that they can try them on, making physical retail a closer substitute for sneakers than for, say, a toilet brush, where very few consumers will demand to try the brush for balance before buying it. And surely consumers may be more willing to buy well-established brands (Nike, Gucci, etc.) directly from the brand’s website than a lesser-known brand sold at an online superstore.

Furniture, for example, is bought and sold in vastly different ways than, say, batteries (by consumers with different preferences for service and timing, by retailers with different relationships with manufacturers, through different channels of distribution, etc.). Whatever the merits to consumers of bundling purchases together from an “online superstore,” it is likely the case that they far less often bundle furniture purchases with other purchases than they do batteries. And surely consumers far more often seek to buy furniture offline or after testing it out in person than they do batteries. Vertically integrated furniture stores like IKEA have certainly done much to “commoditize” the production and sale of furniture in recent decades, but the market remains populated mostly by independent furniture showrooms, traditional manufacturers, and catalog and decorator sales. The same cannot be said for batteries, of course.

It also seems unlikely that consumers purchase Amazon’s proffered products in bundles meaningfully distinct from those they purchase elsewhere. People shopping for kitchen pantry items may well bundle their purchases of these items together. But in the vast majority of cases, they can get that same bundle from a grocery store, even though the grocery store carries many fewer SKUs overall. There is no analog to commercial health plans negotiating prices for a particular “cluster” of hospital services in Amazon’s case—and even if there were, it is certain that any number of other stores can match the actual clusters in which people regularly buy products from Amazon.

2.           Seller-side clustering

The problem of false clustering is even more acute on the seller side in the alleged “online marketplace services” market. Sellers on Amazon comprise at least two distinct types. On the one hand are brands and manufacturers that have a limited range of their own products to offer. These sellers are not resellers of others’ goods, but product creators or brands that use Amazon to sell “direct to consumer” the same sort of products they might otherwise have to sell through a retail intermediary. Within this group there is a further distinction between large, known brands and entrepreneurs selling a unique product (or maybe a few unique products) of their own creation out of their proverbial garage.[65]

On the other hand are retailers—resellers—that offer a wide range of products, none of which they manufacture themselves, but which they may purchase in bulk from manufacturers or offer through drop-shipping. The seller is an intermediary between the actual maker or seller of the product and the customer (in this case, marketing and reaching customers through another intermediary: Amazon). Here, again, there is a further distinction between intermediaries that are virtually invisible or interchangeable pass-throughs of others’ goods and those that attempt to add some value by establishing their own private-label brands or by acting as a trusted intermediary that offers a curated set of products.

Each of these types of sellers has a different demand for the various services bundled by Amazon, and a different set of available alternatives to Amazon. They often compete in different markets, have different relationships with manufacturers, and have differing sets of internal capacities necessitating the purchase of different services (or the purchase of different services in different relative quantities), and entailing a different ability to evaluate their need for different services and differing degrees of reliance on Amazon to complement their capacities. Moreover, the competitive ramifications of constraining each’s ability to sell on Amazon (or increasing the price to do so) is considerably different.

This last point is most obvious when considering the effect on drop-shippers of a possible increase in price on Amazon. What would be the competitive effects if a particular drop-shipper of, say, toilet paper were somehow precluded from Amazon, or harmed by using it? In that case, the seller is largely irrelevant (or worse—simply an additional source of markup). The relevant question is not whether a particular seller can profitably sell the product: “The antitrust laws… were enacted for ‘the protection of competition not competitors.’”[66] Rather, the relevant question is whether the manufacturer of the product can access consumers, and whether consumers can access competing sellers. In the case of toilet paper (or virtually anything else drop-shipped), the answer is manifestly yes. Drop shippers of Charmin could probably disappear completely from Amazon, and consumers would still be able to buy it at competitive prices from Amazon, among a host of competing options, and Proctor & Gamble would have no trouble reaching consumers.

3.           Implications

The implication of all this is that it seems highly dubious that furniture and batteries (to take just one example) face similar enough competitive conditions across online superstores for them to be grouped together in a single “cluster market.” While there may be superficial similarities in the website or technology connecting buyers and sellers, the underlying economics of production, distribution, and consumption seem to vary enormously.

The complaint offers no evidence to support the assertion of similar competitive conditions; no analysis of cross-elasticities of demand or supply across product categories; and no empirical evidence that a price increase for, say, furniture, could be offset by increased sales of batteries. Nor does the complaint consider more granular markets—like furniture, or sporting goods, or books—that would better capture these critical differences.

Indeed, it’s quite possible that narrower markets would demonstrate that Amazon faces real competition in some areas but not others. Grouping disparate products together risks obscuring situations where market power—and thus potentially anticompetitive effects from Amazon’s conduct—might exist in some product spaces but not others. The failure to properly define the relevant market for antitrust analysis doesn’t inherently imply a particular outcome; it just means no outcome can properly be determined.

The FTC offers no defense for clustering beyond the mere fact that Amazon offers these varied products on its platform. Yet selling through a common intermediary hardly establishes that the underlying competition is sufficiently similar to warrant single-market treatment, let alone that common conduct toward sellers affects all products and sellers equally. If the FTC cannot empirically defend treating distinct products as competitively interchangeable, as transactional complements, or as having the same competitive conditions, its case may collapse under the weight of its own market gerrymandering.

IV.      Out-of-Market Effects

This leaves a final question about the two markets defined in the complaint: can and should they really be considered separately, when conduct in each market has significant effects in the other? My colleagues and I intend to address this question more broadly and in more detail in the future (and, indeed, have already begun to do so[67]). For now, I will share a few tantalizing thoughts about this issue.

If Amazon’s practices vis-à-vis sellers cause the sellers to lower their prices, improve the quality of the products available through the marketplace, or otherwise lower costs and whittle down the seller’s profits, then consumers would benefit. Similarly, if Amazon’s practices with sellers improve the quality of consumers’ experience on its marketplace, then consumers would also benefit. The question is whether gain on one side should offset any harms on the other.

The FTC contends that the markets should be considered separately, despite acknowledging (and even trying to bolster its case with) the reality that the two sides of Amazon’s platform have important effects on each other:

Feedback loops between the two relevant markets further demonstrate the critical importance of scale and network effects in these markets. While the markets for online superstores and online marketplace services are distinct, an online superstore may operate an online marketplace and offer associated online marketplace services to sellers. As a result, the relationship and feedback loops between the two relevant markets can create powerful barriers to entry in both markets.[68]

Despite this, the FTC will likely contend that out-of-market efficiencies are not cognizable. That is, benefits to consumers in the online-superstore market that flow from harm in the online-marketplace-services market do not apply (i.e., harm is harm, and it doesn’t matter if it benefits someone else). This approach, however, presents some obvious problems.

If platforms undertake conduct to maximize the overall value of the platform (and not merely the benefits accruing to any one side in particular), it is inevitable that some decisions will impose constraints on some users in order to maximize the value for everyone. Indeed, the FTC attempts to disparage “Amazon’s flywheel” as a mechanism for exploiting its dominance.[69] For Amazon, meanwhile, that “flywheel” encompasses the importance of ensuring value on one side of the platform in order to increase its value to the other side:

A critical mass of customers is key to powering what Amazon calls its “flywheel.” By providing sellers access to significant shopper traffic, Amazon is able to attract more sellers onto its platform. Those sellers’ selection and variety of products, in turn, attract additional shoppers.[70]

But at times, maximizing the value of the platform may entail imposing constraints on sellers or buyers. Unfortunately, some of these practices are the precise ones the FTC complains of here. Limiting access to the “Buy Box” by sellers of products that are available for less elsewhere, for example, ensures that consumers pay less and builds Amazon’s reputation for reliability;[71] bundling Prime services may mean some consumers pay for services they don’t use in order to get fast shipping, but it also attracts more Prime customers, enabling Amazon to raise revenue sufficient to guarantee same-, one-, or two-day shipping and providing a larger customer base for the benefit of its sellers.[72]

The bifurcated market approach also conflicts with the Supreme Court’s holding in Ohio v. American Express.[73] In Amex, the Court held that there must be net harm to both sides of a two-sided market (like Amazon) before a violation of the Sherman Act may be found. And even the decision’s critics recognize the need to look at effects on both sides of the market (whether they are treated as a single market, as in Amex, or not).[74]

The complaint itself seems to provide enough fodder to suggest that Amazon’s marketplace should be treated as a two-sided market, which the Supreme Court defined as a “platform [that] offers different products or services to two different groups who both depend on the platform to intermediate them.”[75] The complaint is replete with allegations of a “feedback loop” between the two markets, and it does appear that the consumers depend on the sellers and vice versa.

The economic literature shows that two-sided markets exhibit interconnectedness between their sides. It would thus be improper to consider effects on only one side in isolation. Yet that is what artificially narrow market definitions facilitate—letting plaintiffs make out a prima facie case of harm in one discrete area. This selective focus then gets upended once defendants demonstrate countervailing efficiencies outside that narrow market.

But why define markets so narrowly if weighing interrelated effects is ultimately essential? Doing so seems certain to heighten false-positive risks. Moreover, cabining market definitions and then trying to “take account” of interdependencies is analytically incoherent. It makes little sense to start with an approach prone to missing the forest for the trees, only to try correcting the distorted lens part way into the analysis. If interconnectedness means single-market treatment is appropriate, the market definition should match from the outset.

But I think the FTC is aiming not for the most accurate approach, but for the one that (it believes) simply permits it to ignore procompetitive effects in other markets, despite its repeated acknowledgment of the “feedback loops” between them.[76] Certainly, FTC Chair Lina Khan is well aware of the possible role that Amex could play, and has even stated previously that she believes Amex does apply to Amazon.[77] Instead, the agency is hoping (incorrectly, I believe) that the Court’s decision in Amex won’t apply, and that its decisions in PNB and Topco will ensure that each market be considered separately and without allowance for “out-of-market” effects occurring between them.[78] Such an approach would make it much easier for the FTC to win its case, but would do nothing to ensure an accurate result.

The district court in Amex, in fact, took a similar approach (finding in favor of the plaintiffs), holding that the case involved “two separate yet complementary product markets.”[79] Citing Topco and PNB, the district court asserted that, “[a]s a general matter . . ., a restraint that causes anticompetitive harm in one market may not be justified by greater competition in a different market.”[80] Similarly, Justice Stephen Breyer, also citing Topco, concluded in his Amex dissent that a burden-shifting analysis wouldn’t incorporate consideration of both sides of the market: “A Sherman Act §1 defendant can rarely, if ever, show that a procompetitive benefit in the market for one product offsets an anticompetitive harm in the market for another.”[81]

Some scholars assert that PNB and Topco apply to preclude offsetting, “out-of-market” efficiencies in monopolization cases, but it is by no means clear that the PNB limitation applies in Sherman Act cases. As a matter of precedent, PNB applies only to mergers evaluated under the Clayton Act. And the claim that the Court in Topco has extended the holding in PNB to the Sherman Act rests (at best) on dicta.[82]

It is true that the Court limited Amex to what it called “transaction” markets.[83] But courts are almost certainly going to have to deal with interrelated effects that occur in less-simultaneous markets, and they will almost certainly have to do so either by extending Amex’s single-market approach, or by accepting out-of-market efficiencies in one market as relevant to the antitrust analysis of an ostensibly distinct market on the other side of the platform. The FTC’s Amazon complaint presents precisely this dynamic.

Legal doctrine aside, ignoring benefits in one interconnected market while focusing on harms in another will lead to costly overdeterrence of procompetitive conduct.

Indeed, the FTC’s complaint identifies not just ambiguous conduct (conduct that may constrain one side but benefit the other side and the platform overall), but it points to the very act of providing benefits to consumers as a means of harming competition.[84]

What if Amazon makes it harder for new entrants on the “marketplace” side to enter profitably, because it offers benefits on the consumer side that most competitors can’t match? The FTC would have you believe that is a harm, full stop, because of the seller-side effect. But that would also effectively mean that simply increasing efficiency and lowering prices would amount to harm, because it would also make it harder for new entrants to match Amazon. How can conduct that provides a clear benefit to consumers constitute an antitrust harm?[85]

In essence, the FTC maintains this illogical position by cordoning off the two sides of Amazon’s platforms into separate markets and then asserting that benefits in one cannot justify “harms” in the other, despite recognizing the close interrelatedness between the two markets:

Sellers who buy marketplace services from Amazon provide much of the product selection that helps Amazon attract and keep its shoppers. As more shoppers turn to Amazon for its product selection, more sellers use its platform to gain access to its ever-expanding consumer base, which attracts more shoppers, and so on. . . . The interplay between Amazon’s shoppers and sellers increases barriers to new entry and expansion in both relevant markets and limits existing rivals’ ability to compete. In this way, scale builds on itself, and is cumulative and self-reinforcing.[86]

This is artificial and nonsensical. What Amazon does is maximize the value of the platform to the benefit of all users, on net. That some of those benefits accrue at certain times to only one set of users cannot be taken to undermine the value of Amazon’s overall, long-term platform-improving conduct.

Finally, it is worth noting that, even where nominal market distinctions across platform users have been argued by plaintiffs and upheld by courts, analysis of anticompetitive effects has generally turned to out-of-market effects.

Consider the famous case of Aspen Skiing Co. v. Aspen Highlands Skiing Corp. In that case, analyzing the competitive effect of the defendant’s conduct regarding access by a competitor to an “all Aspen” ski pass required looking at effects in the output market for downhill skiing, as well as the input market for mountain access needed to provide those tickets.[87] Indeed, as the Court noted, “[t]he question whether Ski Co.’s conduct may properly be characterized as exclusionary cannot be answered by simply considering its effect on Highlands. In addition, it is relevant to consider its impact on consumers and whether it has impaired competition in an unnecessarily restrictive way.”[88] If Aspen Skiing were evaluated as the FTC seeks in this case, there would be two distinct markets at issue, and harm could be proven by assessing the effect on the input market alone, regardless of the effect on consumers.

Indeed, especially where vertically related markets are involved (which is, of course, how the two sides of Amazon’s platform are related), courts have recognized that weighing effects on competition requires a cross-market perspective across both upstream and downstream segments.

Conclusion

The FTC’s proposed market definitions in its case against Amazon exhibit several critical flaws that undermine the complaint. The alleged “online superstore” and “online marketplace services” markets are excessively narrow, excluding manifest competitors and alternatives. The FTC improperly groups together distinctly different products and sellers into questionable “cluster markets” without empirical evidence to support treating them as economically integrated. And the complaint arbitrarily cordons the two markets off from each other, despite acknowledging their interconnectedness, likely in a deliberate effort to avoid weighing out-of-market efficiencies and procompetitive effects flowing between them.

Ultimately, the burden lies with the FTC to defend these narrow market definitions as economically sound. But based on the limited information available thus far, the proposed markets appear to be gerrymandered to suit the FTC’s case, rather than reflective of actual competitive realities.

Whether deliberately tactical or not, the problems with the FTC’s market definition invite skepticism regarding the overall merits of the agency’s case. If the relevant markets prove indefensible upon fuller examination of the facts, the theory of harm in the case may well collapse. At a minimum, the FTC faces an uphill battle if its case indeed rests more on artful pleading than rigorous economics.

 

[1] Gregory J. Werden, Why (Ever) Define Markets? An Answer to Professor Kaplow, 78 Antitrust L.J. 729, 741 (2013) (emphasis added).

[2] See, e.g., Josh Sisco, The FTC Puts Your Lunch on Its Plate, Politico (Nov. 21, 2023), https://www.politico.com/news/2023/11/21/feds-probe-10b-deal-for-subway-sandwich-chain-00128268.

[3] Complaint, F.T.C., et al. v. Amazon.com, Inc., Case No. 2:23-cv-01495-JHC (W.D. Wa., Nov. 2, 2023) at ¶¶ 119-208, available at https://www.ftc.gov/legal-library/browse/cases-proceedings/1910129-1910130-amazoncom-inc-amazon-ecommerce (“Amazon Complaint”).

[4] Id. at ¶ 124.

[5] Id.

[6] Id.

[7] Nike Store (last visited Dec. 6, 2023), https://www.nike.com.

[8] Wayfair (last visited Dec. 6, 2023), https://www.wayfair.com.

[9] E-Commerce Retail Sales as a Percent of Total Sales (ECOMPCTSA), FRED Economic Data (last updated Nov. 17, 2023), https://fred.stlouisfed.org/series/ECOMPCTSA.

[10] Amazon Complaint, supra note 3, at ¶ 185.

[11] See, e.g., How Google Shopping Works, Google (last visited Dec. 6, 2023), https://support.google.com/faqs/answer/2987537; Shopify Official Website, Shopify (last visited Dec. 6, 2023), https://www.shopify.com/; Instagram Shopping, Instagram (last visited Dec. 6, 2023), https://business.instagram.com/shopping.

[12] See Geoffrey A. Manne & E. Marcellus Williamson, Hot Docs vs. Cold Economics: The Use and Misuse of Business Documents in Antitrust Enforcement and Adjudication, 47 Ariz. L. Rev. 609 (2005).

[13] For a discussion of this problem in the context of mergers (but with relevance to market definition in Section 2 cases), see Daniel J. Gilman, Brian Albrecht and Geoffrey A. Manne, The Conundrum of Out-of-Market Effects in Merger Enforcement, Truth on the Market (Jan. 16, 2024), https://truthonthemarket.com/2024/01/16/the-conundrum-of-out-of-market-effects-in-merger-enforcement.

[14] See Amazon Complaint, supra note 3, at ¶ 117.

[15] See id. at ¶ 163.

[16] See id. at ¶ 123 (“Online superstores offer shoppers a unique set of features”).

[17] See id. at ¶ 171. (“Other commercially available data, including recently reported statistics from eMarketer Insider Intelligence, a widely cited industry market research firm, confirms Amazon’s sustained dominance across this same set of companies, with an estimated market share of more than 82% of GMV in 2022.”).

[18] See Matthew Johnston, 10 Biggest Retail Companies, Investopedia (last updated May 8, 2023), https://www.investopedia.com/articles/markets/122415/worlds-top-10-retailers-wmt-cost.asp.

[19] Stephanie Chevalier, Market Share of Leading Retail E-Commerce Companies in the United States in 2023, Statista (Nov. 6, 2023), https://www.statista.com/statistics/274255/market-share-of-the-leading-retailers-in-us-e-commerce.

[20] See Matthew Johnston, supra note 18.

[21] See E-Commerce Retail Sales as a Percent of Total Sales, supra note 9.

[22] Werden, supra note 1, at 741.

[23] See Geoffrey A. Manne, Premium Natural and Organic Bulls**t, Truth on the Market (Jun. 6, 2007), https://truthonthemarket.com/2007/06/06/premium-natural-and-organic-bullst (“[E]conomically relevant market definition turns on demand elasticity among consumers who are often free to purchase products from multiple distribution channels, [and] a myopic focus on a single channel of distribution to the exclusion of others is dangerous.”).

[24] Hicks v. PGA Tour, Inc., 897 F.3d 1109, 1120-21 (9th Cir. 2018) (citing Newcal Indus., Inc. v. Ikon Office Sol., 513 F.3d 1038, 1045 (9th Cir. 2008)).

[25] See Amazon Complaint, supra note 3, at ¶¶ 128-33.

[26] Dick’s Sporting Goods (last visited Dec. 6, 2023), https://www.dickssportinggoods.com; REI Co-op Shop (last visited Dec. 6, 2023), https://www.rei.com; Bass Pro Shops (last visited Dec. 6, 2023), https://www.basspro.com/shop.

[27] Jackets, Columbia (last visited Dec. 10, 2023), https://www.columbia.com/c/outdoor-jackets-coats.

[28] See Columbia Coats & Jackets, Macy’s (last visited Dec. 10, 2023), https://www.macys.com/shop/womens-clothing/womens-coats/Brand/Columbia?id=269; Women’s Columbia Coats, Nordstrom (last visited Dec. 10, 2023), https://www.nordstrom.com/browse/women/clothing/coats-jackets?filterByBrand=columbia.

[29] See Amazon Complaint, supra note 3, at ¶¶ 148-59.

[30] See Daniela Coppola, Average Number of Products Bought Per Order Worldwide from January 2022 to December 2022, Statista (Feb. 1, 2023), https://www.statista.com/statistics/1363180/monthly-average-units-per-e-commerce-transaction.

[31] See Khadeeja Safdar, supra note 38.

[32] Google Product Discovery Statistics, Think with Google (last visited Dec. 6, 2023), https://www.thinkwithgoogle.com/marketing-strategies/search/google-product-discovery-statistics (“49% of shoppers surveyed say they use Google to discover or find a new item or product”). Also notable, “51% of shoppers surveyed say they use Google to research a purchase they plan to make online.” Product Research Statistics, Think with Google (last visited Dec. 6, 2023), https://www.thinkwithgoogle.com/marketing-strategies/search/product-research-search-statistics.

[33] See Danny Goodwin, 50% Of Product Searches Start on Amazon, Search Engine Land (May 16, 2023), https://searchengineland.com/50-of-product-searches-start-on-amazon-424451.

[34] Id.

[35] See Shopify (last visited Dec. 6, 2023), https://www.shopify.com; BigCommerce (last visited Dec. 6, 2023), https://www.bigcommerce.com.

[36] Amazon Complaint, supra note 3, at ¶ 198 (“SaaS providers’ services are not reasonably interchangeable with online marketplace services.”).

[37] See Brown Shoe Co., Inc. v. United States, 370 U.S. 294, 325 (1962) (“The outer boundaries of a product market are determined by the reasonable interchangeability of use or the cross-elasticity of demand between the product itself and substitutes for it.”).

[38] See, e.g., Khadeeja Safdar, Nike to Stop Selling Directly to Amazon, Wall Street J. (Nov. 13, 2019), https://www.wsj.com/articles/nike-to-stop-selling-directly-to-amazon-11573615633.

[39] See Tomas Kacevicius (@intred), Twitter (Jun. 19, 2019, 7:05 PM), https://x.com/intred/status/1141527349193842688?s=20 (“[M]ore than 820K merchants are currently using #Shopify, making it the 3rd largest online retailer in the US.”).

[40] Amazon Complaint, supra note 3, at ¶ 199 (emphasis added).

[41] Id. at ¶ 5.

[42] See infra Section III.

[43] Juozas Kaziukenas, Google Shopping Is Again an E-Commerce Aggregator, Marketplace Pulse (Apr. 28, 2020), https://www.marketplacepulse.com/articles/google-shopping-is-again-an-e-commerce-aggregator.

[44] See Mohammad. Y, Instagram Commerce Statistics and Shopping Trends in 2023, OnlineDasher (last updated Sep. 19, 2023), https://www.onlinedasher.com/instagram-shopping-statistics.

[45] Id.

[46] Checkout on Instagram, Instagram for Business (last visited Dec. 7, 2023), https://business.instagram.com/shopping/checkout.

[47] See Shopify Fulfillment Network, Shopify (last visited Dec. 6, 2023), https://www.shopify.com/fulfillment; Outsourced Fulfillment, ShipBob (last visited Dec. 7, 2023), https://www.shipbob.com/product/outsourced-fulfillment.

[48] Promedica Health Sys., Inc. v. Fed. Trade Comm’n, 749 F.3d 559, 565 (6th Cir. 2014).

[49] Id. at 567.

[50] Id. (quoting 2B Areeda, Antitrust Law, ¶ 565c at 408).

[51] See EU Commission, Universal Music Group / EMI Music, Case No. COMP/M.6458, Decision, 21 September 2012, ¶¶ 141-58.

[52] See, e.g., In the Matter of HCA Healthcare/Steward Health Care System, FTC Docket No. 9410 (Jun. 2, 2022), available at https://www.ftc.gov/legal-library/browse/cases-proceedings/2210003-hca-healthcaresteward-health-care-system-matter.

[53] U.S. v. Philadelphia Nat. Bank, 374 U.S. 321, 356 (1963) (“PNB”) (“We agree with the District Court that the cluster of products (various kinds of credit) and services (such as checking accounts and trust administration) denoted by the term ‘commercial banking,’ composes a distinct line of commerce.”).

[54] Universal Music Group / EMI Music, supra note 51, at ¶ 141.

[55] PNB, 374 U.S. at 356 (emphasis added).

[56] United States v. Philadelphia National Bank, 201 F. Supp. 348, 361 (E.D. Pa. 1962).

[57] Id. at 363.

[58] In the Matter of Penn State Hershey Medical Center and Pinnacle Health System, FTC Docket No. 9368 (Dec. 7, 2015), available at https://www.ftc.gov/system/files/documents/cases/151214hersheypinnaclecmpt.pdf.

[59] Consent Brief of Amici Curiae Economics Professors in Support of Plaintiffs/Appellants Urging Reversal, FTC v. Penn State Hershey Medical Center, et al., Case No. 16-2365 (3rd Cir., Jun. 8, 2016), available at https://www.hbs.edu/ris/Profile%20Files/Amicus%20Brief%20in%20re%20Hershey-Pinnacle%20Proposed%20Merger%206.2016_e38a4380-c58b-4bb4-aecd-26fc7431ecba.

[60] Fed. Trade Comm’n v. Penn State Hershey Med. Ctr., 838 F.3d 327 (3d Cir. 2016).

[61] Complaint, FTC v. Staples Inc. and Office Depot, Inc., Case No. 1:97CV00701 (D.D.C., Apr. 10, 1997), available at https://www.ftc.gov/legal-library/browse/cases-proceedings/9710008-staples-inc-office-depot-inc.

[62] See Orley Ashenfelter, David Ashmore, Jonathan B. Baker, Suzanne Gleason, & Daniel S. Hosken, Empirical Methods in Merger Analysis: Econometric Analysis of Pricing in FTC v. Staples, 13 Int’l J. Econ. of Bus. 265 (2006).

[63] F.T.C. v. Staples, Inc., 970 F. Supp. 1066 (D.D.C. 1997).

[64] And, for at least one court, this is the only basis on which a cluster market is appropriate. See Green Country Food v. Bottling Group, 371 F.3d 1275, 1284 (10th Cir. 2004) (“A cluster market exists only when the ‘cluster’ is itself an object of consumer demand.”) (citing Westman Comm’n Co. v. Hobart Int’l, Inc., 796 F.2d 1216, 1221 (10th Cir. 1986) (rejecting cluster market approach where cluster was not itself the object of consumer demand)).

[65] For example, successful Chinese food product startup Fly By Jing was started by one woman in 2018. She sells only her own products and does so not only on Amazon, but also on her own website and, among countless other places, Costco. See Fly By Jing Amazon Storefront, Amazon.com (last visited Dec. 8, 2023), https://www.amazon.com/stores/page/F2C02352-02C6-4804-81C4-DEA595C644DE; Fly By Jing (last visited Dec. 8, 2023), https://flybyjing.com/shop; Fly By Jing (@flybyjing), Instagram (Feb. 22, 2022), https://www.instagram.com/reel/CaSnvVzlkUW/ (“Sichuan Chili Crisp Now in Costco”).

[66] Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 488 (1977) (quoting Brown Shoe, 370 U.S. at 320).

[67] See Gilman, Albrecht & Manne, supra note 13.

[68] Amazon Complaint, supra note 3, at ¶ 119.

[69] Id. at ¶ 9.

[70] Id. at ¶ 215.

[71] Id. at ¶ 269.

[72] Id. at ¶ 218.

[73] 138 S. Ct. 2274 (2018) (“Amex”).

[74] See, e.g., Michael Katz and Jonathan Sallet, Multisided Platforms and Antitrust Enforcement,127 Yale L.J. 2142 (2018). Katz and Sallet criticize the concept of treating both sides of a two-sided market in one relevant market: “Because users on different sides of a platform have different economic interests, it is inappropriate to view platform competition as being for a single product offered at a single (i.e., net, two-sided) price.” Id. at 2170. But they also contend that effects on both sides must be considered: “[In order] to reach sound conclusions about market power, competition, and consumer welfare, any significant linkages and feedback mechanisms among the different sides must be taken into account.” Id.

[75] Amex, 138 S. Ct. at 2280.

[76] See Amazon Complaint, supra note 3, at ¶¶ 119, 176, 179, 209, 215, & 217.

[77] Lina Khan, The Supreme Court Just Quietly Gutted Antitrust Law, Vox (Jul. 3, 2018), https://www.vox.com/the-big-idea/2018/7/3/17530320/antitrust-american-express-amazon-uber-tech-monopoly-monopsony (“On the surface, the Court’s language [in Amex] suggests that the special rule would apply to Amazon’s marketplace for third-party merchants.”).

[78] PNB, 374 U.S. 321; United States v. Topco Associates, Inc., 405 U.S. 596 (1972) (“Topco”).

[79] United States, et al. v. Am. Express Co., et al., 88 F. Supp. 3d 153, 171 (E.D.N.Y. 2015).

[80] Id., 88 F. Supp. 3d at 247 (citing Topco, 405 U.S. at 610; PNB, 374 U.S. at 370).

[81] Amex, 138 S. Ct. at 2303 (quoting Topco, 405 U.S. at 611).

[82] See Geoffrey A. Manne, In Defence of the Supreme Court’s ‘Single Market’ Definition in Ohio v American Express, 7 J. Antitrust Enf. 104, 115-17 (2019) (“The Court in Topco cited PNB in dictum, not for a doctrinal proposition relating to the operation of the rule of reason, but for a general, conceptual point about the asserted difficulty of courts adjudicating between conflicting economic rights. . . . Nowhere does the Court in Topco suggest that it is inappropriate within a rule-of-reason analysis to weigh out-of-market efficiencies against in-market effects.”).

[83] Ohio v. Am. Express Co., 138 S. Ct. at 2280 (“Thus, credit-card networks are a special type of two-sided platform known as a ‘transaction’ platform. The key feature of transaction platforms is that they cannot make a sale to one side of the platform without simultaneously making a sale to the other.”) (citations omitted).

[84] See, e.g., Amazon Complaint, supra note 3, at ¶ 222 (“Amazon’s restrictive all-or-nothing Prime strategy artificially heightens entry barriers because rivals and potential rivals cannot compete for shoppers . . . solely on the merits of their online superstores or marketplace services. Instead, they must enter multiple unrelated industries to attract Prime subscribers away from Amazon or incur substantially increased costs to convince Prime subscribers to sign up for a second shipping subscription or otherwise pay for shipping a second time. This substantial expense significantly constrains the number of firms who have any meaningful chance to compete against Amazon and raises the costs of any that even try. . . . Amazon’s restrictive strategy artificially heightens barriers to entry, such that an equally or even a more efficient or innovative rival would be unable to fully compete by offering a better online superstore or better online marketplace services.”).

[85] See Brian Albrecht, Is Amazon’s Scale a Harm?, Truth on the Market (Oct. 13, 2023), https://truthonthemarket.com/2023/10/13/is-amazons-scale-a-harm/.

[86] Amazon Complaint, supra note 3, at ¶¶ 214 & 216.

[87] In Aspen Skiing, the “jury found that the relevant product market was ‘[d]ownhill skiing at destination ski resorts,’” Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585, 596 n.20 (1985). The conduct at issue, however, occurred on the input side of the market.

[88] Id. at 605 (emphasis added).

Continue reading
Antitrust & Consumer Protection

The Antitrust Assault on Ad Tech: A Law & Economics Critique

ICLE White Paper For years, regulators and competition watchdogs have expressed concern about competition in the digital advertising business. They note that digital advertising appears to be dominated by a few dominant firms, such as Google, Facebook, and—to a lesser extent—Amazon.

Executive Summary

For years, regulators and competition watchdogs have expressed concern about competition in the digital advertising business. They note that digital advertising appears to be dominated by a few dominant firms, such as Google, Facebook, and—to a lesser extent—Amazon. Some claim that this dominance allows these firms—and Google, in particular—to engage in anticompetitive conduct to extend their market power and to earn supercompetitive profit at the expense of advertisers, publishers, and consumers. This paper investigates the digital advertising market and assesses some of these claims. We conclude based on the information that is publicly available that many of the most significant claims made against Google’s advertising technology (ad tech) business are based on a misunderstanding of U.S. antitrust law, or of the details of the  ad tech market itself.

Digital advertising provides the economic underpinning for much of the Internet. Targeted digital advertising on independent websites is often facilitated by intermediaries that match advertisers and websites automatically, displaying ads to users for whom they are most relevant. This intermediation has advanced enormously over the past three decades. Some now allege, however, that the digital advertising market is monopolized by its largest participant: Google.

In particular, a lawsuit filed in December 2020 by the State of Texas and nine other U.S. states (later joined by five more states in March 2021) alleges anticompetitive conduct related to Google’s online display-advertising business. It has been reported that the U.S. Justice Department (DOJ) may bring a similar lawsuit before the end of 2022.

Moreover, a bipartisan group of U.S. senators introduced the Competition and Transparency in Digital Advertising Act in May 2022. If passed by Congress and signed into law, the bill would require some of the largest Internet firms—such as Google, Facebook, and Amazon—to break up their digital advertising businesses. A summary of the bill claims that Google is the “leading or dominant” firm in “every part” of the ad tech “stack” and that it uses this dominance to “extract monopoly rents” from advertisers and publishers.

Our paper focuses on the open-display digital advertising business. Display ads are designed to be visually engaging, combining text, images, and a hyperlink to a website. These are distinct from search ads, which are text ads displayed along with organic Internet search results. Most of today’s digital display advertising appears on heavily trafficked owned-and-operated sites such as TikTok, Instagram, YouTube, and Amazon, in which the company providing the advertising space is the same company that operates the platforms that place the ads. In contrast, open-display space is supplied by independent publishers—such as The New York Times (nytimes.com), MLB.com, or The Weather Channel (weather.com)—and is usually facilitated by intermediaries.

This paper begins with an overview of digital advertising. The market’s history is one of dynamic innovation, with many new developments arising to solve problems created by previous innovations, address new innovations, and respond to market developments. The market’s structure has changed dramatically as advertisers, publishers, and consumers have responded to new technologies. These changes and innovations all must balance the competing demands of advertisers, publishers, and consumers to maximize the total value of the advertising platform. Thus, any antitrust evaluation of digital advertising must consider whether the market’s structure and the conduct of its participants may be procompetitive responses to prior market changes, as well as the extent to which the overall market structure may mitigate superficially problematic elements within it.

Of particular importance, digital advertising intermediaries that are vertically integrated into some or all components of the digital advertising “stack” of services use the prices charged to each side of the market to optimize overall use of the platform. As a result, pricing in these markets operates differently than pricing in traditional markets: pricing on one side of the platform is often used to subsidize participation on another side of the market, increasing the value to all sides combined. Consequently, pricing (or other terms of exchange) on one side of the market may appear to diverge from the competitive level when viewed for that side alone. While one side of the market may pay superficially higher fees, this cost can be offset by the benefits from increased participation on the other side of the market. In this way, using subsidies to increase participation on another side of the market creates valuable network effects for the side of the market facing the higher fees.

In the second half of the paper, we address some of the specific allegations of anticompetitive structures and conduct made in the Texas Complaint and by critics of the digital advertising industry. We conclude that a flawed premise underlies many of these allegations. It is a version of the “big is bad” argument, in which conduct by dominant incumbent firms that makes competition more difficult for certain competitors is viewed as inherently anticompetitive—even if the conduct confers benefits on users. Under this approach, the largest firms are seen as acting anticompetitively if they do not share their innovations or reveal their business processes to competing firms. As a result, creating new and innovative products, lowering prices, reducing costs through vertical integration, and enhancing interoperability among existing products is miscast as anticompetitive conduct.

In contrast, we note that U.S. antitrust law is intended to foster innovation that creates benefits for consumers, including innovation by incumbents. The law does not proscribe efficiency-enhancing unilateral conduct on the grounds that it might also inconvenience competitors, or that there is some other arrangement that could be “even more” competitive.

Moreover, U.S. antitrust law does not second guess unilateral conduct simply because it may hinder rivals. Any such conduct would first have to be shown to be anticompetitive—that is, to harm consumers or competition, not merely certain competitors. In multisided markets, this means finding not simply that some firms on one side of the market are harmed, but that the combined net effect of challenged conduct across all sides of the market is harmful. The Texas Complaint, however, is built on the alleged harm of reduced revenue to publishers, without considering the corresponding benefit of lower prices to advertisers (and the consumers of advertised products and services).

Based on the information publicly available, we conclude that many of the most significant claims made against Google’s ad tech business are based on a misunderstanding of U.S. antitrust law, or of the details of the ad tech market itself. Although we cannot be sure how the Texas et al v. Google case will develop once its allegations are fleshed out into full arguments, many of its claims and assumptions appear wrongheaded. If the court rules in favor of these, the result will be to condemn procompetitive conduct and potentially to impose costly, inefficient remedies that function as a drag on innovation.

Legislators, too, who may be concerned about Google’s conduct and tempted to impose regulatory requirements on tech companies should bear in minds the risk of the “Nirvana fallacy,” in which real-life conduct is compared against a hypothetical “competition-maximizing” benchmark and anything that falls short is deemed worthy of intervention. That fanciful approach would pervert businesses’ incentives to innovate and compete and would make an unobtainable “perfect” that exists only in the minds of some economists and lawyers the enemy of a “good” that exists in the market.

Introduction

For years, regulators and competition watchdogs have expressed concern about competition in the digital advertising business. They note that digital advertising appears to be dominated by a few exceptionally large firms, such as Google, Facebook, and—to a lesser extent—Amazon. Some claim that this dominance allows these firms—and Google, in particular—to engage in anticompetitive conduct to extend their market power and to earn supercompetitive profits at the expense of advertisers, publishers, and consumers. This paper investigates the digital advertising market and assesses some of these claims. We conclude, based on the information that is publicly available, that many of the most significant claims made against Google’s advertising technology (ad tech) business are based on a misunderstanding of U.S. antitrust law, or of the details of the ad tech market itself.

Digital advertising provides the economic backbone for much of the Internet. By providing websites and apps a means to monetize their products without having to charge user fees, advertising enables access to entertaining and informative content without payment. Targeted advertising allows companies to inform potential customers of new products, giving new entrants a way to compete with popular incumbents, while effective targeting avoids wasting the time of those who aren’t likely to be interested. Advertising can endow products with new characteristics in customers’ minds and make consumers aware of product features they may not have known about.

Advertising on independent websites is often facilitated by intermediaries that match advertisers and websites automatically, targeting ads at the users to whom they are most relevant. This intermediation has advanced enormously over the past three decades. Some now allege, however, that the digital-advertising market is monopolized by its largest participant: Google. These allegations originate from various sources, including policy discussions, lawsuits, draft legislation, and academic reports.

In particular, a lawsuit filed in December 2020 by the State of Texas and nine other U.S. states (later joined by five more states in March 2021) alleges, among other things, anticompetitive conduct related to Google’s online display-advertising business.[1] This action is one of three currently pending lawsuits brought against Google by government antitrust enforcers in the United States; the other two relate to Google’s distribution agreements and search design.[2] It has been reported that the U.S. Department of Justice (DOJ) may bring a similar lawsuit before the end of 2022.[3] Along similar lines, the European Commission opened an investigation into Google’s display-advertising services in June 2021[4] and the German competition authority has published a report regarding its inquiry into non-search advertising.[5] Among other things, the European Commission is investigating whether Google “has made it harder for rival online advertising services to compete in the so-called ad tech stack.”[6]

These ongoing cases follow regulatory reports and hearings examining the market, including a year-long study by the United Kingdom’s Competition and Markets Authority (CMA). The CMA investigation of digital advertising (including search and social-media advertising) has thus far produced recommendations for a code of conduct and “pro-competitive interventions” into the market, as well as a new regulatory body to oversee these measures.[7] The Australian Competition and Consumer Commission is also conducting its own study of the digital advertising market,[8] and both houses of the U.S. Congress have held hearings on the market in recent years.[9] In October 2020, the Democratic majority staff of the U.S. House Judiciary Committee’s Subcommittee on Antitrust, Commercial, and Administrative Law issued a report that recommended, among other things, regulation for the display advertising market.[10]

The digital  advertising industry has also drawn the attention of legislators. In May 2022, a bipartisan group of U.S. senators introduced a bill that would require some of the largest Internet firms to break up their digital advertising businesses.[11] Dubbed the Competition and Transparency in Digital Advertising Act (CTDAA), the measure was introduced by Sen. Mike Lee (R-Utah) and co-sponsored by Sens. Amy Klobuchar (D-Minn.), Ted Cruz (R-Texas), and Richard Blumenthal (D-Conn.). Sen. Lee’s summary of the CTDAA identifies several allegations against the largest firms in the digital advertising business, with an emphasis on Google.[12] The summary claims that Google is the “leading or dominant” firm in “every part” of the “ad tech stack” and that it “exploits” conflicts of interest to “extract monopoly rents” from advertisers and publishers. Because of these monopoly rents, consumers are harmed in the form of higher prices for advertised goods and services and lower quality of online content, according to Lee’s press release.[13]

A 2020 paper published by the Omidyar Network—itself based on an interim CMA report produced during the authority’s then-pending investigation[14]—alleged anticompetitive practices within Google’s display advertising business and laid out a “roadmap” for a prospective antitrust case.[15] Other legal and economic consultants have also voiced concerns about Google’s role in the digital advertising industry.[16] These critiques were published before the Texas Complaint and provide more detail underlying the allegations and arguments described in the Texas Complaint. For that reason and because it may influence further lawsuits and regulatory interventions in the digital advertising market, including the DOJ’s, we also evaluate several of the Roadmap’s findings and conclusions.

This paper investigates the digital advertising market and assesses the aforementioned claims that it is uncompetitive. It explains some of the complex dynamics that underpin this market, thereby shedding light on the weaknesses and deficiencies in many of the arguments made about it, particularly those behind the Texas Complaint. This analysis is relevant to the entire Internet and to the wider economy, not just to Google and the display-advertising market. Many of the allegations made in the Texas Complaint would, if upheld by a court, have profound implications for antitrust law by establishing precedents that successful platforms effectively have a legal duty to act as essential facilities for their competitors; that efficiency-enhancing innovation by incumbent platforms is anticompetitive (particularly when it is not shared with competitors); and that courts or regulators can impose remedies that put these duties into effect without consideration of the harmful tradeoffs and unforeseen consequences that could themselves constrain competition and innovation. Such an approach would severely affect not only Google and the ad tech industry, but also businesses operating in unrelated markets and industries.

We begin with an overview and history of digital advertising. It is a history of dynamic innovation, with many new developments arising to solve problems created by previous innovations, address new innovations, and respond to market developments. These innovations must balance the competing demands of advertisers and publishers. The market’s structure has changed dramatically as advertisers, publishers, and consumers have responded to recent technologies. Because of this dynamism, we argue that it is a mistake to conclude that market structure and firm conduct at some specific point in time was ideal or better from a competition point of view, or that deviations from that paradigm represent a problem for competition enforcers to correct.

In the second half of the paper, we address some of the specific allegations of anticompetitive structures and conduct made in the Texas Complaint and by critics of the digital advertising industry. We conclude that a flawed premise underlies many of these allegations. Fundamentally, the allegations stem from an assertion that conduct engaged in by dominant, incumbent firms that makes competition more difficult for competitors is anticompetitive—even if the conduct confers benefits on users. This often amounts to a claim that the largest firms are acting anticompetitively by innovating and developing their business processes and products in ways that create benefits for one or more digital advertising constituents and for the advertising ecosystem more generally, including by creating new and innovative products, lowering prices, reducing costs through vertical integration, and enhancing interoperability between existing products.

I. Overview of the Digital Advertising Industry

For many people, digital ads are “just there.” They appear on one’s Facebook timeline, get inserted into one’s Twitter feed, or show up in the middle of an online news article. Unseen to most users is a complex stack of services that match advertisers with advertising space, using real-time auctions and other algorithms to deliver ads targeted to produce a user response, such as buying a product, supporting a cause, or visiting another website.

In this section we explain that digital advertising is just one small part of a much broader advertising and marketing industry. We provide a concise history of how the business evolved from simple banner ads to highly targeted display ads. Through this evolution, digital advertising has become a multisided market where intermediaries must balance the demands of advertisers, publishers, and users to maximize the total value of the advertising platform. Because of this balancing, it is a mistake for policymakers and regulators to focus only on a single set of users or a single segment of the stack of digital advertising services.

A.   Digital advertising is part of a broader advertising and marketing market

The Texas Complaint alleges that Google has market power in six distinct product markets, each of which the states claim to be nationwide in geographic scope. Four of these distinct product markets involve intermediation in the sale of “open” display ads on third-party websites and two involve intermediation in the sale of “in-app” display ads on mobile devices.[17] As we note in our earlier paper, it is likely that the states allege overly narrow product-market definitions.[18] In particular, structure and conduct viewed within a broader digital advertising market, overall advertising market, or marketing market indicates than no single firm has significant market power.

Digital advertising comprises about half of U.S. advertising revenues (Figure 1), while advertising itself accounts for about half of marketing activities. Marketing includes advertising, as well as events, sales promotion, direct marketing, telemarketing, product placement, and merchandising. Within digital advertising, advertisers have a broad set of options about where and how to run ads, including:

  • Search ads, in which the ad is displayed as a search-engine result (e.g., Google, Bing, DuckDuckGo);
  • Display ads on a site owned and operated by the firm that sells the ad space (e.g., Facebook, YouTube, Amazon Marketplace);
  • “Open” display ads on a third party’s site (e.g., The New York Times, Dallas Morning News, Runner’s World); or
  • “In-app” display ads served on mobile apps.

While total advertising spending in the United States has increased by about 15% since 2009, as a share of the economy spending has been relatively flat at slightly less than 1% of GDP.[19] As mentioned, about half of total U.S. advertising dollars go to digital channels, up from approximately 10% a decade ago. Approximately 30% of ad spending goes to TV and less than one-quarter goes to radio, newspapers, magazines, billboards, and other “offline” forms of media.[20]

Figure 1: U.S. Advertising Spending Over Time

Source: Benedict Evans, News by the Ton

It is well-understood that television broadcasters and cable networks compete with digital services.[21] And they do so on virtually all dimensions: for user attention, for labor, for content and other inputs—and for advertising. The same is true of competition for advertising among digital publications, newspapers, radio, magazines, video games, music streaming, and podcasts. The fact that offline and online advertising—to say nothing of marketing more broadly—employ distinct processes does not consign them to separate markets. Indeed, it is widely accepted that online advertising has drawn advertisers from offline markets, as previous technological innovations drew advertisers from other channels before them.[22] Moreover, while there is evidence that, in some cases, offline and online advertising may be complements as well as substitutes,[23] the distinction between these cases is becoming less and less meaningful as the revolution in measuring the effectiveness of advertising has changed how marketers approach different levels of what is known as the marketing “funnel.”[24] For example, economist David Evans’ review of the literature concludes that digital advertising is a segment of the broader advertising business in which different forms of advertising compete and complement each other:

Advertisers base decisions about the level and allocation of their budgets on formal or informal analyses of the rate of return on investment. For these ad campaigns, the different advertising methods can be substitutes to the extent they provide alternative ways of delivering messages to an audience, and complements to the extent they can reinforce each other.[25]

Economists Avi Goldfarb and Catherine Tucker demonstrate that display-advertising pricing is sensitive to the availability of offline alternatives.[26] Although advertising technology and both supplier and consumer preferences continue to evolve, the weight of evidence suggests a far more unified and integrated economically relevant market between offline and online advertising than their common semantic separation would suggest:

We believe our studies refute the hypothesis that online and offline advertising markets operate independently and suggest a default position of substitution. Online and offline advertising markets appear to be closely related. That said, it is important not to draw any firm conclusions based on historical behavior.[27]

In summary, there is evidence that open-display and in-app ads compete with search ads, while digital ads compete with offline advertising. Thus, courts and regulators should be skeptical of overly narrow market definitions focused on only small slices of a much larger relevant market for advertising.

B.   A simplified description of digital display advertising

The combination of software and processes that facilitate digital advertising transactions is known as the “ad tech stack.” The stack consists of several software components to match advertisers with publishers.

Advertiser ad servers are used by advertisers and media agencies to store ads, deliver them to publishers, track their activity, and assess their effectiveness (by, for example, tracking conversions). Demand-side platforms (DSPs) automate the purchase of advertising inventory by collecting bids in real-time auctions from multiple advertiser ad servers. DSP bids are based on the advertiser’s objectives, data about the final user, and data on impressions or conversions. Publisher ad servers manage publishers’ inventory and determine whether and where to serve a particular ad on a publisher’s site. Supply-side platforms (SSPs) automate the sale of publishers’ inventory, typically through real-time auctions involving multiple DSPs.

In general, the process of buying and selling digital ads through open-display auctions works as follows (Figure 2):

  1. When a user opens a webpage (or uses an app), the publisher’s ad server sends a bid request to SSPs for the advertising impressions available on that page for that user.
  2. The SSPs send bid requests to multiple DSPs for the ad impressions.
  3. DSPs evaluate the advertising opportunity based on user data and the objectives of their advertisers’ campaigns and send bids to the SSPs.
  4. SSPs rank the bids received based on price and other priorities set by the publisher. The SSPs send their winning bids to the publisher.
  5. The publisher ad server compares bids received from SSPs, along with any pre-existing direct deals between the publisher and specific advertisers and decides which ad to serve on the page.

Figure 2: A Simplified View of the Digital Advertising Stack

Source: OECD, CMA

While this process applies to most programmatic transactions, there are many variations. For example, there are diverse ways in which SSPs are contacted and asked to submit their bids. To the extent that a publisher and advertiser have a pre-existing, direct agreement, there are differences in how these arrangements are handled and integrated with deals arranged through automated platforms. The specific approach used to match ads with ad inventory will reflect a balance among different sides of a multisided market. One approach might increase the prices received by sellers (publishers) but expose buyers (advertisers) to increased risk of overpayment. Other methods might reduce risks to advertisers, but also reduce the prices received by publishers.

C.   A brief history of digital advertising

This history of the digital advertising market is a history of iterative innovation, with new developments and services arising to solve problems created by previous innovations and to respond to changing market conditions. At the heart of these innovations is an attempt to balance the competing demands of advertisers, publishers, and consumers. Given that this is a dynamic market, it would be mistaken to conclude that the market structure at some specific point was the “correct” one from a competition point of view. Moreover, it would be a mistake to conclude that deviations from some previous “ideal” world present a problem that can be corrected by disruptive regulation.

Digital advertising originally worked similarly to conventional print and broadcast advertising. Online publishers would negotiate with advertisers (or their ad agencies) to sell ad space on their websites, giving relevant advertisers information about their readership gathered through market research. All users would see the same ads. The resulting system was poorly targeted, inefficient, and carried high fixed costs, including the cost of things like market research and the transaction costs of publishers hiring sales teams and advertisers hiring ad agencies to do business with one another. Inevitably, these fixed costs meant that only larger publishers and larger advertisers could engage in the online market profitably.

In 1993, O’Reilly & Associates Inc. introduced its Global Network Navigator (GNN) magazine and other ad-supported online publications, which first rolled out clickable ads. O’Reilly is credited with the first attempt to create an “advertising medium” on the Internet.[28] The price of ads ranged from $500 for a one-page business profile to $5,000 for up to 25 pages about the company placing the ad.[29] A year later, Wired magazine’s digital affiliate HotWired ran what later became known as the web’s first banner ads. The ad—for AT&T’s “You Will” campaign—cost $30,000 for a three-month dedicated placement in a section of HotWired’s site.[30]

The first step toward automating this process came with the introduction of ad-server software on both the publisher and advertiser sides, which allowed each side to automate parts of the ad-placement process. Ad servers allowed publishers to automatically describe the type of content on their pages, which in turn allowed advertisers to place ads tailored to that content. An article about hiking could automatically indicate to a department store to place ads selling walking boots, for instance. It also allowed the publisher to sell to many advertisers without having to transact directly with any of them. Ad servers also allowed advertisers to browse and manage campaigns across a large, aggregated number of publishing sites, instead of having to interact with sales teams one by one. This process was, however, still negotiated directly, and often left publishers with unused “remnant” advertising space that they had not been able to sell.

To solve this problem, ad networks entered the market. These functioned as intermediaries between advertisers’ and publishers’ ad servers, aggregating unsold remnant ad space and allowing advertisers to buy that space en masse without having to deal directly with each publisher. Ad networks did not replace direct sales, but they allowed for residual space to be bought and sold more easily, increasing the amount of ad inventory available and lowering the fixed costs to use it. This, in turn, made the market feasible for smaller publishers, who would otherwise be unable to attract direct deals to sell ad space, and for advertisers to conduct large-scale ad campaigns across many publishers (including small ones).

In 1995, GNN was acquired by AOL.[31] That same year, marketing-communications agency Poppe Tyson spun off its Internet advertising division as DoubleClick, with the objective of “responding to advertisers’ need to be able to buy millions of impressions on the Internet without having to buy from hundreds of different sites.”[32] The company created “subnets” of publishers to target specific categories of consumers.[33]

Also in 1995, ad agency WebConnect, the first ad network, began to collaborate with its clients to identify the websites that their ideal consumers visit. WebConnect then placed ads on the websites where they were more likely to be seen by the audience most relevant to their clients. The company also produced a tool to prevent “ad fatigue,” which occurs when users are repeatedly shown the same ad.[34]

In 1998, GoTo.com, which was renamed Overture in 2001, launched its ad-supported search engine.[35] Search result rankings were based on an open-market bidding process. Advertisers on GoTo were informed of the amounts other advertisers were bidding for a click-through within the results for a given search term, and any advertiser could increase its bid to obtain a higher ranking, a process GoTo described as “pay for performance.”[36] One of GoTo’s key innovations was linking advertising pricing to click-throughs, rather than to page views.[37]

To drive home just how efficient these ads could be, [GoTo.com’s founder Bill Gross] came up with an audacious pricing scheme. Instead of paying for page-views—an old-media model that had come to dominate the Web—advertisers would pay only when people actually clicked on their ads. And their placement on the GoTo.com results page would be determined through an auction, so that more desirable keywords would command higher prices, while less common keywords could be had for as little as a penny per click. As a search engine, GoTo.com had nothing on Google. But as a way of making money on searches, it was ingenious.[38]

During the dot-com boom of the late 1990s, banner ads spread throughout the Internet, though growth was tempered by user complaints that the ads slowed page loading.[39] The dotcom bust wiped out many of the firms that were the biggest buyers of digital banner ads. In response, Wired predicted that digital advertising would undergo a “facelift.”[40]

Old media revenues will wither as mainstream advertisers storm the Net. Instead of stuffing junk mail into the mailbox outside your house, they’ll send it directly to your inbox. And companies get smarter, choosing sites that take better aim at their quarry, er, potential customers.

“It’s very much a targeted medium,” Robin Neilfield, co-founder of NetPlus Marketing. “You have to analyze the content on a site, you can’t just buy based on demographics.”[41]

Because ad networks were not comprehensive—they did not carry the entire inventory of the Internet—publishers began to use yield managers (later known as SSPs) to compare bids for their ad space and to decide which to accept. SSPs effectively allowed publishers to aggregate demand from a larger number of ad networks, which themselves aggregated demand from advertisers. This process allowed ad space to be more easily commoditized, with an SSP helping to identify an ad space’s relevance to potential advertisers.

As indirect sales became possible, ad exchanges emerged to sell ad space using real-time auctions. Ad space could be tagged according to characteristics like audience type, relevance to the advertiser, and/or prominence and quality of the ad, with bids gauged accordingly.

Finally, DSPs on the advertiser side allowed advertisers to engage with many ad networks and ad exchanges at one time. These also allowed advertisers to track campaigns and measure performance of different ads with different publishers, and to adjust their campaigns accordingly. Most ad exchanges now have DSP functionality built in.

In 2000, Google introduced a new self-service advertising product called AdWords (now Google Ads) that allowed businesses to purchase text ads on search-results pages. At the time it was reported that AdWords was designed to attract small-to-midsize advertisers with budgets of $5,000 or less.[42] AdWords differed from GoTo/Overture in a major way. GoTo/Overture placed ads within search results, with results ranked by bid. In contrast, AdWords placed ads separate from search results with pricing based on pageviews.[43] In this way, Google could display ads without compromising the relevance of search results. In 2002, it launched AdWords Select, its pay-per-click, auction-based search-advertising product.[44]

In 2003, Google acquired Applied Semantics, whose AdSense display advertising product allowed it to sell targeted ads on third-party websites.[45] With AdSense, the display-ad server was able to read text on a publisher’s site and serve relevant ads, considering factors like the user’s geographic location, age, demographics, and the search made.[46] AdSense was the forerunner of programmatic display advertising, the process of automating the buying and selling of ad inventory in real time through an automated bidding system. In 2005, Google introduced the Quality Score model, which considers an ad’s click-through rate, as well as the bid price, in placing ads.[47]

YouTube was launched in 2005 and acquired by Google the following year, when the company also introduced video ads. In 2007, Google acquired DoubleClick.[48]

Around 2015, “header bidding” began to roll out, with publishers Meredith Corp. and Townhall Media as two of largest early adopters.[49] Before header bidding, it was difficult for every demand-side partner to submit a bid for every ad request. As a result, publishers relied on approaches such as “ad waterfalls”[50] to try to get the most from each partner. Because of the way ad waterfalls are configured (based on historical, not real-time, data), publishers believed ad waterfalling led to winning bids that were below what some bidders might be willing to pay.[51] Client-side[52] header bidding was adopted as a way to increase real-time price competition among multiple SSPs, leading to higher returns for publishers and a more efficient allocation of ad space to advertisers.

Despite the widespread adoption of header bidding—as of the second quarter of 2021, about two-thirds of publishers were using it[53]—the technology has its own challenges. For example, the addition of extra code on the webpage, which client-side header bidding requires, can slow down the publisher’s website, driving away users.[54]

As an alternative to client-side header bidding, server-side header bidding was introduced. Prebid launched in 2015 as an independent and open-source option. Google released Open Bidding in April 2016 and Amazon introduced Transparent Ad Marketplace (“TAM”) at the end of 2016. In these alternatives, the auction among SSPs takes place in a remote server controlled by a third party (the provider of the server-side header-bidding solution) instead of in the user’s browser. This helps to improve site-load speed. On the other hand, this solution leads to less revenue for publishers and reduces the availability of data to advertisers and publishers.[55]

Over the past decade, the price of digital advertising has fallen steadily, while output has risen. U.S. digital-ad spending grew from $26 billion in 2010 to $189 billion in 2021, an average annual increase of 20%.[56] Over the same period, the Producer Price Index for Internet-advertising sales declined by an annual average of 4%.[57] The rise in spending in the face of falling prices indicates that the number of ads bought and sold increased by approximately 25% annually. The combination of increasing quantity, decreasing cost, and increasing total revenues are consistent with a growing and increasingly competitive market, rather than one of rising concentration and reduced competition.

D.   Digital advertising is a multisided market

The digital advertising market can be thought of as a complex multi-step and multisided market that involves three key parties—advertisers, publishers, and intermediaries—and is aimed at a fourth: consumers. In contrast, critics of the current structure of and conduct in the digital advertising industry have characterized it as a “straightforward and traditional” market in which publishers supply an inventory of ad space and advertisers are buyers of the ad space.[58] In this simplistic account of the market, for a given supply of inventory, publishers would seek to maximize the price received per ad, while advertisers would seek to minimize the price paid per ad. Targeting of ads would be based on the demographics of a publisher’s readership or the content of the publication, rather than the individual characteristics of each reader. In general, this is how the market initially operated before the introduction of clickable ads. But even this simple formulation is quite complex. Advertisers expect to maximize the return on their investment in advertising. Even at a low price, advertising expenditures would be wasted if that investment were not converted to increased sales of the advertiser’s product or service.

The invention of clickable ads with which users could interact changed the objective function of digital advertising. Publisher revenues and advertising costs became linked to individual consumers acting on an ad by, for example, clicking on it. Rather than paying or receiving a price-per-ad based on the size of a publication’s user base, advertising expenditures became a function of a price-per-click (or other action) and the number of clicks. This meant that the rewards for relevance—as well as the complexities of determining relevance—were greater because some viewers might be persuaded to act there and then.

In this multisided market, ad intermediaries must balance the interests of at least three constituencies: (1) advertisers creating ads and placing them; (2) publishers defining inventory and displaying ads; and (3) users consuming published content who view and act on ads. Intermediaries in these markets often benefit from network effects, through which the value of the platform to each user depends in part on the number and quality of other users on the platform.[59]

The quality and relevance of users is assessed by collecting information on the users as they browse the web. This information can include which ads they have viewed and clicked in the past, their geographical location, as well as their demographics, financial situation, and topics of interest. Broadly speaking, a larger network with diverse users provides more information and is better able to target ads to relevant users, benefiting advertisers, publishers, and consumers.

Network effects are not always positive, however, nor are they always captured by the platform that facilitates them.[60] While access to consumer data can help to improve the quality of the ads displayed—and increase the value of those ads to advertisers and publishers—claims that such access provides increasing returns to scale are not borne out by the burgeoning empirical literature on the topic. Summarizing these empirical findings, economist Catherine Tucker concludes that “empirically there is little evidence of economies of scale and scope in digital data in the instances where one would expect to find them.”[61]

Intermediaries in multisided markets often face difficult optimization problems caused by the interrelated demands of participants on different sides of the market, each group of whom benefits from the existence and size of the other, but whose interests conflict across many margins.[62] This highlights the key distinction between “straightforward and traditional” markets and multisided markets.

Ad tech intermediaries that are vertically integrated into some or all components of the ad tech stack use prices charged to each side of the market to optimize overall use of the platform. As a result, pricing in these markets operates differently than pricing in traditional markets. Pricing on one side of the platform is often used to subsidize participation on another side of the market, increasing the value to all sides combined. Consequently, pricing (or other terms of exchange) may appear to one side of the market to diverge from the competitive level when viewed for that side alone. While one side of the market may pay higher fees, this cost can be offset by the benefits from increased participation on the other side of the market. Thus, using subsidies to increase participation on another side of the market creates valuable network benefits for the side of the market facing the higher fees.

For example, among the criticisms of digital advertising business practices is the use of “second-price auctions” rather than “first-price auctions.”[63] First-price auctions are those most familiar to people: multiple bidders offer prices, and the highest bidder wins the auction and pays an amount equal to her winning bid. In a second-price auction, the highest bidder wins the auction but pays an amount based on the next-highest bid. In markets with many bidders possessing the same information, first-price auctions and second-price auctions would be expected to produce the same amount of revenue under the well-known auction-theory concept of revenue equivalence.[64]

The choice of auction approach reflects the tensions between different sides of the market in a multisided market. On the one hand, under certain circumstances, a first-price auction tends to increase the prices received by sellers (here, publishers), but exposes buyers (here, advertisers) to an increased risk of overpayment.[65] On the other hand, under certain conditions, a second-price auction reduces risks to advertisers, but also reduces the prices received by publishers.[66] It would be expected that an ad tech intermediary would balance these competing interests to maximize total revenues flowing through the ad tech stack, to maximize its profitability. In such a multisided market, it would be a mistake to focus only on one side of the market and ignore the effects that decisions such as this have on the other participants.

The extent to which ad tech intermediaries—in particular, vertically integrated services like Google’s—act to optimize the overall value of the platform is critical to understanding how these markets work. It also highlights how misleading it can be to assume that these processes can be analyzed as “straightforward and traditional” markets.

II. Antitrust Primer: Effective Competition Is not an Antitrust Offense

A flawed premise underlies much of the Texas Complaint, the Omidyar Network’s Roadmap report, and the CTDAA legislation. Fundamentally, most of the charges that each of these level against Google and Facebook’s ad tech businesses derive from an assertion that conduct engaged in by dominant incumbent firms that makes competition more difficult for competitors is anticompetitive—even if the conduct confers benefits on users. This often amounts to a claim that the largest firms are acting anticompetitively by innovating and developing their business processes and products in ways that create benefits for one or more digital advertising constituents and for the advertising ecosystem more generally. These include creating new and innovative products, lowering prices, reducing costs through vertical integration, and enhancing interoperability between existing products, among other things.

This approach entails an argument—made explicit in the Texas Complaint and the Omidyar Roadmap—that Google harms competition by creating obstacles for rivals without offsetting “incremental efficiencies.”67F[67] According to the report, this means that, even if Google’s practices produce benefits for such constituents as advertisers, publishers, or consumers, they could possibly be reimagined to create even more competition or achieve the same benefits in ways that better prop up rivals. According to the Roadmap, the practices should therefore be condemned as anticompetitive.68F[68]

But that is not how the law—or the economics—works. Such an approach converts manifestly beneficial aspects of Google’s ad tech business into anticompetitive defects, essentially arguing that successful competition creates barriers to entry that merit correction through antitrust enforcement. The CTDAA takes this argument a step further by imposing “best interests,” “best execution,” and “transparency” obligations on large firms and mandating divesture of parts of the largest firms to facilitate more entry by competitors. This approach turns U.S. antitrust law (and basic economics) on its head. As some of the most famous words of U.S. antitrust jurisprudence have it:

A market may, for example, be so limited that it is impossible to produce at all and meet the cost of production except by a plant large enough to supply the whole demand. Or there may be changes in taste or in cost which drive out all but one purveyor. A single producer may be the survivor out of a group of active competitors, merely by virtue of his superior skill, foresight and industry. In such cases a strong argument can be made that, although, the result may expose the public to the evils of monopoly, the Act does not mean to condemn the resultant of those very forces which it is its prime object to foster: finis opus coronat. The successful competitor, having been urged to compete, must not be turned upon when he wins.69F[69]

U.S. antitrust law is intended to foster innovation that creates benefits for consumers, including innovation by incumbents. The law does not proscribe efficiency-enhancing unilateral conduct on the grounds that it might also inconvenience competitors, or that there is some other arrangement that could be “even more” competitive. Under U.S. antitrust law, firms are “under no duty to help [competitors] survive or expand.”70F[70]

To be sure, the arguments are couched in terms of anticompetitive effect rather than being described merely as commercial disagreements over the distribution of profits. But these effects are simply inferred, based on assumptions that Google and Facebook’s vertically integrated business models entail an inherent ability and incentive to harm rivals. For example, Google is alleged to be able to surreptitiously derive benefits from display advertisers by “leveraging” its search-advertising capabilities,71F[71] or by “withholding YouTube inventory,”72F[72] rather than altruistically opening it up to rival ad networks, or by using its access to data to improve its products without sharing that data with competitors.

All these charges may be true, but none is inherently anticompetitive. Under U.S. law, companies are not obligated to deal with rivals and certainly are not obligated to do so on rivals’ preferred terms.73F[73] In the Texas Complaint, for example, the court, citing Charych v. Siriusware, noted, “[D]efendants were under no obligation to develop an interface that was compatible with plaintiffs’ product.”74F[74] As long ago as 1919, the U.S. Supreme Court held that “[i]n the absence of any purpose to create or maintain a monopoly, the [Sherman Act] does not restrict the long recognized right of [a] trader or manufacturer engaged in an entirely private business, freely to exercise his own independent discretion as to parties with whom he will deal.”75F[75] More recently (in 2004) the Court held:

Firms may acquire monopoly power by establishing an infrastructure that renders them uniquely suited to serve their customers. Compelling such firms to share the source of their advantage is in some tension with the underlying purpose of antitrust law, since it may lessen the incentive for the monopolist, the rival, or both to invest in those economically beneficial facilities.76F[76]

Moreover, U.S. antitrust law does not second guess unilateral conduct simply because it may hinder rivals. Any such conduct would first have to be shown to be anticompetitive—that is, to harm consumers or competition, not merely certain competitors.77F[77] In two-sided markets, this means finding not simply that some firms on one side of the market were harmed, but rather that the combined net effect of challenged conduct across all sides of the market was harmful.78F[78]

In the platform context, understanding whether there is harm to competition at all requires an assessment of the effects of conduct on all sides of the platform. “[N]o economic basis exists for establishing a presumption that ‘harm’ on one side of a two-sided platform is sufficient to demonstrate that market output has been restricted, or that consumer welfare has otherwise been harmed.” In fact, “[s]eparating the two markets allows legitimate competitive activities in the market for general purposes to be penalized no matter how output-enhancing such activities may be.”79F[79]

The Texas Complaint, however, is built on the alleged harm of reduced revenue to publishers, without considering the corresponding benefit of lower prices to advertisers, or the net effect on consumers.80F[80]

Beyond that, antitrust law does not condemn conduct on the basis that an enforcer (or a court) is able to identify or hypothesize alternative conduct that might provide similar benefits at lower cost. In alleging that there are ostensibly “better” ways that Google could have pursued its product design, pricing, and terms of dealing, both the Texas Complaint and Omidyar Roadmap do just that—assert that, had the firm only selected a different path, an alternative could have produced even more benefits or an even more competitive structure. This line of thinking seems to be one motivation for the CTDAA’s remedies.

The reason that the possibility of “better” theoretical arrangements cannot serve as the basis for antitrust intervention is that there are limits to what can be achieved through intervention, not least because of limitations on legislators’ and enforcers’ knowledge about the competitive dynamics of the markets they seek to regulate.81F[81] A practice’s departure from a theoretical competitive benchmark may be inextricably linked to the social benefits it generates. When this is the case, enforcement that requires the practice or product to change in order to adhere to a theoretical standard may end up undermining the benefits of the practice in the first place. That is particularly true in the context of the sort of “vertical foreclosure” arguments leveled against Google in the advertising space, in which it is alleged that the combination of different levels of the ad-supply chain by Google limits the ability of competitors to enter and compete effectively.82F[82] It is surely conceivable that the product improvements conferred by the combination of different functions into a single platform—e.g., greater efficiency, realization of network effects, more effective targeting—could be replicated by a different means that might also facilitate “even more competition.” But such an approach is fraught with the risk of serious and costly error.83F[83]

The alleged cure of tinkering with benefit-producing unilateral conduct by applying an “even more competition” benchmark is worse than the supposed disease. The adjudicator is likely to misapply such a benchmark, deterring the very conduct the law seeks to promote. As then-Judge Stephen Breyer explained in the context of above-cost low pricing (another “defect” that both the Texas Complaint and the Roadmap claim constitutes anticompetitive conduct by Google84F[84]), “the consequence of a mistake” is “to penalize a procompetitive price cut,” conduct that, from an antitrust perspective, is “the most desirable activity.”85F[85] That commentators or enforcers may be able to imagine alternative, theoretically more desirable, conduct is beside the point.

Similarly, subjecting the kinds of product-design decisions at issue in the Google case to refined balancing of benefits and harms would deter innovation. “To weigh the benefits of an improved product design against the resulting injuries to competitors is not just unwise, it is unadministrable. There are no criteria that courts can use to calculate the ‘right’ amount of innovation, which would maximize social gains and minimize competitive injury.”86F[86] Put simply, “no one can determine with any reasonable assurance whether one product is ‘superior’ to another.”87F[87]

For these reasons, a “product improvement by itself does not violate Section 2, even if it is performed by a monopolist and harms competitors as a result.”88F[88] “Any other conclusion would unjustifiably deter the development and introduction of those new technologies so essential to the continued progress of our economy.”89F[89] A benefit-creating product design, even if it hinders rivals, is “necessarily tolerated by the antitrust laws.”90F[90]

Nor does U.S. law condemn a firm’s decision not to share a product improvement with rivals on terms rivals might prefer, even when such sharing might lead to greater competition in the short term. “Compelling” innovators “to share the source of their advantage” with rivals, among other evils, “may lessen the incentive for the monopolist, or rival, or both” to invest in innovation.91F[91] Except in extremely limited circumstances, firms can decide the terms on which they offer their products and services.

Directly rejecting the Roadmap’s suggestion—and the CTDAA’s mandate—of compelling dealings on terms that might produce greater competition, the Supreme Court has decreed that the “Sherman Act . . . does not give judges carte blanche to insist that a monopolist alter its way of doing business whenever some other approach might yield greater competition.”92F[92] Firms are not obliged to go into new lines of business or abandon existing lines to throw lifelines to rivals.93F[93]

The law similarly encourages vertical integration, because it tends to foster innovation-enhancing synergies and lower prices by eliminating double marginalization.94F[94] As the Roadmap elsewhere admits, it is “not in itself uncommon” to see vertical integration result in “fewer and fewer companies,” even “in competitive markets.”95F[95] Thus, vertical integration by internal expansion—even by a monopolist—is presumptively lawful.96F[96] The Roadmap and the CTDAA, however, simply disregard this, instead presumptively condemning vertical integration that hinders rivals by creating efficiencies.97F[97] Again, this is simply not a defensible interpretation of U.S. antitrust law, nor should it be.

III. Allegations Against Google in Digital Display Advertising

Critics of the digital advertising industry—and Google’s role in it—have leveled numerous allegations. These include claims that Google “leverages” its ownership of YouTube to obtain and exert market power in the buying and selling of other digital-display ads. Some claim that Google anticompetitively uses cross-subsidies, charging supercompetitive prices at one end of the ad tech stack to subsidize supra-competitive prices at another end of the stack. It is also alleged that Google has superior information about consumers that it will not provide to competitors, giving Google an anticompetitive advantage. It is claimed that steep entry barriers—some allegedly erected by Google—inhibit entry and allow the company to achieve a supercompetitive “take rate” from its intermediation services. While the lawsuits may provide additional information and data to support these claims, we argue that, with the limited public information available to us, it is not clear that any of them constitute anticompetitive conduct.

A.   ‘Leveraging’ market power in video streaming into the digital open-display market

The Omidyar Roadmap argues that Google, by virtue of its vertical integration throughout the intermediary stack and into the supply side (as the owner of YouTube), has the incentive and ability to derive unwarranted benefits from its display advertising business. It alleges, for example, that, by offering a single interface for placing both search and display ads, “Google leverages its monopoly position in search to coerce advertisers into using Google’s display products.”98F[98] In support it cites the CMA as saying:

Google may also be able to leverage its market power in search into the open display market. Smaller advertisers often choose to single-home to minimize transaction costs. Advertisers that wish to single-home have a strong incentive to use Google Ads as they can use it to access Google search advertising and YouTube inventory as well as the open display market.99F[99]

An earlier version of the Texas Complaint echoed these claims:

Google’s practice of withholding YouTube video inventory from rival ad buying tools… effectively locks single-homing small advertisers into Google’s ad buying tool. In addition, other providers of ad buying tools indicate that it does not make economic sense to try to compete with Google Ads for small advertisers, because they cannot achieve sufficient scale with smaller advertisers who want to buy display, YouTube, and even search ads, through just one tool.100F[100]

And, similarly, the Roadmap also argues that most sources of demand for Google Ads purchase ad space through AdX because Google “designed its exchange in such a way that it operates more efficiently with requests from Google’s own ad server than it does when requests come from rival ad servers.”101F[101]

All these assertions describe efficiency-enhancing behavior as anticompetitive. The report does not allege that Google preferences its own ad exchange in ways that harm advertisers; rather, the company’s products simply work better together (which is not unusual when different software products must interact) and it is thus in advertisers’ best interests for Google to act this way.

U.S. law, rightly, does not consider efficiencies obtained from vertical integration in this way to be anticompetitive. Nor do efficiencies that rivals cannot beat qualify as “barriers to entry.” The alternative—requiring Google to refrain from using the cheapest and/or fastest option available, because doing so makes its product better than all competitors—would mean reduced innovation, higher overall costs, and no benefit to either advertisers or publishers.

Later, the Roadmap makes another similar allegation: that Google “leverages” its ownership of YouTube, and the fact that only Google’s DSP can place ads on YouTube, to give itself an anticompetitive advantage in open-display advertising because rival DSPs are inherently limited by being unable to place ads on YouTube. An earlier version of the Texas Complaint echoed this claim.102F[102]

The Roadmap characterizes this conduct as “a contractual way to deny interoperability,”103F[103] but there is no contractual restraint here. How Google distributes YouTube’s ad inventory is a unilateral distribution decision permitted under U.S. law. And Google’s policy is not unusual in any way: many other websites that carry video advertising—including Hulu, Instagram, and Twitter—self-distribute their own inventory and do not make it available for resale by third parties.104F[104] Google does not have a duty to maximize its competitors’ profits by allowing them to resell YouTube inventory.

Access to YouTube is also not essential to a DSP’s success. Before Google stopped third-party platforms from buying YouTube ad inventory, it reported that only a “small amount” of buying was being done through Google’s AdX, which allowed third-party platforms to bid. At the time, AdExchanger reported that “[b]y ’small amount,’ that reportedly means 5%.”105F[105] A competing DSP, TubeMogul, said that this decision was an “unfortunate development” but “immaterial, since less than 5% of total ad spend through our software in Q2 was directed to YouTube.”106F[106]

This is consistent with the fact that there are several successful DSP competitors that compete with Google, despite not having access to YouTube’s ad inventory. The Trade Desk went public for $1 billion in 2016, processed more than $6.2 billion in transactions in 2021, and had a market cap of more than $25 billion in the first week of August 2022. 107F[107] Other DSPs, like Amazon’s and Xandr (formerly AppNexus), both continue to compete with Google vigorously without access to YouTube inventory, as the Omidyar Roadmap admits in the case of AppNexus.108F[108]

The Roadmap further -alleges that Google’s owned-and-operated properties—including Search, YouTube, Shopping, Flights, and News—confer an anticompetitive advantage because “Google pays no ‘traffic acquisition costs’” for the advertising space on its own sites: “When Google places ads on YouTube, just as when it places ads on its own search results pages, Google pays no ‘traffic acquisition costs’ because it needn’t pay any publisher for access to the ‘eyeballs’ that will see or interact with the ads it helps place.”109F[109]

Google’s parent Alphabet reported that the company’s traffic-acquisition costs were approximately 20% of its revenues in 2021.110F[110] Over the past few years, 40-50% of Alphabet’s expenditures have been on “cost of revenues,”111F[111] and of these, roughly half have involved traffic-acquisition costs.112F[112] Alphabet defines traffic-acquisition costs as (a) “the amounts paid to our distribution partners who make available our search access points and services” and (b) amounts paid for ads displayed on Google Network Members properties. It identifies “distribution partners” as browser providers, mobile carriers, original equipment manufacturers, and software developers.

Contrary to the Roadmap’s insinuations, there is nothing to suggest that these expenditures become less burdensome as a company increases in scale. Indeed, the opposite may be true, if it is more costly to gain access to marginal users than inframarginal ones, consistent with Google’s traffic-acquisition costs increasing over the years as it has grown.113F[113] While Google does not have to pay itself for the use of its own display inventory, there is clearly an opportunity cost to displaying its own inventory rather that of another firm. The claim that the company faces no traffic-acquisition costs for these properties is inaccurate.

The Roadmap’s focus on traffic-acquisition costs also overlooks content-acquisition costs—the payments to content providers from whom Google licenses video and other content for distribution on ad-driven and subscription services such as YouTube and Google Play. While Google does not pay a publisher for access to “eyeballs” on its owned-and-operated properties, it pays substantial and increasing amounts for content on those properties that attract the “eyeballs.”114F[114] Alphabet CFO Ruth Porat indicates, for example, that YouTube pays content creators “a majority of our revenue.”115F[115] Leaving this expense out of the calculation is another example of the over-simplification that characterizes many of the claims that Google’s ad tech business is a simple (and simply anticompetitive) business.

B.   Excess pricing

Where Google’s critics diverge most significantly from the spirit of U.S. antitrust law is in their overriding concern for how advertising revenues are distributed among the recipients of advertisers’ payments: intermediaries (Google) and publishers. The Texas Complaint alleges that Google has a higher “take rate” than competing exchanges,116F[116] is able to increase its take rates without losing market share,117F[117] and “manipulates auctions to increase its take rate.”118F[118] This follows the Roadmap’s speculation—based on the CMA Interim Report—that Google may take a larger cut of advertising spending than its competitors.119F[119] And these allegations echo claims made in another report that Google introduces “hidden” fees that increase the overall cut it takes from ad auctions.120F[120]

First, it should be noted that the basis for these claims in the Roadmap are drawn from the CMA investigation’s interim report, published in December 2019. In the final report, after further investigation, the CMA abandoned this claim. The final report describes the CMA’s analysis of all the billions of Google Ad Manager open auctions related to U.K. web traffic during the period between March 8 and March 14, 2020. This, according to the CMA, allowed it to observe any possible “hidden” fees, as well. The CMA concludes:

Our analysis found that, in transactions where both Google Ads and Ad Manager (AdX) are used, Google’s overall take rate is approximately 30% of advertisers’ spend. This is broadly in line with (or slightly lower than) our aggregate market-wide fee estimate outlined above. We also calculated the margin between the winning bid and the second highest bid in AdX for Google and non-Google DSPs…. We found that Google’s average winning margin was similar to that of non-Google DSPs. Overall, this evidence does not indicate that Google is currently extracting significant hidden fees. As noted below, however, it retains the ability and incentive to do so.121F[121]

This is a crucial finding that severely undermines the allegations that Google extracts excessive or “hidden” fees. It also undermines the claim that there are “missing funds accruing to Google.” While these conclusions do not eliminate the possibility that the industrywide price could itself be above competitive levels (and it remains to be seen whether the plaintiffs states in the Texas case will produce different evidence), they do mean that the best evidence currently available calls into question the charge that Google exploits a lack of interoperability by prioritizing its own products or that it engages in opaque pricing to conceal hidden charges of which its customers are unaware.

More fundamentally, absent evidence of Google deceiving advertisers and publishers to extract above-competitive margins, claims that its prices are “too high” or its revenue sharing “too low” are at odds with established antitrust law. U.S. antitrust law does not attempt to derive “proper” prices and impose these obligations on companies to ensure a “fair” outcome. Absent anticompetitive defects in the process, even monopolists are free to charge monopoly prices. The alternative would be for some agency—a court or a regulator—to regulate pricing and second guess every business decision made by dominant firms.

C.   Cross-subsidies

At the same time, the Roadmap alleges that Google can “charge low prices at one end of the stack, to drive out competitors, while charging high prices at the other to counterbalance any losses.”122F[122] But even if true, this would not be anticompetitive. It is a widely understood feature of platforms that they can shift prices from one side of a multisided market to another to maximize the platform’s total value. For example, a marketplace may make sellers bear the burden of fraud or mis-selling to give assurance to customers, and grow the consumer side of the platform market, just as a ridesharing app may discount rides to attract customers to build a large enough base to induce drivers onto the app.

This is a normal part of platform economics, which has long recognized that offering one side a low, zero, or negative price can be efficient and procompetitive.123F[123] As the U.S. Supreme Court held in Ohio v. American Express, an integrated competitive-effects analysis should look at the overall effect on output, not the effect on one side of the market; the relevant market must include both sides of the platform or the market would not exist at all.124F[124] There is no reason to think that this kind of behavior would generally be classed as “predatory pricing” in the absence of other behavior, such as raising prices after driving out competitors.125F[125]

But neither the Texas Complaint nor the Roadmap allege that Google’s prices were predatory. On the contrary, their sole claim in this respect is that, after being acquired by Google, DoubleClick lowered its prices (by a factor of ten, according to the Roadmap126F[126]), which it then maintained at these lower levels. This price reduction is facially procompetitive, however. It is unusual, to say the least, to describe a price reduction, with no subsequent price rises, as anticompetitive. If less-efficient competitors were unable to compete with these lower prices, that is competition in action. The law does not preclude nonpredatory low prices, nor even predatory prices without recoupment.127F[127] Sustained price reductions are one of the primary goals of antitrust.

Moreover, the source of the Roadmap’s claim that these price reductions were done “to drive out competitors” was, notably, a company that was not actually driven out of business by these price reductions. The source was an ad server, Smart, which claimed that Google’s price reductions “made the provision of publisher ad server difficult to sustain as a standalone business. This was the main reason why Smart felt the need to expand into the provision of SSP services.”128F[128] A competitor of Google’s responding to price reductions by broadening its own offerings is, again, procompetitive, not anticompetitive.

D.   Data gathering and integration

The Texas Complaint and the Roadmap describe several pro-privacy measures Google has adopted or plans to adopt as being detrimental to its competitors, including the decision to disable third-party cookies (which allow digital advertising companies to track users across the web to serve them relevant targeted ads) on the Chrome browser.129F[129] The Complaint argues that this shift benefits Google to the detriment of other ad tech companies, because (it says) Google, but not its competitors, has other data sources it can use to target ads at users.130F[130] In the same vein, the Roadmap points to Google’s decision not to share with advertisers raw data that it compiles about users.131F[131]

The Complaint ignores regulatory causes of these changes altogether, and the Roadmap dismisses the suggestion that they may be driven by the European General Data Protection Regulation (GDPR), on the basis that “data sharing here in the U.S., where we have no privacy regulatory scheme akin to that which is in place in Europe” has also been curtailed.132F[132] Both forget data-privacy laws in U.S. states, such as the California Consumer Privacy Act (CCPA). And, even if that weren’t the case, the claim that GDPR would have no effect outside Europe ignores that companies may find it easier to comply with such laws by changing their practices globally, rather than on a country-by-country or state-by-state basis.133F[133] Many companies have done this: Microsoft, for example, announced in November 2019 that it would “honor California’s new privacy rights throughout the United States.”134F[134]

Both also ignore the possibility that these provisions may be a response to demand from users of Google Chrome. Google may have good reasons to maintain a reputation for protecting user privacy, particularly because of the wide range of services it provides where user privacy is often of paramount importance to many users: Search, Maps, Gmail, YouTube, and Chrome itself.

Apple and Mozilla, neither of which has a significant online display advertising arm (and thus, have no incentive to block cookies simply to disadvantage display advertisers, as the Complaint alleges Google has done) have taken similar steps to increase user privacy. These are direct competitors of Google Chrome’s, and when Apple made blocking third-party cookies the default in its Safari browser, it was reported by one major outlet as “beating Google by two years to the privacy feature.”135F[135] Indeed, one of the reasons that Google delayed its disabling of third-party cookies was reportedly to implement technologies to “make it easier for advertisers to target certain demographics without laser-sighting down to specific people, ensure that the infrastructure many sites use for logins don’t break, and help provide some level of anonymous tracking so advertisers can know if their ads actually converted into sales.”136F[136]

That Chrome’s competitors, neither of which has an incentive to hurt ad tech companies, have taken the same steps that the Complaint alleges Google is taking for anticompetitive reasons should be compelling evidence that Google, too, is responding to user demand and/or regulation. Under U.S. law, the fact that these are legitimate moves and benefit users interested in privacy—and, indeed, may be a response to competition in the browser market—undermines claims that Google has failed to maximize competition along other dimensions.

The Roadmap also presents a hypothetical circumstance that amounts to an allegation that Google “captures” data from ads served to publishers to “expropriate” publishers’ investments in content that attracts a particular audience:

Some publishers have invested in content that attracts and retains a specific type of consumer, for example, readers of the Wall Street Journal or Golf Digest; this in turn allows them to support their business by selling valuable ads to advertisers looking for exposure to those audiences. Google has two ways to expropriate that value. First, rather than serve an ad on the Wall Street Journal at a high price, it can track the user who visited the Wall Street Journal and wait until she visits a site that sells space at low prices, for example, a local recipe blogger. Google can then sell Wall Street Journal users to advertisers in a way that does not benefit the Wall Street Journal at all and costs advertisers much less. A second strategy used by Google is to take the data describing these differentiated audiences and use it to create an imitation portfolio of consumers that mimic the characteristics of the publisher’s audience. For example, Google could create an audience of consumers similar to the people who read Golf Digest. Then Google sells access to this group of consumers when they visit inexpensive websites. Advertisers are happy to buy these ads because the consumers likely belong to the specialized audience of interest but are available at a much lower price. In these ways the unique audience assembled by the publisher is copied and expropriated.137F[137]

It should be noted that the Roadmap does not conclude that Google engages in these practices, but merely describes strategies Google “can” undertake to “expropriate” publishers’ investments. The Roadmap concedes that advertisers would be “happy” under such hypotheticals, because they are buying effective ads at a “much lower price.” In the Roadmap’s example, the hypothetical recipe blogger is “happy” that it earned revenues from selling an impression and the advertiser is “happy” that it paid a lower price than it would have had the impression been sold to the Wall Street Journal. The Wall Street Journal may not be so “happy” that it did not serve that particular ad, but that display space did not sit empty; the hypothetical lost ad was replaced by another impression that was served. And it is neither Google’s nor antitrust law’s job to make specific publishers better off—nor to make publishers better off at the expense of advertisers—but to ensure that the market as a whole is competitive and acting in consumers’ interests.

These hypotheticals again highlight the tensions discussed above between the different sides of a multisided market. Actions that make advertisers “happy” may come at the expense of publishers’ advertising revenues and actions that increase publishers’ revenues may increase costs to advertisers. One of the goals of a multisided market intermediary such as Google is to balance these competing interests to maximize total revenues flowing through the ad tech stack.

The Roadmap concludes that, through its “entire family of products,” Google collects and analyzes substantial amounts of information about its users. It uses this information to maximize the “effectiveness,” “precision,” and “value” of the ads it intermediates.

First, Google offers an entire family of products—everything from Gmail and Google Maps to the Google Calendar, Google Chrome, Android mobile operating system and the search engine—that gather valuable personal data about its users. Second, the products across the ad stack further collect data on consumer activities that the company then integrates to maximize the effectiveness and precision of ad targeting and attribution and thereby the value of the ads.138F[138]

Rather than “expropriating” publishers’ data, it would be reasonable to conclude that Google is adding value to the data provided by publishers, advertisers, and consumers to better target ads. For example, the Wall Street Journal may not know that a consumer recently did a Google Search for “running shoes.” By adding valuable information from Search, the consumer might be served a relevant running shoe ad on the Wall Street Journal’s site. This benefits the publisher who is paid for serving a valuable impression, the advertiser who sells a pair of shoes, and the consumer who obtains useful information and purchases the product she was seeking.

E.   Accelerated Mobile Pages (AMP) and header bidding

The Texas Complaint, like the Roadmap, alleges that Google designed Accelerated Mobile Pages (AMP) “[t]o respond to the threat of header bidding… [by making it] essentially incompatible with JavaScript and header bidding. Google then used its power in the search market to effectively force publishers into using AMP.”139F[139] But this gets several key facts wrong. First, AMP is an open-source industry collaboration project and Google cannot unilaterally impose a design standard on it. 140F[140] Second, a version of header bidding can work with AMP.141F[141] Third, it is mistaken to assert that “non-AMP-formatted results often do not even show up on the first page of results, regardless of their relevance.”142F[142] AMP has been a prerequisite only for inclusion in the top news story carousel, while other listings are ranked by relevance and speed.

Importantly, the argument ignores the main benefit of AMP to publishers and users: faster load times for mobile users who may be on slow connections. One of header bidding’s chief downsides is that it increases page-load latency. It is obvious why an HTML framework built to maximize load times would not be compatible with header bidding. Because AMP confers undisputed benefits on users and publishers, Google and the other companies involved in the AMP project have no obligation to re-engineer AMP to be compatible with header bidding. Any conclusion otherwise would involve a court deciding that users should be forced to use a slower Internet so that websites can use header bidding.

F.   Alleged barriers to entry in the open digital display ad market

Claims about Google’s alleged market power in display advertising rest on assumptions that the company enjoys the benefits of significant barriers to entry throughout the ad tech stack, thus enabling it to extract supercompetitive rents without fear of competition: “With these barriers in place,” it is claimed, “entry seems nearly futile.”143F[143]

A key element in establishing a company’s durable market power—and thus, its ability to impose anticompetitive costs on its users—is the presence of entry barriers. Even a market with only a single company—a true monopoly—cannot act like a monopoly if entry into its market is easy; if it did profitably raise prices, new competitors would enter the market and undercut it.144F[144]

As the Roadmap concedes, “[m]arket power is not permanent, of course. It can be undercut by, among other things, new entrants that offer better quality or lower prices.”145F[145] This notion of “contestability” is a fundamental part of assessing the competitiveness of markets under U.S. antitrust law.146F[146] In the absence of barriers to entry, it is well-established that assumptions of future competitive harm from ongoing conduct cannot be sustained.147F[147] Thus, the Roadmap bases much of its brief against Google on the presence of barriers to entry, “which heighten[] the prospect that Google can engage in conduct that harms competition without restraint from new entrants or potential new entrants.”148F[148] On the strength of these asserted barriers, the Roadmap’s authors interpret ambiguous conduct as anticompetitive.

According to the Roadmap, “the CMA’s findings reveal a number of significant barriers to entry into the digital advertising market.”149F[149] But most of its assertions in this regard are flawed, either because the CMA did not, in fact, make “findings” in the ways it suggests, or else because it reaches incorrect conclusions that certain conduct constitutes a barrier.150F[150] The Texas Complaint’s assertions of similar barriers to entry are likewise problematic.151F[151]

1.     Consumer location information

Although the Texas Complaint does not discuss it as an explicit barrier to entry,152F[152] one of the Omidyar Roadmap’s key assertions about barriers to entry relates to Google’s access to user-location data. It asserts that:

The CMA concluded that Google has nearly insurmountable advantages in access to location data, due to the location information it receives from the Android operating system, Google search, and other applications…. An entrant into the ad tech stack requires information about the consumer to target an ad effectively. Because Google accounts for nearly the entirety of the mobile search sector in the UK—97%—and controls many of the known sources of location data, such an entrant faces a large barrier to entry.153F[153]

But the CMA does not, in fact, “conclude[] that Google has nearly insurmountable advantages in access to location data,” either in the CMA Interim Report to which the Roadmap refers, nor in the CMA Final Report. The CMA never makes any claim of “insurmountable advantage.” Indeed, it does not use the word “insurmountable” at all, except to note that “rival platforms did not suggest that accessing consumer data was an insurmountable barrier to entry.”154F[154] Rather, to support this claim, the Roadmap cites to a portion of the CMA Interim Report recounting a suggestion made by Microsoft regarding the “critical” value of location data in providing relevant advertising.155F[155]

Moreover, that portion of the CMA Interim Report, as well as the suggestion made by Microsoft, is about search advertising, not display advertising. While the CMA does not characterize this data in the way the Roadmap claims, it does allege that “Google has exclusive access to a large amount of user data that can be used for targeted advertising and for measuring advertising outcomes, collected through its consumer-facing services. Data collected on its search platform is particularly valuable for targeting purposes in open display as it reveals users’ purchasing intent.”156F[156]

While location data may also be valuable for display advertising in a comparable way, it is not clear that the GPS-level data that is so valuable in providing mobile-search ad listings is particularly useful for display advertising, which may be just as well-targeted by less granular city- or county-level location data.

Consider the Roadmap’s illustrative example:

A digital ad for a brick-and-mortar running store in Des Moines is of little use to a runner looking to test out new shoes in Omaha, and, if shown to the Omaha runner, is unlikely to prompt a click, much less a purchase, from the Des Moines store.157F[157]

This is certainly correct. But GPS or even cell-tower location data is not necessary to determine in which city a user is located. Publicly available databases of IP address locations can provide this information, and they are readily and often freely available to all competitors. It is difficult to imagine that display advertising uses location data at any greater level of granularity except in unusual circumstances; it simply would not be particularly useful or effective.158F[158]

Furthermore, to the extent that location data (like other consumer data) may be useful for display advertising, the most significant issue affecting its availability to advertisers is not Google’s presence in the ad tech stack; it is privacy regulations that limit the collection, use, and sharing of such data.159F[159] These privacy regulations, such as the GDPR, limit the ability of digital firms to sell or otherwise pass user data to third-party advertisers. What seems like unequal treatment, in this regard, is really a case of privacy regulation in action.

These laws may have the indirect effect of favoring larger digital conglomerates that can collect user data through one service and use it to target ads in another. In this sense, it could be true that Google has informational advantages over rivals, though in a critically different way than that alleged by the Roadmap. But it can hardly be considered anticompetitive if the source of such advantage is legal constraints on information sharing. Indeed, an empirical study by economists Avi Goldfarb and Catherine Tucker found that (pre-GDPR) privacy regulation in the EU “restricted advertisers’ ability to collect data on Web users in order to target ad campaigns. We ?nd that, on average, display advertising became far less effective at changing stated purchase intent after the EU laws were enacted, relative to display advertising in other countries.”160F[160] Along similar lines, Nils Wernerfelt and his co-authors show that access to data from different sources significantly improves ad targeting.161F[161] In turn, this may give a competitive advantage to firms that operate several successful web services and applications.

As Israeli competition law scholars Michal Gal and Oshrit Aviv found with respect to the GDPR, privacy regulations can function as a barrier to entry in several further ways.162F[162] These include creating new economies of scale associated with regulatory compliance, increased litigation risk, and uncertainty around interpretation of the rules. Because they serve to make reputation more central, they also can lead users to become more likely to entrust their data to incumbents but not to unknown, new entrants.163F[163]

2.     Attribution measurement

“Attribution” refers to the method by which advertisers can see which ad led a user to an action, such as visiting a website or making a purchase.164F[164] The Roadmap alleges that Google can design attribution to mislead advertisers by, for example, favoring search ads over display ads.165F[165] This would lead to more of the advertiser’s money going to Google instead of (in part) to a publisher, and (assuming, as the Roadmap implies, that this makes ad campaigns less effective166F[166]) can harm advertisers by misleading them into choosing a less-effective advertising channel.

The Roadmap provides no evidence this is taking place. What it describes is more a complaint about the nature of search advertising in general: that companies will sometimes end up paying for ads in lieu of identical organic search results for their pages. That is an argument to be had elsewhere, but there are clear reasons why it may be in an advertiser’s interest to advertise even in these situations. Paid search ads may give them greater control over how a link is displayed to a user (for example, with text the advertiser has chosen, rather than text that the search engine has retrieved) or guarantee a prominent listing for searches where the advertiser’s URL listing is not always guaranteed to be on top.

Apart from the broader objection to the nature of search advertising, the Roadmap’s authors also object to Google setting an attribution default in its DSP. But advertisers can choose from several different attribution models, not just the default one that the Roadmap objects to, which attributes to search ads the “last click.” Other options include “last non-direct click,” which “ignores direct traffic and attributes 100% of the conversion value to the last channel that the customer clicked through from before buying or converting”; “last Google Ads click”; “First interaction”; and others that give attribution weights according to where and when the customer saw or used the ads during their purchase or conversion “journey.”167F[167] These are precisely the kinds of models that the Roadmap’s authors implicitly believe are more appropriate for campaigns heavy on display advertising. Advertisers can also create their own custom models, and Google has published guides for advertisers to help them choose among the different models.168F[168]

The Roadmap’s objection is thus reduced to being about the choice by Google to use the “last click” attribution model as the default. But some model has to be the default, and “last click” is also the default on, for example, Microsoft Advertising.169F[169] Indeed, according to digital-ad-intermediary company, Outbrain (one of Google’s competitors), it is the most common attribution default across the industry.170F[170] For the Roadmap’s objection to carry any weight, a case based on this claim would need to demonstrate that it was unusual for Google to use “last click” as the default attribution. Even then, given the ease with which advertisers can change the attribution model, the charge would be thin.

3.     ‘Opaque’ pricing

Both the Texas Complaint and the Roadmap allege that Google’s “opaque pricing” constitutes a barrier to entry by impeding “advertisers from switching to a lower-cost for higher-quality” buying tool.171F[171] As the Roadmap puts it, a “new or potential new PAS or DSP cannot credibly claim to be able to undercut the Google products on price if the publishers and advertisers cannot tell how much Google actually is charging.”172F[172] The Texas Complaint further alleges that “Google compounds its exclusionary auction manipulations by purposefully keeping its auction mechanics, terms, and pricing, opaque and ‘nontransparent.’ This makes it nearly impossible for publishers and advertisers to discover Google’s misrepresentations, and even harder for rivals to neutralize or offset.”173F[173] Both the Texas Complaint and the Roadmap also suggest that competition is undermined when publishers and advertisers do not know the fee structure of the intermediaries they are using.

But it is not unusual for businesses’ costs and prices to be private to their competitors, and it is not a barrier to competition. Grocery stores do not need to know how much it cost a farmer to grow an orange or how much their rivals are paying for transportation, unless they are attempting to anticompetitively coordinate their prices; they just need to work to make their own costs as low as possible and to reduce their prices to consumers by as much as possible. Similarly efficient firms are perfectly able to offer competitive prices simply by making the best offer based on their own fundamentals; only less-efficient firms will struggle (as they should).

Along these lines, for competition to work effectively in display advertising, Google’s competitors do not need to know what Google is charging; they need to offer a price and product that is more attractive overall to potential customers than Google’s is. Similarly, a publisher does not need to know how much an advertiser bid to place an ad, nor does the advertiser need to know how much the publisher received to serve the ad. The advertiser’s competition concern is whether an effective ad can be served at a lower price from a different intermediary and the publisher’s competition concern is whether it can earn greater revenues from a different intermediary. One should not be surprised that Google does not reveal information on which competing intermediaries can free ride. Indeed, this is widely considered to be one of the hallmarks of vigorous competition.

Conclusion

As we have argued, many of the most significant claims made against Google’s ad tech products are based on a misunderstanding of U.S. antitrust law, or of the details of the ad tech market itself. Although we cannot be sure how the Texas, et al. v. Google case will develop once the allegations in the Complaint are fleshed out into full arguments, many of its initial claims and assumptions are wrongheaded. Based on the information currently available, if the court rules in favor of these, the result will be to condemn procompetitive conduct and potentially to impose costly, inefficient remedies that function as a drag on innovation.

Legislators, too, who may be concerned about Google’s conduct and tempted to impose regulatory requirements on it and other tech companies should bear in minds the risk of the Nirvana fallacy, in which real-life conduct is compared against a hypothetical “competition maximizing” benchmark, and anything that falls short is deemed problematic and in need of intervention.174F[174] That approach would pervert the incentives of businesses to innovate and compete, and would make an unobtainable “perfect” that exists only in the minds of some economists and lawyers the enemy of a “good” that exists in the market right now.

 

[1] Third Amended Complaint, Texas v. Google, 21-md-3010-PKC (S.D.N.Y. Jan 14, 2022) at 105 (hereinafter, “Texas Complaint”).

[2] See Complaint, United States v. Google LLC, No. 1:20-CV-03010 (D.D.C. Oct. 20, 2020); see also, Complaint, State of Colorado, et al. v. Google LLC, 1:20-CV-03715 (D.D.C. Dec. 17, 2020).

[3] DoJ Expected to File Antitrust Lawsuit Against Google in Weeks—Bloomberg News, U.S. News (Jul. 14, 2022), https://money.usnews.com/investing/news/articles/2022-07-14/doj-expected-to-file-antitrust-lawsuit-against-google-in-weeks-bloomberg-news.

[4] Antitrust: Commission Opens Investigation into Possible Anticompetitive Conduct by Google in the Online Advertising Technology Sector, European Commission (Jun. 22, 2021), https://ec.europa.eu/commission/presscorner/detail/en/ip_21_3143.

[5] Bundeskartellamt Publishes Report on Non-Search Online Advertising for Public Discussion, Bundeskartellamt (Aug. 29, 2022), https://www.bundeskartellamt.de/SharedDocs/Meldung/EN/Pressemitteilungen/2022/29_08_2022_SU_Online_Werbung.html?nn=3599398.

[6] Id.

[7] Online Platforms and Digital Advertising Market Study, U.K. Competition and Markets Authority (Jul. 1, 2020), https://www.gov.uk/cma-cases/online-platforms-and-digital-advertising-market-study (hereinafter “CMA Market Study”); Online Platforms and Digital Advertising, Market Study Final Report, U.K. Competition and Markets Authority (Jul. 1, 2020), https://assets.publishing.service.gov.uk/media/5fa557668fa8f5788db46efc/Final_report_Digital_ALT_TEXT.pdf (hereinafter “CMA Final Report”), at 21 & 37.

[8] See Josh Frydenberg, Competition and Consumer (Price Inquiry—Digital Advertising Services) Direction 2020 (Feb. 10, 2020); Ad Tech Inquiry Issues Paper 5, Australian Competition & Consumer Commission (Mar. 10, 2020); Digital Advertising Services Inquiry, Australian Competition & Consumer Commission (Feb. 26, 2021), https://www.accc.gov.au/focus-areas/inquiries-finalised/digital-advertising-services-inquiry/submissions-to-interim-report.

[9] Investigation of Competition in Digital Markets, Majority Staff Report and Recommendations, Subcommittee on Antitrust, Commercial and Administrative Law of the Committee on the Judiciary (Oct. 4, 2020), available at https://templatelab.com/competition-in-digital-markets/4493-519; Hearing on Stacking the Tech: Has Google Harmed Competition in Online Advertising?, Committee of the Judiciary, Subcommittee on Antitrust, Competition Policy and Consumer Rights (Sep. 15, 2020), https://www.judiciary.senate.gov/meetings/stacking-the-tech-has-google-harmed-competition-in-online-advertising.

[10] Investigation of Competition in Digital Markets, Majority Staff Report and Recommendations, id., at 20.

[11] Competition and Transparency in Digital Advertising Act, S.4258, 117th Congress (2021-2022), https://www.congress.gov/bill/117th-congress/senate-bill/4258/text.

[12] Lee Introduces Digital Advertising Act, Mike Lee US Senator for Utah (May 19, 2022), https://www.lee.senate.gov/2022/5/lee-introduces-digital-advertising-act.

[13] Support Mounts for Lee’s Digital Advertising Act, Mike Lee US Senator for Utah (May 27, 2022), https://www.lee.senate.gov/2022/5/support-mounts-for-lee-s-digital-advertising-act.

[14] Online Platforms and Digital Advertising Market Study Interim Report, U.K. Competition and Markets Authority (Dec. 18, 2019), https://assets.publishing.service.gov.uk/media/5ed0f75bd3bf7f4602e98330/Interim_report_—_web.pdf (hereinafter, “CMA Interim Report”).

[15] Fiona M. Scott Morton & David C. Dinielli, Roadmap for a Digital Advertising Monopolization Case Against Google, Omidyar Network (May 2020), https://omidyar.com/wp-content/uploads/2020/09/Roadmap-for-a-Case-Against-Google.pdf (hereinafter, “Roadmap” or “Omidyar Roadmap”). One of the Roadmap’s authors testified at a Senate hearing on the display-advertising market, and the report has been widely cited. See, e.g., Gilad Edelman, Here’s What an Antitrust Case Against Google Might Look Like, Wired (May 18, 2020), https://www.wired.com/story/antitrust-case-against-google-roadmap-paper.

[16] Damien Geradin & Dimitrios Katsifis, An EU Competition Law Analysis of Online Display Advertising in the Programmatic Age, 15 Eur. Comp. J. 55 (2019); Damien Geradin & Dimitrios Katsifis, “Trust Me, I’m Fair”: Analysing Google’s Latest Practices in Ad Tech from the Perspective of EU Competition Law, 16 Eur. Comp. J. 11 (2020); Damien Geradin & Dimitrios Katsifis, Online Platforms and Digital Advertising Market Study: Observations on CMA’s Interim Report, TILEC Discussion Paper No. DP2020-044 (Feb. 13, 2020), https://ssrn.com/abstract=3537864; Damien Geradin & Dimitrios Katsifis, Competition in Ad Tech: A Response to Google, TILEC Discussion Paper No. DP2020-038 (Jun. 3, 2020), https://ssrn.com/abstract=3617839.

[17] The markets alleged in the Texas Complaint involve (1) publisher ad servers, (2) ad exchanges, (3) ad-buying tools for large advertisers, (4) ad-buying tools for small advertisers, (5) in-app mediation tools, and (6) in-app networks. The complaint does not relate to other forms of advertising on the Internet, such as targeted text-based ads sold by search engines, video ads that run before or during video content, or shareable ads on social media platforms.

[18] This section is distilled from our much longer discussion of the broader market surrounding digital advertising. See Eric Fruits, Geoffrey A. Manne & Lazar Radic, Relevant Market in the Google AdTech Case, ICLE Issue Brief 2022-06-01 (2022), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4125569.

[19] Benedict Evans, News by the Ton: 75 Years of US Advertising (Jun. 15, 2020), https://www.ben-evans.com/benedictevans/2020/6/14/75-years-of-us-advertising; Benedict Evans, TV, Merchant Media and the Unbundling of Advertising (Mar. 18, 2022), https://www.ben-evans.com/benedictevans/2022/3/18/unbundling-advertising.

[20] See Fruits, Manne & Radic, supra note 18.

[21] Michael Schneider & Kate Aurthur, R.I.P. Cable TV: Why Hollywood Is Slowly Killing Its Biggest Moneymaker, Variety (Jul. 21, 2020), https://variety.com/2020/tv/news/cable-tv-decline-streaming-cord-cutting-1234710007 (“[B]asic cable feasted on a dual revenue stream of subscriber fees and advertising revenue. But that gravy train started going off the rails when the streaming services arrived.”).

[22] At the same time, as Benedict Evans notes, not all digital advertising is drawn from offline sources: “[I]f you talk to people at both Google and Facebook and in the agency world, you’ll hear that a lot of the money spent on Google and Facebook is money that was never spent on traditional advertising—it’s coming from SMEs [small and medium enterprises] and local businesses that might have spent in classified at most but probably wouldn’t have done even that.” Evans, News by the Ton, supra note 19 (emphasis in original).

[23] See Xi He, Rigoberto Lopez & Yizao Liu, Are Online and Offline Advertising Substitutes or Complements? Evidence from U.S. Food Industries, 15 J. Agricultural & Food Indus. Org. 1 (2017).

[24] David Bardey, Jorge Tovar & Nicolas Santos, Characterization of the Relevant Market in the Media Industry: Some New Evidence, Toulouse School of Economics Working Paper 16-719 (2016), https://www.tse-fr.eu/publications/characterization-relevant-market-media-industry-some-new-evidence (“The results show substitution and complementary patterns across certain media outlets. An increase in price for advertising in radio, for instance, leads to higher demand for newspapers and outdoors. Similarly, complementarity relationships between media outlets are observed, suggesting that advertising across the various media platforms is, overall, interwoven.”).

[25] David S. Evans, The Online Advertising Industry: Economics, Evolution, and Privacy, 23 J. Econ. Persp. 37, 49 (2009).

[26] Avi Goldfarb & Catherine Tucker, Search Engine Advertising: Channel Substitution When Pricing Ads to Context, 57 Management Sci. 458 (2011) (The authors find that the price of “ambulance chaser” lawyer ads was significantly more expensive in states prohibiting direct-mail solicitation by attorneys. This leads them to conclude that “online advertising substitutes for online advertising”).

[27] Avi Goldfarb & Catherine Tucker, Substitution Between Offline and Online Advertising Markets, 7 J. Competition L. & Econ. 37, 43 (2011).

[28] Daniel S. Levine, Ad-Supported Cyber-Magazines to Launch on Internet, Adweek (Sep. 10, 1993).

[29] Id.

[30] Brian Morrissey, How the Banner Ad Was Born, Digiday (Apr. 12, 2013), https://digiday.com/marketing/how-the-banner-ad-was-born.

[31] Chris Lapham, AOL and GNN Partner to Build Launch Pad, CMC Magazine (Jul. 1, 1995), https://www.december.com/cmc/mag/1995/jul/cutting.html.

[32] Kim Cleland, Poppe Creates Web Net, Advertising Age (Oct. 30, 1995).

[33] Id.

[34] The History of Online Advertising, OKO Ad Management (Jul. 19, 2019), https://oko.uk/blog/the-history-of-online-advertising.

[35] The company changed its name to Overture, which was acquired by Yahoo! in 2003.

[36] GoTo.com Announces First Round of Financing, Totaling More Than $ 6 Million, Led by Draper Fisher Jurvetson, Business Wire (May 19, 1998), https://www.internetnews.com/marketing/goto-com-raises-6-million-in-first-round-of-financing.

[37] Will Oremus, Google’s Big Break, Slate (Oct. 13, 2013), https://slate.com/business/2013/10/googles-big-break-how-bill-gross-goto-com-inspired-the-adwords-business-model.html.

[38] Id.

[39] Dean Schmid, The History of Display Advertising: Everything You Need to Know, DisruptorDaily.com (Aug. 14, 2017), https://www.disruptordaily.com/the-history-of-display-advertising-everything-you-need-to-know.

[40] Julia Scheeres, Death of Banner Ads Exaggerated, Wired (Jan. 26, 2001), https://www.wired.com/2001/01/death-of-banner-ads-exaggerated.

[41] Id.

[42] Breaking News, AdAge (Oct. 23, 2000).

[43] Mark Evans, Investors Leap off Overture Roller Coaster: Rival Google Elbows In, National Post (Feb. 21, 2002).

[44] Oremus, supra note 37.

[45] Google Grabs Applied Semantics, EuropeMedia (Apr. 25, 2003); Google Expands Advertising Monetization Program for Websites, Google Press Release (Jun. 18, 2003), http://googlepress.blogspot.com/2003/06/google-expands-advertising-monetization.html.

[46] Dean Schmid, The History of Display Advertising: Everything You Need to Know, DisruptorDaily (Aug. 14, 2017), https://www.disruptordaily.com/the-history-of-display-advertising-everything-you-need-to-know.

[47] Kate Walsh, Search Marketing: Understanding the Basics, B2B Marketing Magazine (March 2006).

[48] Louise Story & Miguel Helft, Google Buys DoubleClick for $3.1 Billion, The New York Times (Apr. 14, 2007), https://www.nytimes.com/2007/04/14/technology/14DoubleClick.html.

[49] Sarah Sluis, The Year Header Bidding Went Mainstream, AdExchanger (Dec. 27, 2016); Townhall Media Selects OpenX for Patent-Pending Header Bidding Solution, BusinessWire (Sep. 18, 2015), https://www.businesswire.com/news/home/20150918005110/en/Townhall-Media-Selects-OpenX-for-Patent-Pending-Header-Bidding-Solution.

[50] As the name suggests, ad waterfalls enable publishers to sell their inventory seriatim, beginning with premium, direct sales and flowing through the most historically profitable ad servers in succession to unload unsold inventory before offering its remnant inventory in the open display channel. See, e.g., Maciej Zawadzinski, What Is Waterfalling and How Does it Work?, Clearcode (Aug. 20, 2021), https://clearcode.cc/blog/what-is-waterfalling.

[51] See, e.g., Header Bidding, OKO Ad Management, https://oko.uk/topic/header-bidding (retrieved July 27, 2022).

[52] Client-side header bidding is so-named because it operates via a small piece of java script embedded in the header of a publisher’s website and executed within the user’s browser (i.e., client). See, e.g., Maciej Zawadzinski, What Is Header Bidding and How Does it Work?, Clearcode (Aug. 20, 2021), https://clearcode.cc/blog/what-is-header-bidding.

[53] Header Bidding Facts and Statistics 2021, Automatad (Jun. 27, 2021), https://headerbidding.co/header-bidding-statistics. Today, 70% of the top 10,000 U.S. publishers use header bidding. See Header Bidding (HBIX) Tracker, kevel (retrieved Nov. 1, 2022), https://www.kevel.com/hbix.

[54] See, e.g., CMA Final Report, supra note 7, at Appendix M, ¶ 33.

[55] See, e.g., Vishveshwar Jatain, Header Bidding Integrations: Client Vs. Server-Side, Explained, Blockthrough (Apr. 15, 2021), https://blockthrough.com/blog/header-bidding-integrations-client-vs-server-side-explained.

[56] IAB and PwC, IAB Internet Advertising Revenue Report, 2010 Full Year Results (Apr. 2011), available at https://www.iab.com/wp-content/uploads/2015/05/IAB_Full_year_2010_0413_Final.pdf; Megan Graham, Digital Ad Revenue Jumped 35% in the U.S. Last Year, Biggest Gain Since 2006, Wall Street Journal (Apr. 12, 2022), https://www.wsj.com/articles/digital-ad-revenue-jumped-35-in-the-u-s-last-year-biggest-gain-since-2006-11649759401.

[57] Producer Price Index by Commodity: Advertising Space and Time Sales: Internet Advertising Sales, Excluding Internet Advertising Sold by Print Publishers, U.S. Bureau of Labor Statistics, https://fred.stlouisfed.org/series/WPU365; Producer Price Index, December 2009—February 2021, U.S. Bureau of Labor Statistics, https://fred.stlouisfed.org/graph/?g=vtTd.

[58] Scott Morton & Dinielli, supra note 15, at 9.

[59] Importantly, however, network effects are not monolithic; nor do they increase forever. For different types of networks at different points in their growth, adding more users might not increase the value of the platform and could even reduce the platform’s benefits. See, e.g., D’Arcy Coolican & Li Jin, The Dynamics of Network Effects, Andreesen Horowitz (Dec. 13, 2018), https://a16z.com/2018/12/13/network-effects-dynamics-in-practice.

[60] See Stan J. Liebowitz & Stephen E. Margolis, Network Externality: An Uncommon Tragedy, 8 J. Econ. Persp. 133 (1994).

[61] Catherine Tucker, Digital Data, Platforms and the Usual [Antitrust] Suspects: Network Effects, Switching Costs, Essential Facility, 54 Rev. Indus. Org. 683, 686 (2019).

[62] See, e.g., David S. Evans, Economics of Vertical Restraints for Multi-Sided Platforms, University of Chicago Institute for Law & Economics Olin Research Paper No. 626 (Jan. 2, 2013), https://ssrn.com/abstract=2195778.

[63] See, e.g., Stylianos Despotakis, R. Ravi & Amin Sayedi, First-Price Auctions in Online Display Advertising, 58 J. Marketing Research 888 (2021). See also Display Advertising Switched to First-Price Auctions After Adoption of Header Bidding, New Study Finds, Tepper School of Business (Apr. 22, 2020), https://www.cmu.edu/tepper/news/stories/2020/april/display-advertisting-research-ravi.html.

[64]  Jonathan Levin, Auction Theory (Oct. 2004), available at https://web.stanford.edu/~jdlevin/Econ%20286/Auctions.pdf.

[65] Maciej Zawadzi?ski, How Do First-Price and Second-Price Auctions Work in Online Advertising?, Clearcode (Aug. 12, 2021), https://clearcode.cc/blog/first-price-second-price-auction.

[66] Id.

[67] Texas Complaint, supra note 1, at ¶ 351 (“Overall, the lack of transparency prevents more efficient competition that would drive greater innovation, increase the quality of intermediary services, increase output, and create downward pricing pressure on intermediary fees.”); Scott Morton & Dinielli, supra note 15, at 18 (“Based on the public facts known at the moment, however, it does not seem plausible that the incremental efficiencies created by the conduct described here could outweigh all the harms to competition resulting from this broad pattern of behaviors.”); Scott Morton & Dinielli, supra note 15, at 38 (“It also is true that Google has allowed some rivals to survive (although not necessarily to thrive). It is possible that Google adopted a strategy of incomplete foreclosure specifically so that it can paint an illusion of healthy competition when the reality is quite different. Indeed, to the extent Google has adopted ‘pro-competitive’ concessions, the narrative here demonstrates that they simply have not succeeded in addressing the harms or lowering the barriers to entry.”).

[68] Id. at 3 (“It is clear even to us as lay people that there are less anticompetitive ways of delivering effective digital advertising—and thereby preserving the substantial benefits from this technology—than those employed by Google.”).

[69] United States v. Aluminum Co. of America, 148 F.2d 416, 430 (2nd Cir. 1945) (Learned Hand, J.) (emphasis added).

[70] Cal. Computer Prods., Inc. v. Int’l Bus. Machine Corp., 613 F.2d 727, 744 (9th Cir. 1979) (“IBM, assuming it was a monopolist, had the right to redesign its products to make them more attractive to buyers whether by reason of lower manufacturing cost and price or improved performance. It was under no duty to help [its competitors] survive or expand.”).

[71] Scott Morton & Dinielli, supra note 15 at 18.

[72] Texas Second Amended Complaint at ¶ 113.

[73] See Verizon Commc’ns Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398, 408 (2004). The exception—”at or near the outer boundary of § 2 liability” (id. at 409)—is the extremely narrow case in which a monopolist (i) sacrifices profits, by (ii) terminating a prior course of dealing, (iii) for no purpose except to harm competition. See Novell v. Microsoft, 731 F.3d 1064, 1074-75 (10th Cir. 2013) (Gorsuch, J.) (holding that a refusal-to-deal claim requires terminating “a preexisting voluntary” course of dealing where the “monopolist decided to forsake short-term profits,” and “the monopolist’s conduct” is “irrational but for its anticompetitive effect”).

[74] Opinion and Order, Texas, et al. v. Google, 21-md-3010-PKC (S.D.N.Y, Sep. 13, 2022) (citing Charych v. Siriusware, Inc., 790 Fed. App’x 299, 302 (2nd Cir. 2019)).

[75] United States v. Colgate & Co., 250 U.S. 300, 307 (1919).

[76] Trinko, 540 U.S. at 407-08

[77] See Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 488 (1977) (“The antitrust laws, however, were enacted for ‘the protection of competition not competitors.’”) (quoting Brown Shoe Co. v. United States, 370 U.S. 294, 320 (1962)).

[78] See Ohio v. American Express, 138 U.S. 2274, 2285 (2018) (“Due to indirect network effects, two-sided platforms cannot raise prices on one side without risking a feedback loop of declining demand…. Price increases on one side of the platform [] do not suggest anticompetitive effects without some evidence that they have increased the overall cost of the platform’s services.”).

[79] Geoffrey A. Manne, In Defence of the Supreme Court’s ‘Single Market’ Definition in Ohio v American Express, 7 J. Antitrust Enforcement 104, 111 (2019) (quoting Brief for Amici Curiae Antitrust Law & Economics Scholars in Support of Respondents at 19, Ohio v. American Express, 138 U.S. 2274 (2018) (No. 16-1454) and United States, et al. v. American Express, 838 F.3d 179, 198 (2nd Cir. 2016)).

[80] Among innumerable examples, see Texas Complaint, supra note 1, at ¶ 297 (“Google’s harm to the competitive process has harmed customers in this market, i.e., online publishers.”). Notably, the Texas Complaint does, in places, recognize that identifying the incidence of benefits and harms in multisided markets is complex—it just fails to carry its analysis to its logical conclusion. Thus, in ¶157 the Complaint notes that “[t]he higher advertising revenue publishers make from exchanges permits publishers to offer consumers better quality content and lower-priced or free access to their content.” (Emphasis added). Undoubtedly, this is true. But if it is correct, then it must also be correct that, at the same time, the correspondingly higher prices advertisers pay for advertising through exchanges limits their ability to provide marketing benefits directly to consumers and may increase the price to consumers of the advertised goods. It is an empirical question which effect is larger, but the mere possibility that one set of consumers could benefit from a different arrangement is insufficient on its own to identify harm when another set of consumers would be harmed by it.

[81] See Harold Demsetz, Information and Efficiency: Another Viewpoint, 12 J. L. & Econ. 1, 1-2 (1969) (“In practice, those who adopt the nirvana viewpoint seek to discover discrepancies between the ideal and the real and if discrepancies are found, they deduce that the real is inefficient…. The nirvana approach is… susceptible… to committing three logical fallacies—the grass is always greener fallacy, the fallacy of the free lunch, and the people could be different fallacy.”) (emphasis in original).

[82] See, generally, Thomas Nachbar, Less Restrictive Alternatives and the Ancillary Restraints Doctrine, Virginia Law and Economics Research Paper No. 2020-18 (2021) (forthcoming U. Seattle L. Rev.) at 57-8, available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3723807 (“The more general risk to tech markets comes from the intangible nature of the products and services they produce. Although many of the cases cited for less restrictive alternatives are horizontal cases, it is the vertical context (which normally receives more permissive antitrust review) in which less restrictive alternatives present the greatest likelihood of destabilizing current law because of the difficulty of specifying what is and is not less restrictive with regard to the intangible products produced by today’s ‘big tech’ economy. To the extent that less restrictive alternatives present problems of incrementalism, those problems will be exacerbated in the ‘big tech’ markets.”).

[83] See Geoffrey A. Manne, Error Costs in Digital Markets, Global Antitrust Institute Report on the Digital Economy (Joshua D. Wright & Douglas H. Ginsburg, eds., 2020) 33, 76, available at https://gaidigitalreport.com/wp-content/uploads/2020/11/Manne-Error-Costs-in-Digital-Markets.pdf (“The concern with error costs is especially high in dynamic markets in which it is difficult to discern the real competitive effects of a firm’s conduct from observation alone. And for several reasons, antitrust decision-making in the context of innovation tends much more readily toward distrust of novel behavior, thus exacerbating the risk and cost of over-enforcement.”).

[84] Among many other examples, see Texas Second Amended Complaint at ¶138 (“Then, through Dynamic Allocation, Google’s ad server passed inside information to Google’s exchange and permitted Google’s exchange to purchase valuable impressions at artificially depressed prices. Competing exchanges were deprived of the opportunity to compete for inventory and left with the low-value impressions passed over by Google’s exchange.”); Omidyar Roadmap, supra note 15, at 20 (“[A]fter purchasing DoubleClick, which became its publisher ad server, Google apparently lowered its prices to publishers by a factor of ten, at least according to one publisher’s account related to the CMA. Low prices for this service can force rivals to depart, thereby directly reducing competition.”).

[85] Barry Wright Corp. v. ITT Grinnell Corp., 724 F.2d. 227, 235 (1st Cir. 1983) (Breyer, C.J.).

[86] Allied Orthopedic Appliances, Inc. v. Tyco Health Care Grp. LP, 592 F.3d 991, 1000 (9th Cir. 2010).

[87] Berkey Photo, Inc. v. Eastman Kodak Co., 603 F.2d 263, 287 (2nd Cir. 1979). See also, Manne & Wright, Innovation and the Limits of Antitrust, 6 J. Comp. L. & Econ. 153–202 (March 2010), https://doi.org/10.1093/joclec/nhp032.

[88] Allied Orthopedic, 592 F.3d at 999-1000; see also, California Computers Prods. v. IBM, 613 F.2d 727, 744 (9th Cir. 1979); Foremost Pro Color, Inc. v. Eastman Kodak Co., 703 F.2d 534, 543-45 (9th Cir. 1983).

[89] Foremost Pro Color, 703 F.2d at 543.

[90] Allied Orthopedic, 592 F.3d at 1000.

[91] Verizon Commc’ns, Inc. v. Law Offices of Curtis V. Trinko, 540 U.S. 398, 400-41 (2004).

[92] Trinko, 540 U.S. at 415-16.

[93] Id.; see also New York Merc. Exch., Inc. v. Intercontinental Exch. Inc., 323 F.Supp.2d 559 (S.D.N.Y. 2004) (dismissing Section 2 claim and reiterating limited exceptions in which forced sharing is appropriate).

[94] See, e.g., Jack Walters & Sons Corp. v. Morton Bldg., Inc., 737 F.2d 698, 710 (7th Cir. 1984) (“We just said that vertical integration is not an improper objective. But this puts the matter too tepidly; vertical integration usually is procompetitive. If there are cost savings from bringing into the firm a function formerly performed outside it, the firm will be made a more effective competitor.”). There is a robust body of empirical research indicating that vertical integration is generally procompetitive or benign. For a summary of the leading meta-studies by DOJ and FTC economists and others, see Koren W. Wong-Ervin, Antitrust Analysis of Vertical Mergers: Recent Developments and Economic Teachings, The Antitrust Source (February 2019), https://www.americanbar.org/content/dam/aba/publishing/antitrust_source/2018-2019/atsource-february2019/feb19_wong_ervin_2_18f.pdf. See also, Francine Lafontaine & Margaret Slade, Vertical Integration and Firm Boundaries: The Evidence, 45 J. Econ. Lit. 677 (2007) (“In spite of the lack of unified theory, overall a fairly clear empirical picture emerges. The data appear to be telling us that efficiency considerations overwhelm anticompetitive motives in most contexts. Furthermore, even when we limit attention to natural monopolies or tight oligopolies, the evidence of anticompetitive harm is not strong.”). See also, generally, Geoffrey A. Manne, Kristian Stout & Eric Fruits, The Fatal Economic Flaws of the Contemporary Campaign Against Vertical Integration, 68 Kansas L. Rev. 923 (2020).

[95] Scott Morton & Dinielli, supra note 15, at 17.

[96] See, e.g., Port Dock & Stone Corp. v. Oldcastle Ne., Inc., 507 F.3d 117, 123-25 (2nd Cir. 2007) (affirming dismissal of a Section 2 claim and finding that even a monopolist’s “vertical expansion into another level of the same product market will ordinarily be for the purpose of increasing its efficiency, which is a prototypical valid business purpose”). Moreover, single-firm conduct that supposedly projects power into another market, even through anticompetitive means, does not violate Sherman Act Section 2 unless the practices threaten monopoly power in that distinct second market. Harming competition is not enough. See Trinko, 540 U.S. at 415 n.4 (citing Spectrum Sports, Inc. v. McQuillan, 506 U.S. 447, 459 (1993)).

[97] Thus, the Omidyar Roadmap condemns Google’s supposed integration of data “to maximize the effectiveness and precision of ad targeting and attribution and thereby the value of an ad,” Scott Morton & Dinielli, supra note 15, at 20, even though the conduct makes Google’s offering to advertisers more attractive.

[98] Id. at 18-19.

[99] CMA Interim Report, supra note 14, at ¶ 5.89.

[100] Texas Second Amended Complaint at ¶ 113.

[101] Scott Morton & Dinielli, supra note 15, at 22.

[102] Texas Second Amended Complaint at ¶¶ 284-91 (“Cutting off access to YouTube foreclosed competition in the ad buying tool markets and protected Google’s market power in these markets. Many DSPs stopped growing, many others went out of business, and the market overall has been closed to entry.”).

[103] Scott Morton & Dinielli, supra note 15, at 22.

[104] See, e.g., Ryan Joe, The Big Story: Call of the Peacock, AdExchanger (Jan. 22, 2020) at 31:05-31:26, https://www.adexchanger.com/podcast/the-big-story/the-big-story-call-of-the-peacock (indicating that NBC’s Peacock streaming service will have only direct sales when it launches); Kevin Weiss, What Is the Amazon Demand Side Platform (DSP)?, Amplio (July 2019), https://www.ampliodigital.com/blog/what-is-the-amazon-demand-side-platform-dsp (“Amazon DSP is the only way to access advertising inventory exclusively available on Amazon’s collection of owned online properties and devices like: Kindle; Fire TV; IMDb; Amazon Owned & Operated properties”); Tim Cross, Xandr Launches New Demand-Side Platform ‘Xandr Invest’, VideoAdNews (Jun. 10, 2019) https://videoadnews.com/2019/06/10/xandr-launches-new-demand-side-platform-xandr-invest (“Xandr [AT&T-Time Warner’s ad tech division] has announced it will be the exclusive source of inventory from Community, its recently announced video marketplace which includes content from various WarnerMedia brands as well as Vice, Hearst Magazines, Newsy, Philo, Tubi and XUMO.”).

[105] Neal Mohan, Focusing Investments to Improve Buying on YouTube, Google (Aug. 6, 2015) https://doubleclick-advertisers.googleblog.com/2015 (“To continue improving the YouTube advertising experience for as many of our clients as possible, we’ll be focusing our future development efforts on the formats and channels used by most of our partners. To enable that, as of the end of the year, we’ll no longer support the small amount of YouTube buying happening on the DoubleClick Ad Exchange.”); see also, Kelly Liyakasa, Google to Yank YouTube Inventory out of AdX by Year’s End, AdExchanger (Aug. 6, 2015), https://www.adexchanger.com/ad-exchange-news/google-to-yank-youtube-inventory-out-of-adx-by-years-end.

[106] Liyakasa, id.

[107] See Lara O’Reilly, Ad Tech Company The Trade Desk Goes Public at $28.75 Per Share—A Huge Pop on its $18 Price Target, Business Insider (Sep. 21, 2016), https://www.businessinsider.com/the-trade-desk-ipo-2016-9; Trey Titone, The Bill That Could Break Up Google and Shake Up Ad Tech, Ad Tech Explained (May 23, 2022), https://adtechexplained.com/competition-and-transparency-in-digital-advertising-act-ctda; Trade Desk Market Cap, YCharts, https://ycharts.com/companies/TTD/market_cap.

[108] Scott Morton & Dinielli, supra note 15, at 16 n.70 (identifying AppNexus as a “vigorous competitor to Google”).

[109] Id. at 2, 28-29.

[110] Annual Report (Form 10-K) for Year Ending December 31, 2021, Alphabet Inc. (Feb. 02, 2022), https://www.sec.gov/ix?doc=/Archives/edgar/data/0001652044/000165204422000019/goog-20211231.htm.

[111] Id.

[112] Id.

[113] Id.

[114] Id.

[115] Rachel Kaser, YouTube Claims to Share Billions in Ad Money with Creators, Unlike Instagram, The Next Web (Feb. 5, 2020), https://thenextweb.com/facebook/2020/02/05/youtube-claims-share-billions-ad-money-creators-unlike-instagram.

[116] Texas Complaint, supra note 1, at ¶¶ 61, 156, 253, 288.

[117] Id. ¶ 157.

[118] Id. ¶ 21.

[119] Scott Morton & Dinielli, supra note 15, at 14 (“The CMA estimates Google’s take rate, or price, at 40%, which it deems a supra-competitive price for the services provided by the Google-controlled players in ad tech stack. A recent study by the Incorporated Society of British Advertisers (ISBA) found that publishers received 51% of the price, while the amount they could track going to intermediaries was 34%. The study could not find where the remaining 15% of the price went. As we will describe below, Google has such a dominant position across all elements of the ecosystem, it seems likely that these missing funds are accruing to Google at least in part, which would support the CMA’s findings.”).

[120] Geradin & Katsifis, “Trust Me, I’m Fair,” supra note 16 (“[L]ack of competition across the ad tech chain enables Google to exploit advertisers and publishers by charging hidden fees for ad intermediation on top of its disclosed commission…. Unfortunately, we conclude that Google’s latest switch does nothing to increase auction transparency. Worse, it seems to strengthen Google’s ability to extract hidden margins from its customers, while undermining the competitive pressure exercised by header bidding.”).

[121] CMA Final Report, supra note 7, at 275 (emphasis added).

[122] Scott Morton & Dinielli, supra note 15, at 20.

[123] See, e.g., Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 1 J. Eur. Econ. Ass’n 990 (2003); Bruno Jullien, Price Skewness and Competition in Multi-Sided Markets, IDEI Working Paper 504 (March 2008), available at https://core.ac.uk/download/pdf/6375977.pdf.

[124] See Joshua. D. Wright & John. M. Yun, Burdens and Balancing in Multisided Markets: The First Principles Approach of Ohio v. American Express, 54 Rev. Industrial Organization 717 (2019); Manne, In Defence of the Supreme Court’s ‘Single Market’ Definition in Ohio v American Express, supra note 79.

[125] As described here, true pricing is theoretically possible but difficult in practice: “To successfully engage in predatory pricing means taking enormous and rising losses that grow for the ‘predatory’ firm as customers switch to it from its competitor. And once the rival firm has exited the market, if the predatory firm raises prices above average cost (i.e., to recoup its losses), there is no guarantee that a new competitor will not enter the market selling at the previously competitive price. And the competing firm can either shut down temporarily or, in some cases, just buy up the ‘predatory’ firm’s discounted goods to resell later.” Sam Bowman, Buck’s “Third Way”: A Different Road to the Same Destination, Truth on the Market (Oct. 27, 2020), https://truthonthemarket.com/2020/10/27/bucks-third-way-a-different-road-to-the-same-destination.

[126] Scott Morton & Dinielli, supra note 15, at 20.

[127] See, e.g., Barry Wright, 724 F.2d at 234-35.

[128] CMA Interim Report, supra note 14, at Appendix H, ¶ 194.

[129] Texas Complaint, supra note 1, at ¶ 477.

[130] Id. at ¶¶ 473-476.

[131] Scott Morton & Dinielli, supra note 15, at 28.

[132] Id. at 28.

[133] As one study on the effects of GDPR (in this case, on app development) notes, “While 42.1 percent of EU-developed apps exit in the year following GDPR, the analogous ?gure averages between 37.7 and 50 percent in the other six countries, con?rming the di?culty in ?nding an untreated part of the world.” Rebecca Janßen, Reinhold Kesler, Michael E. Kummer, & Joel Waldfogel, GDPR and the Lost Generation of Innovative Apps, NBER Working Paper 30028 (May 2022) at 19-20, available at https://www.nber.org/papers/w30028.

[134] Julie Brill, Microsoft Will Honor California’s New Privacy Rights Throughout the United States, Microsoft Blog (Nov. 11, 2019), https://blogs.microsoft.com/on-the-issues/2019/11/11/microsoft-california-privacy-rights.

[135] Nick Statt, Apple Updates Safari’s Anti-Tracking with Full Third-Party Cookie Blocking, The Verge (Mar. 24, 2020), https://www.theverge.com/2020/3/24/21192830/apple-safari-intelligent-tracking-privacy-full-third-party-cookie-blocking.

[136] Dieter Bohn, Google to “Phase Out” Third-Party Cookies in Chrome, but not for Two Years, The Verge (Jan. 24, 2020), https://www.theverge.com/2020/1/14/21064698/google-third-party-cookies-chrome-two-years-privacy-safari-firefox.

[137] Scott Morton & Dinielli, supra note 15, at 30.

[138] Id. at 20.

[139] Texas Complaint, supra note 1, at ¶¶ 407-408. See also, Scott Morton & Dinielli, supra note 15, at 26.

[140] See David Besbris, Introducing the Accelerated Mobile Pages Project, for a Faster, Open Mobile Web, Google (Oct. 7, 2015), https://blog.google/products/search/introducing-accelerated-mobile-pages/.

[141] See Automated Team, Header Bidding on AMP—A Complete Guide, Automated (Jan. 10, 2020), https://headerbidding.co/header-bidding-amp.

[142] Scott Morton & Dinielli, supra note 15, at 26.

[143] Id. at 17.

[144] See William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1, 14 (1982) (“In the limit, when entry and exit are completely free, efficient incumbent monopolists and oligopolists may in fact be able to prevent entry. But they can do so only by behaving virtuously, that is, by offering to consumers the benefits which competition would otherwise bring. For every deviation from good behavior instantly makes them vulnerable to hit-and-run entry.”).

[145] Scott Morton & Dinielli, supra note 15, at 15.

[146] See, generally, William J. Baumol, John C. Panzar, & Robert D. Willig, Contestable Markets and the Theory of Industry Structure (1982).

[147] United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001) (“Because a firm cannot possess monopoly power in a market unless that market is also protected by significant barriers to entry… it follows that a firm cannot threaten to achieve monopoly power in a market unless that market is, or will be, similarly protected.”).

[148] Scott Morton & Dinielli, supra note 15, at 15.

[149] Id.

[150] See CMA Final Report, supra note 7, at 252–55 for a discussion of barriers to entry.

[151] See Texas Complaint, supra note 1, at ¶ 127: In addition to these barriers, Google’s own anticompetitive conduct imposes additional barriers to entry and expansion. As addressed below in Section VII.A, from 2010 to present, Google has tied its ad server to its ad exchange, requiring publishers to use Google’s ad server in order to receive live, competitive bids from Google’s ad exchange. This tie effectively forces almost every large publisher to use Google’s ad server. And because it is difficult-to-impossible for a publisher to use multiple ad servers simultaneously, requiring publishers to use Google’s ad server effectively prohibits them from using a competitor’s ad server. Google’s anticompetitive conduct creates an unnatural and nearly insurmountable barrier to entry.

[152] An earlier version of the Texas Complaint did make assertions regarding Google’s abuse of monopoly power through the “use [of] its data advantages to trade on inside information” (Texas Second Amended Complaint at ¶ 311), by which the state plaintiffs may mean (or have meant) to encompass location data, among other things.

[153] Scott Morton & Dinielli, supra note 15, at 15.

[154] CMA Interim Report, supra note 14, at 189 (emphasis added).

[155] Id., at ¶ 3.71 (“Microsoft suggested that accessing at-scale location data from user devices is a critical input to providing relevant, localized results. It indicated its belief that Google has unique advantages in this area, due to the location data that it receives from the Android operating system and the location data it receives when users access Google Search or other apps like Google Maps/Waze.”).

[156] CMA Final Report, supra note 7, at ¶ 5.268.

[157] Scott Morton & Dinielli, supra note 15, at 15.

[158] To be sure, location data can be helpful in assessing the efficacy of advertising by, for example, enabling an advertiser to better evaluate whether an advertisement led users to go to the advertiser’s physical location. But this function hardly seems necessary to a well-functioning market, and other sources of such information (e.g., questionnaires) are available.

[159] Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation), OJ L 119, 4.5, 2016; See, e.g., Bert Peeters, Processing of Location Data: Navigating the EU Data Protection Framework, CiTiP Blog (Feb. 4, 2021), https://www.law.kuleuven.be/citip/blog/processing-of-location-data-navigating-the-eu-data-protection-framework. (“The general understanding seems to be that, while European law does not qualify location data as a ‘special category’ of data under article 9 of the GDPR, location data should for all intents and purposes be treated with the utmost of care.”).

[160] Avi Goldfarb & Catherine Tucker, Privacy Regulation and Online Advertising, 57 Mgmt. Sci. 57, 57 (2011).

[161] Nils Wernerfelt, Anna Tuchman, Bradley Shapiro, & Robert Moakler, Estimating the Value of Offsite Data to Advertisers on Meta, University of Chicago, Becker Friedman Institute for Economics Working Paper No. 114 (August 22, 2022) at 1, available at https://ssrn.com/abstract=4198438 (“Taken together, our results suggest a substantial benefit of offsite data across a wide range of advertisers, an important input into policy in this space.”).

[162] Michal S. Gal & Oshrit Aviv, The Competitive Effects of the GDPR, 16 J. Competition & Econ. 349 (May 18, 2020). See also James Campbell, Avi Goldfarb, & Catherine Tucker, Privacy Regulation and Market Structure, 24 J. Econ. & Mgmt. Strategy 47, 68 (2015) (“[A] potential risk in privacy regulation is the entrenchment of the existing incumbent firms and a consequent reduction in the incentives to invest in quality. These incentives are stronger when firms have little consumer-facing price flexibility, as is the case in online media.”).

[163] Gal & Aviv, id. at 16.

[164] See, e.g., Cheok Lup, Explaining Marketing Attribution Models [Scenario Example], tinkerEdge (Nov. 12, 2015), https://www.tinkeredge.com/blog/web-analytics/explaining-marketing-attribution-models. (“On Day #1: User wants to purchase a coffee table for his new house, and perform a keyword search on Google. He clicks on one of the organic listings on Google Search Engine Result Page (SERP) to land onto Overstock.com. On Day # 2: He continues his search for his coffee table, and clicks on one of the PPC ads on Google SERP to land onto Overstock.com again. He subscribes to the email newsletter this time. On Day #3: He receives an eDM [electronic direct mail] from Overstock.com with a promotional offer of 30% discount sale, and clicks the “Buy Now” button from the eDM to enter the website. Unable to resist the discount offer, he decides to make a purchase of the furniture from the website.”). Attribution metrics determine which channel gets credit for the ultimate sale.

[165] Scott Morton & Dinielli, supra note 15, at 29

[166] “Furthermore, the default makes the advertiser believe that search ads are very effective relative to display ads, so the advertiser has no reason to change the default.” Id.

[167] See About the Default MCF Attribution Models: Learn How Each MCF Model Assigns Conversion Credit, Google Analytics Help (last visited Nov. 1, 2022), Attribution Models, Google, https://support.google.com/analytics/answer/1665189?hl=en. For the Google Analytics 4 version of these attribution models currently being implemented, see [GA4] About Attribution and Attribution Modeling, Google Analytics Help (last visited Nov. 1, 2022), https://support.google.com/analytics/answer/10596866. Google even created a guide called “Beyond Last Click Attribution” to help advertisers select the most appropriate model. See Beyond Last Click Attribution: Official Guide to Attribution Modeling in Google Ads, Google Ads Help (last visited Nov. 1, 2022), https://support.google.com/google-ads/answer/7003286.

[168] See Joan Arensman & Wilfred Yeung, Move Beyond Last Click Attribution in AdWords, Google Blog (May 10, 2016), https://adwords.googleblog.com/2016/05/move-beyond-last-click-attribution.html.

[169] How Does Conversion Tracking Work?, Microsoft Advertising (n.d.) https://help.ads.microsoft.com/#apex/ads/en/56710/2.

[170] Nir Elharar, How to Choose the Right Marketing Attribution Model for Your Content, Outbrain (Apr. 8, 2019), https://www.outbrain.com/blog/marketing-attribution-model-content.

[171] Texas Complaint, supra note 1, at ¶ 195.

[172] Scott Morton & Dinielli, supra note 15, at 17

[173] Texas Complaint, supra note 1, at ¶ 351.

[174] See Demsetz, supra note 81.

Continue reading
Antitrust & Consumer Protection

Who Moderates the Moderators?: A Law & Economics Approach to Holding Online Platforms Accountable Without Destroying the Internet

ICLE White Paper A comprehensive survey of the law & economics of online intermediary liability, which concludes that any proposed reform of Section 230 must meaningfully reduce the incidence of unlawful or tortious online content such that its net benefits outweigh its net costs.

Executive Summary

A quarter-century since its enactment as part of the Communications Decency Act of 1996, a growing number of lawmakers have been seeking reforms to Section 230. In the 116th Congress alone, 26 bills were introduced to modify the law’s scope or to repeal it altogether. Indeed, we have learned much in the last 25 years about where Section 230 has worked well and where it has not.

Although the current Section 230 reform debate popularly—and politically—revolves around when platforms should be forced to host certain content politically favored by one faction (i.e., conservative speech) or when they should be forced to remove certain content disfavored by another (i.e., alleged “misinformation” or hate speech), this paper does not discuss, nor even entertain, such reform proposals. Rather, such proposals are (and should be) legal non-starters under the First Amendment.

Indeed, such reforms are virtually certain to harm, not improve, social welfare: As frustrating as imperfect content moderation may be, state-directed speech codes are much worse. Moreover, the politicized focus on curbing legal and non-tortious speech undermines the promise of making any progress on legitimate issues: The real gains to social welfare will materialize from reforms that better align the incentives of online platforms with the social goal of deterring or mitigating illegal or tortious conduct.

Section 230 contains two major provisions: (1) that an online service provider will not be treated as the speaker or publisher of the content of a third party, and (2) that actions taken by an online service provider to moderate the content hosted by its services will not trigger liability. In essence, Section 230 has come to be seen as a broad immunity provision insulating online platforms from liability for virtually all harms caused by user-generated content hosted by their services, including when platforms might otherwise be deemed to be implicated because of the exercise of their editorial control over that content.

To the extent that the current legal regime permits social harms online that exceed concomitant benefits, it should be reformed to deter those harms if such reform can be accomplished at sufficiently low cost. The salient objection to Section 230 reform is not one of principle, but of practicality: are there effective reforms that would address the identified harms without destroying (or excessively damaging) the vibrant Internet ecosystem by imposing punishing, open-ended legal liability? We believe there are.

First and foremost, we believe that Section 230(c)(1)’s intermediary-liability protections for illegal or tortious conduct by third parties can and should be conditioned on taking reasonable steps to curb such conduct, subject to procedural constraints that will prevent a tide of unmeritorious litigation.

This basic principle is not without its strenuous and thoughtful detractors, of course. A common set of objections to Section 230 reform has grown out of legitimate concerns that the economic and speech gains that have accompanied the rise of the Internet over the last three decades would be undermined or reversed if Section 230’s liability shield were weakened. Our paper thus establishes a proper framework for evaluating online intermediary liability and evaluates the implications of the common objections to Section 230 reform within that context. Indeed, it is important to take those criticisms seriously, as they highlight many of the pitfalls that could attend imprudent reforms. We examine these criticisms both to find ways to incorporate them into an effective reform agenda, and to highlight where the criticisms themselves are flawed.

Our approach is rooted in the well-established law & economics analysis of liability rules and civil procedure, which we use to introduce a framework for understanding the tradeoffs faced by online platforms under differing legal standards with differing degrees of liability for the behavior and speech of third-party users. This analysis is bolstered by a discussion of common law and statutory antecedents that allow us to understand how courts and legislatures have been able to develop appropriate liability regimes for the behavior of third parties in different, but analogous, contexts. Ultimately, and drawing on this analysis, we describe the contours of our recommended duty-of-care standard, along with a set of necessary procedural reforms that would help to ensure that we retain as much of the value of user-generated content as possible, while encouraging platforms to better police illicit and tortious content on their services.

The Law & Economics of Online Intermediary Liability

An important goal of civil tort law is to align individual incentives with social welfare such that costly behavior is deterred and individuals are encouraged to take optimal levels of precaution against risks of injury. Not uncommonly, the law even holds intermediaries—persons or businesses that have a special relationship with offenders or victims—accountable when they are the least-cost avoider of harms, even when those harms result from the actions of third parties.

Against this background, the near-complete immunity granted to online platforms by Section 230 for harms caused by platform users is a departure from normal rules governing intermediary behavior. This immunity has certainly yielded benefits in the form of more user-generated online content and the ability of platforms to moderate without fear of liability. But it has also imposed costs to the extent that broad immunity fails to ensure that illegal and tortious conduct are optimally deterred online.

The crucial question for any proposed reform of Section 230 is whether it could pass a cost-benefit test—that is, whether it is likely to meaningfully reduce the incidence of unlawful or tortious online content while sufficiently addressing the objections to the modification of Section 230 immunity, such that its net benefits outweigh its net costs. In the context of both criminal and tort law generally, this balancing is sought through a mix of direct and collateral enforcement actions that, ideally, minimizes the total costs of misconduct and enforcement. Section 230, as it is currently construed, however, eschews entirely the possibility of collateral liability, foreclosing an important mechanism for properly adjusting the overall liability scheme.

But there is no sound reason to think this must be so. While many objections to Section 230 reform—that is, to the imposition of any amount of intermediary liability—are well-founded, they also frequently suffer from overstatement or unsupported suppositions about the magnitude of harm. At the same time, some of the expressed concerns are either simply misplaced or serve instead as arguments for broader civil-procedure reform (or decriminalization), rather than as defenses of the particularized immunity afforded by Section 230 itself.

Unfortunately, the usual course of discussion typically fails to acknowledge the tradeoffs that Section 230—and its reform—requires. These tradeoffs embody value judgments about the quantity and type of speech that should exist online, how individuals threatened by tortious and illegal conduct online should be protected, how injured parties should be made whole, and what role online platforms should have in helping to negotiate these tradeoffs. This paper’s overarching goal, even more important than any particular recommendation, is to make explicit what these tradeoffs entail.

Of central importance to the approach taken in this paper, our proposals presuppose a condition frequently elided by defenders of the Section 230 status quo, although we believe nearly all of them would agree with the assertion: that there is actual harm—violations of civil law and civil rights, violations of criminal law, and tortious conduct—that occurs on online platforms and that imposes real costs on individuals and society at-large. Our proposal proceeds on the assumption, in other words, that there are very real, concrete benefits that would result from demanding greater accountability from online intermediaries, even if that also leads to “collateral censorship” of some lawful speech.

It is necessary to understand that the baseline standard for speech and conduct—both online and offline—is not “anything goes,” but rather self-restraint enforced primarily by incentives for deterrence. Just as the law may deter some amount of speech, so too is speech deterred by fear of reprisal, threat of social sanction, and people’s baseline sense of morality. Some of this “lost” speech will be over-deterred, but one hopes that most deterred speech will be of the harmful or, at least, low-value sort (or else, the underlying laws and norms should be changed). Moreover, not even the most valuable speech is of infinite value, such that any change in a legal regime that results in relatively less speech can be deemed per se negative.

A proper evaluation of the merits of an intermediary-liability regime must therefore consider whether user liability alone is insufficient to deter bad actors, either because it is too costly to pursue remedies against users directly, or because the actions of platforms serve to make it less likely that harmful speech or conduct is deterred. The latter concern, in other words, is that intermediaries may—intentionally or not—facilitate harmful speech that would otherwise be deterred (self-censored) were it not for the operation of the platform.

Arguably, the incentives offered by each of the forces for self-restraint are weakened in the context of online platforms. Certainly everyone is familiar with the significantly weaker operation of social norms in the more attenuated and/or pseudonymous environment of online social interaction. While this environment facilitates more legal speech and conduct than in the offline world, it also facilitates more illegal and tortious speech and conduct. Similarly, fear of reprisal (i.e., self-help) is often attenuated online, not least because online harms are often a function of the multiplier effect of online speech: it is frequently not the actions of the original malfeasant actor, but those of neutral actors amplifying that speech or conduct, that cause harm. In such an environment, the culpability of the original actor is surely mitigated and may be lost entirely. Likewise, in the normal course, victims of tortious or illegal conduct and law enforcers acting on their behalf are the primary line of defense against bad actors. But the relative anonymity/pseudonymity of online interactions may substantially weaken this defense.

Many argue, nonetheless, that holding online intermediaries responsible for failing to remove offensive content would lead to a flood of lawsuits that would ultimately overwhelm service providers, and sub-optimally diminish the value these firms provide to society—a so-called “death by ten thousand duck-bites.” Relatedly, firms that face potentially greater liability would be forced to internalize some increased—possibly exorbitant—degree of compliance costs even if litigation never materialized.

There is certainly some validity to these concerns. Given the sheer volume of content online and the complexity, imprecision, and uncertainty of moderation processes, even very effective content-moderation algorithms will fail to prevent all actionable conduct, which could result in many potential claims. At the same time, it can be difficult to weed out unlawful conduct without inadvertently over-limiting lawful activity.

But many of the unique features of online platforms also cut against the relaxation of legal standards online. Among other things—and in addition to the attenuated incentives for self-restraint mentioned above—where traditional (offline) media primarily host expressive content, online platforms facilitate a significant volume of behavior and commerce that isn’t purely expressive. Tortious and illegal content tends to be less susceptible to normal deterrence online than in other contexts, as individuals can hide behind varying degrees of anonymity. Even users who are neither anonymous nor pseudonymous can sometimes prove challenging to reach with legal process. And, perhaps most importantly, online content is disseminated both faster and more broadly than offline media.

At the same time, an increase in liability risk for online platforms may lead not to insurmountable increases in litigation costs, but to other changes that may be less privately costly to a platform than litigation, and which may be socially desirable. Among these changes may be an increase in preemptive moderation; smaller, more specialized platforms and/or tighter screening of platform participants on the front end (both of which are likely to entail stronger reputational and normative constraints); the establishment of more effective user-reporting and harm-mitigation mechanisms; the development and adoption of specialized insurance offerings; or any number of other possible changes.

Thus the proper framework for evaluating potential reforms to Section 230 must include the following considerations: To what degree would shifting the legal rules governing platform liability increase litigation costs, increase moderation costs, constrain the provision of products and services, increase “collateral censorship,” and impede startup formation and competition, all relative to the status quo, not to some imaginary ideal state? Assessing the marginal changes in all these aspects entails, first, determining how they are affected by the current regime. It then requires identifying both the direction and magnitude of change that would result from reform. Next, it requires evaluating the corresponding benefits that legal change would bring in increasing accountability for tortious or criminal conduct online. And, finally, it necessitates hazarding a best guess of the net effect. Virtually never is this requisite analysis undertaken with any real degree of rigor. Our paper aims to correct that.

A Proposal for Reform

What is called for is a properly scoped reform that applies the same political, legal, economic, and other social preferences offline as online, aimed at ensuring that we optimally deter illegal content without losing the benefits of widespread user-generated content. Properly considered, there is no novel conflict between promoting the flow of information and protecting against tortious or illegal conduct online. While the specific mechanisms employed to mediate between these two principles online and offline may differ—and, indeed, while technological differences can alter the distribution of costs and benefits in ways that must be accounted for—the fundamental principles that determine the dividing line between actionable and illegal or tortious content offline can and should be respected online, as well. Indeed, even Google has argued for exactly this sort of parity, recently calling on the Canadian government to “take care to ensure that their proposal does not risk creating different legal standards for online and offline environments.”

Keeping in mind the tradeoffs embedded in Section 230, we believe that, in order to more optimally mitigate truly harmful conduct on Internet platforms, intermediary-liability law should develop a “duty-of-care” standard that obliges service providers to reasonably protect their users and others from the foreseeable illegal or tortious acts of third parties. As a guiding principle, we should not hold online platforms vicariously liable for the speech of third parties, both because of the sheer volume of user-generated content online and the generally attenuated relationship between online platforms and users, as well as because of the potentially large costs to overly chilling free expression online. But we should place at least the same burden to curb unlawful behavior on online platforms that we do on traditional media operating offline.

Nevertheless, we hasten to add that this alone would likely be deficient: adding an open-ended duty of care to the current legal system could generate a volume of litigation that few, if any, platform providers could survive. Instead, any new duty of care should be tempered by procedural reforms designed to ensure that only meritorious litigation survives beyond a pre-discovery motion to dismiss.

Procedurally, Section 230 immunity protects service providers not just from liability for harm caused by third-party content, but also from having to incur substantial litigation costs. Concern for judicial economy and operational efficiency are laudable, of course, but such concerns are properly addressed toward minimizing the costs of litigation in ways that do not undermine the deterrent and compensatory effects of meritorious causes of action. While litigation costs that exceed the minimum required to properly assign liability are deadweight losses to be avoided, the cost of liability itself—when properly found—ought to be borne by the party best positioned to prevent harm. Thus, a functional regime will attempt to accurately balance excessive litigation costs against legitimate and necessary liability costs.

In order to achieve this balance, we recommend that, while online platforms should be responsible for adopting reasonable practices to mitigate illegal or tortious conduct by their users, they should not face liability for communication torts (e.g., defamation) arising out of user-generated content unless they fail to remove content they knew or should have known was defamatory.  Further, we propose that Section 230(c)(2)’s safe harbor should remain in force and that, unlike for traditional media operating offline, the act of reasonable content moderation by online platforms should not, by itself, create liability exposure.

In sum, we propose that Section 230 should be reformed to incorporate the following high-level elements, encompassing two major components: first, a proposal to alter the underlying intermediary-liability rules to establish a “duty of care” requiring adherence to certain standards of conduct with respect to user-generated content; and second, a set of procedural reforms that are meant to phase in the introduction of the duty of care and its refinement by courts and establish guardrails governing litigation of the duty.

Proposed Basic Liability Rules

Online intermediaries should operate under a duty of care to take appropriate measures to prevent or mitigate foreseeable harms caused by their users’ conduct.

Section 230(c)(1) should not preclude intermediary liability when an online service provider fails to take reasonable care to prevent non-speech-related tortious or illegal conduct by its users

As an exception to the general reasonableness rule above, Section 230(c)(1) should preclude intermediary liability for communication torts arising out of user-generated content unless an online service provider fails to remove content it knew or should have known was defamatory.

Section 230(c)(2) should provide a safe harbor from liability when an online service provider does take reasonable steps to moderate unlawful conduct. In this way, an online service provider would not be held liable simply for having let harmful content slip through, despite its reasonable efforts.

The act of moderation should not give rise to a presumption of knowledge. Taking down content may indicate an online service provider knows it is unlawful, but it does not establish that the online service provider should necessarily be liable for a failure to remove it anywhere the same or similar content arises.

But Section 230 should contemplate “red-flag” knowledge, such that a failure to remove content will not be deemed reasonable if an online service provider knows or should have known that it is illegal or tortious. Because the Internet creates exceptional opportunities for the rapid spread of harmful content, a reasonableness obligation that applies only ex ante may be insufficient. Rather, it may be necessary to impose certain ex post requirements for harmful content that was reasonably permitted in the first instance, but that should nevertheless be removed given sufficient notice.

Proposed Procedural Reforms

In order to effect the safe harbor for reasonable moderation practices that nevertheless result in harmful content, we propose the establishment of “certified” moderation standards under the aegis of a multi-stakeholder body convened by an overseeing government agency. Compliance with these standards would operate to foreclose litigation at an early stage against online service providers in most circumstances. If followed, a defendant could provide its certified moderation practices as a “certified answer” to any complaint alleging a cause of action arising out of user-generated content. Compliant practices will merit dismissal of the case, effecting a safe harbor for such practices.

In litigation, after a defendant answers a complaint with its certified moderation practices, the burden would shift to the plaintiff to adduce sufficient evidence to show that the certified standards were not actually adhered to. Such evidence should be more than mere res ipsa loquitur; it must be sufficient to demonstrate that the online service provider should have been aware of a harm or potential harm, that it had the opportunity to cure or prevent it, and that it failed to do so. Such a claim would need to meet a heightened pleading requirement, as for fraud, requiring particularity.

Finally, we believe any executive or legislative oversight of this process should be explicitly scheduled to sunset. Once the basic system of intermediary liability has had some time to mature, it should be left to courts to further manage and develop the relevant common law.

Our proposal does not demand perfection from online service providers in their content-moderation decisions—only that they make reasonable efforts. What is appropriate for YouTube, Facebook, or Twitter will not be the same as what’s appropriate for a startup social-media site, a web-infrastructure provider, or an e-commerce platform. A properly designed duty-of-care standard should be flexible and account for the scale of a platform, the nature and size of its user base, and the costs of compliance, among other considerations. Indeed, this sort of flexibility is a benefit of adopting a “reasonableness” standard, such as is found in common law negligence. Allowing courts to apply the flexible common law duty of reasonable care would also enable the jurisprudence to evolve with the changing nature of online intermediaries, the problems they pose, and the moderating technologies that become available.

Read the full working paper here.

Continue reading
Innovation & the New Economy

Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins

Scholarship Dystopian thinking is pervasive within the antitrust community. Unlike entrepreneurs, antitrust scholars and policy makers often lack the imagination to see how competition will emerge and enable entrants to overthrow seemingly untouchable incumbents.

Introduction

The dystopian novel is a powerful literary genre. It has given us such masterpieces as Nineteen Eighty-Four, Brave New World, Fahrenheit 451, and Animal Farm. Though these novels often shed light on some of the risks that contemporary society faces and the zeitgeist of the time when they were written, they almost always systematically overshoot the mark (whether intentionally or not) and severely underestimate the radical improvements commensurate with the technology (or other causes) that they fear. Nineteen Eighty-Four, for example, presciently saw in 1949 the coming ravages of communism, but it did not guess that markets would prevail, allowing us all to live freer and more comfortable lives than any preceding generation. Fahrenheit 451 accurately feared that books would lose their monopoly as the foremost medium of communication, but it completely missed the unparalleled access to knowledge that today’s generations enjoy. And while Animal Farm portrayed a metaphorical world where increasing inequality is inexorably linked to totalitarianism and immiseration, global poverty has reached historic lows in the twenty-first century, and this is likely also true of global inequality. In short, for all their literary merit, dystopian novels appear to be terrible predictors of the quality of future human existence. The fact that popular depictions of the future often take the shape of dystopias is more likely reflective of the genre’s entertainment value than of society’s impending demise.

But dystopias are not just a literary phenomenon; they are also a powerful force in policy circles. For example, in the early 1970s, the so-called Club of Rome published an influential report titled The Limits to Growth. The report argued that absent rapid and far-reaching policy shifts, the planet was on a clear path to self-destruction:

If the present growth trends in world population, industrialization, pollution, food production, and resource depletion continue unchanged, the limits to growth on this planet will be reached sometime within the next one hundred years. The most probable result will be a rather sudden and uncontrollable decline in both population and industrial capacity.

Halfway through the authors’ 100-year timeline, however, available data suggests that their predictions were way off the mark. While the world’s economic growth has continued at a breakneck pace, extreme poverty, famine, and the depletion of natural resources have all decreased tremendously.

For all its inaccurate and misguided predictions, dire tracts such as The Limits to Growth perhaps deserve some of the credit for the environmental movements that followed. But taken at face value, the dystopian future along with the attendant policy demands put forward by works like The Limits to Growth would have had cataclysmic consequences for, apparently, extremely limited gain. The policy incentive is to strongly claim impending doom. There’s no incentive to suggest “all is well,” and little incentive even to offer realistic, caveated predictions.

As we argue in this Article, antitrust scholarship and commentary is also afflicted by dystopian thinking. Today, antitrust pessimists have set their sights predominantly on the digital economy—“big tech” and “big data”—alleging a vast array of potential harms. Scholars have argued that the data created and employed by the digital economy produces network effects that inevitably lead to tipping and more concentrated markets. In other words, firms will allegedly accumulate insurmountable data advantages and thus thwart competitors for extended periods of time. Some have gone so far as to argue that this threatens the very fabric of western democracy. Other commentators have voiced fears that companies may implement abusive privacy policies to shortchange consumers. It has also been said that the widespread adoption of pricing algorithms will almost inevitably lead to rampant price discrimination and algorithmic collusion. Indeed, “pollution” from data has even been likened to the environmental pollution that spawned The Limits to Growth: “If indeed ‘data are to this century what oil was to the last one,’ then—[it’s] argue[d]—data pollution is to our century what industrial pollution was to the last one.”

Some scholars have drawn explicit parallels between the emergence of the tech industry and famous dystopian novels. Professor Shoshana Zuboff, for instance, refers to today’s tech giants as “Big Other.” In an article called “Only You Can Prevent Dystopia,” one New York Times columnist surmised:

The new year is here, and online, the forecast calls for several seasons of hell. Tech giants and the media have scarcely figured out all that went wrong during the last presidential election—viral misinformation, state-sponsored propaganda, bots aplenty, all of us cleaved into our own tribal reality bubbles—yet here we go again, headlong into another experiment in digitally mediated democracy.

I’ll be honest with you: I’m terrified . . . There’s a good chance the internet will help break the world this year, and I’m not confident we have the tools to stop it.

Parallels between the novel Nineteen Eighty-Four and the power of large digital platforms were also plain to see when Epic Games launched an antitrust suit against Apple and its App Store in August 2020. Indeed, Epic Games released a short video clip parodying Apple’s famous “1984” ad (which upon its release was itself widely seen as a critique of the tech incumbents of the time).

Similarly, a piece in the New Statesman, titled “Slouching Towards Dystopia: The Rise of Surveillance Capitalism and the Death of Privacy,” concluded that: “Our lives and behaviour have been turned into profit for the Big Tech giants—and we meekly click ‘Accept.’ How did we sleepwalk into a world without privacy?”

Finally, a piece published in the online magazine Gizmodo asked a number of experts whether we are “already living in a tech dystopia.” Some of the responses were alarming, to say the least:

I’ve started thinking of some of our most promising tech, including machine learning, as like asbestos: … it’s really hard to account for, much less remove, once it’s in place; and it carries with it the possibility of deep injury both now and down the line.

. . . .

We live in a world saturated with technological surveillance, democracy-negating media, and technology companies that put themselves above the law while helping to spread hate and abuse all over the world.

Yet the most dystopian aspect of the current technology world may be that so many people actively promote these technologies as utopian.

Antitrust pessimism is not a new phenomenon, and antitrust enforcers and scholars have long been fascinated with—and skeptical of—high tech markets. From early interventions against the champions of the Second Industrial Revolution (oil, railways, steel, etc.) through the mid-twentieth century innovations such as telecommunications and early computing (most notably the RCA, IBM, and Bell Labs consent decrees in the US) to today’s technology giants, each wave of innovation has been met with a rapid response from antitrust authorities, copious intervention-minded scholarship, and waves of pessimistic press coverage. This is hardly surprising given that the adoption of antitrust statutes was in part a response to the emergence of those large corporations that came to dominate the Second Industrial Revolution (despite the numerous radical innovations that these firms introduced in the process). Especially for unilateral conduct issues, it has long been innovative firms that have drawn the lion’s share of cases, scholarly writings, and press coverage.

Underlying this pessimism is a pervasive assumption that new technologies will somehow undermine the competitiveness of markets, imperil innovation, and entrench dominant technology firms for decades to come. This is a form of antitrust dystopia. For its proponents, the future ushered in by digital platforms will be a bleak one—despite abundant evidence that information technology and competition in technology markets have played significant roles in the positive transformation of society. This tendency was highlighted by economist Ronald Coase:

[I]f an economist finds something—a business practice of one sort or another—that he does not understand, he looks for a monopoly explanation. And as in this field we are very ignorant, the number of ununderstandable practices tends to be rather large, and the reliance on a monopoly explanation, frequent.

“The fear of the new—and the assumption that ‘ununderstandable practices’ emerge from anticompetitive impulses and generate anticompetitive effects—permeates not only much antitrust scholarship, but antitrust doctrine as well.” While much antitrust doctrine is capable of accommodating novel conduct and innovative business practices, antitrust law—like all common law-based legal regimes—is inherently backward looking: it primarily evaluates novel arrangements with reference to existing or prior structures, contracts, and practices, often responding to any deviations with “inhospitality.” As a result, there is a built-in “nostalgia bias” throughout much of antitrust that casts a deeply skeptical eye upon novel conduct.

“The upshot is that antitrust scholarship often emphasizes the risks that new market realities create for competition, while idealizing the extent to which previous market realities led to procompetitive outcomes.” Against this backdrop, our Article argues that the current wave of antitrust pessimism is premised on particularly questionable assumptions about competition in data-intensive markets.

Part I lays out the theory and identifies the sources and likely magnitude of both the dystopia and nostalgia biases. Having examined various expressions of these two biases, the Article argues that their exponents ultimately seek to apply a precautionary principle within the field of antitrust enforcement, made most evident in critics’ calls for authorities to shift the burden of proof in a subset of proceedings.

Part II discusses how these arguments play out in the context of digital markets. It argues that economic forces may undermine many of the ills that allegedly plague these markets—and thus the case for implementing a form of precautionary antitrust enforcement. For instance, because data is ultimately just information, it will prove exceedingly difficult for firms to hoard data for extended periods of time. Instead, a more plausible risk is that firms will underinvest in the generation of data. Likewise, the main challenge for digital economy firms is not so much to obtain data, but to create valuable goods and hire talented engineers to draw insights from the data these goods generate. Recent empirical findings suggest, for example, that data economy firms don’t benefit as much as often claimed from data network effects or increasing returns to scale.

Part III reconsiders the United States v. Microsoft Corp. antitrust litigation—the most important precursor to today’s “big tech” antitrust enforcement efforts—and shows how it undermines, rather than supports, pessimistic antitrust thinking. It shows that many of the fears that were raised at the time didn’t transpire (for reasons unrelated to antitrust intervention). Rather, pessimists missed the emergence of key developments that greatly undermined Microsoft’s market position, and greatly overestimated Microsoft’s ability to thwart its competitors. Those circumstances—particularly revolving around the alleged “applications barrier to entry”—have uncanny analogues in the data markets of today. We thus explain how and why the Microsoft case should serve as a cautionary tale for current enforcers confronted with dystopian antitrust theories.

In short, the Article exposes a form of bias within the antitrust community. Unlike entrepreneurs, antitrust scholars and policy makers often lack the imagination to see how competition will emerge and enable entrants to overthrow seemingly untouchable incumbents. New technologies are particularly prone to this bias because there is a shorter history of competition to go on and thus less tangible evidence of attrition in these specific markets. The digital future is almost certainly far less bleak than many antitrust critics have suggested and yet the current swath of interventions aimed at reining in “big tech” presume. This does not mean that antitrust authorities should throw caution to the wind. Instead, policy makers should strive to maintain existing enforcement thresholds, which exclude interventions that are based solely on highly speculative theories of harm.

Read the full white paper here.

Continue reading
Antitrust & Consumer Protection

Technology Mergers and the Market for Corporate Control

Scholarship Forthcoming in the Missouri Law Journal, ICLE scholars scrutinize recent scholarship regarding so-called "kill zones" and "killer acquisitions" and the pitfalls that would accompany attempts to change existing merger rules and thresholds to account for them.

Abstract

A growing number of policymakers and scholars are calling for tougher rules to curb corporate acquisitions. But these appeals are premature. There is currently little evidence to suggest that mergers systematically harm consumer welfare. More importantly, scholars fail to identify alternative institutional arrangements that would capture the anticompetitive mergers that evade prosecution without disproportionate false positives and administrative costs. Their proposals thus fail to meet the requirements of the error-cost framework.

Several high-profile academic articles and reports claim to have identified important gaps in current merger enforcement rules, particularly with respect to tech and pharma acquisitions involving nascent and potential competitors—so-called “killer acquisitions” and “kill zones.” As a result of these perceived deficiencies, scholars and enforcers have called for tougher rules, including the introduction of lower merger filing thresholds and substantive changes, such as the inversion of the burden of proof when authorities review mergers and acquisitions in the digital platform industry. Meanwhile, and seemingly in response to the increased political and advocacy pressures around the issue, U.S. antitrust enforcers have recently undertaken several enforcement actions directly targeting such acquisitions.

As this paper discusses, however, these proposals tend to overlook the important tradeoffs that would ensue from attempts to decrease the number of false positives under existing merger rules and thresholds. While merger enforcement ought to be mindful of these possible theories of harm, the theories and evidence are not nearly as robust as many proponents suggest. Most importantly, there is insufficient basis to conclude that the costs of permitting the behavior they identify is greater than the costs would be of increasing enforcement to prohibit it.

Our work draws from two key strands of economic literature that are routinely overlooked (or summarily dismissed) by critics of the status quo. For a start, as Frank Easterbrook argued in his pioneering work on The Limits of Antitrust, antitrust enforcement is anything but costless. In the case of merger enforcement, not only is it expensive for agencies to detect anticompetitive deals, but overbearing rules may deter beneficial merger activity that creates value for consumers. Indeed, not only are most mergers welfare-enhancing, but barriers to merger activity have been shown to significantly, and negatively, affect early company investment.

Second, critics mistake the nature of causality. Scholars routinely surmise that incumbents use mergers to shield themselves from competition. Acquisitions are thus seen as a means of eliminating competition. But this overlooks an important alternative: It is at least plausible that incumbents’ superior managerial or other capabilities make them the ideal purchasers for entrepreneurs and startup investors who are looking to sell. This dynamic is likely to be amplified where the acquirer and acquiree operate in overlapping lines of business. In other words, competitive advantage, and the ability to profitably acquire other firms, might be caused by business acumen rather than anticompetitive behavior.

Thus, significant and high-profile M&A activity involving would-be competitors may be the procompetitive byproduct of a well-managed business, rather than anticompetitive efforts to stifle competition. Critics systematically overlook this possibility. Indeed, Henry Manne’s seminal work on Mergers and Market for Corporate Control—the first to argue that mergers are a means of applying superior management practices to new assets—is almost never cited by contemporary researchers in this space. Our paper attempts to set the record straight.

Read the full white paper here.

Continue reading
Antitrust & Consumer Protection

A Dynamic Analysis of Broadband Competition

ICLE White Paper The instinct to promote broadband network buildout is understandable, but precisely how that infrastructure funding is deployed will determine whether such proposals succeed or fail.

The 117th Congress is considering whether to devote significant federal resources toward promoting broadband access in underserved communities. Legislative proposals to do so include President Joe Biden’s draft American Jobs Plan—a $2.3 trillion budget-reconciliation package that sets aside $100 billion for broadband infrastructure. They also include the Accessible, Affordable Internet for All Act, which would create a $79.5 billion federal program.

The instinct to promote network buildout is understandable, particularly in the wake of the COVID-19 pandemic and the various socioeconomic disparities it highlighted. But precisely how that infrastructure funding is deployed will determine whether such proposals succeed or fail.

In fact, the U.S. broadband market is already healthy, and in most cases, competitive outcomes are close to optimal. Charges that broadband markets are dominated by monopolies or oligopolies and that they are therefore stagnant, over-priced, and of low quality do not comport with the empirical and economic realities. To take but one example, even with the overall rise of prices across the economy, and in the face of surging demand during the COVID-19 pandemic, U.S. broadband prices fell.

Concentration is a poor predictor of competitiveness, and broadband markets with even a small number of competitors can be—and are—quite healthy. Indeed, the multi-year, multi-billion-dollar investment plans broadband firms execute—amid constant pressure from alternative modes of Internet access like 5G, fixed wireless, and satellite—tell the story of a highly competitive, dynamic market.

To be sure, there are a few areas where there has been no meaningful wireline broadband buildout: Approximately 4.4 percent of the U.S. population does not have access to at least 25/3 Mbps fixed service. Even then, however, many of those areas are served by wireless Internet service providers (WISPs), cellular broadband, and/or satellite service.

But while the digital divide—both rural and urban—may be real, that fact alone does not justify wholesale intervention into broadband markets. Instead, the actual scope of the problem should be assessed, and policies tailored to remedy specific needs. The policies required to reach that stubborn 4.4 percent tail of broadband rollout are likely to be very different than those that facilitated the buildout of the first 95.6 percent.

Policies designed to close the digital divide should have two broad features: they should reach consumers where they are, and they should not disrupt the otherwise healthy broadband market. Reaching consumers where they are means targeting subsidies directly to consumers to make it more viable for existing providers to build out into new areas. Such policies should be technology-neutral and designed to stimulate demand to jumpstart markets that have otherwise proven too costly for any provider to enter. Avoiding disruption of healthy markets entails refraining from interventions that artificially introduce new competitors, skew investment planning by broadband providers, or dictate how and where providers should build networks.

There is much that can be done to encourage better and timelier broadband rollout, but not all solutions are equally effective. As we detail below, policymakers must choose carefully among competing options to realize the best possible result.

This paper aims to address common misconceptions associated with broadband competition that, in turn, undercut practical solutions for connecting the unconnected. It first details some of those misconceptions and contrasts them with the realities of current broadband markets. It then provides an overview of how to properly understand healthy competition in local broadband markets. It then provides a critique of commonly advanced proposals that are based on fundamental misunderstandings of how broadband markets work. And finally, it offers an approach to policy that incorporates a variety of solutions for connecting the unconnected.

Read the full white paper here.

* NOTE: Section 1.b was updated July 13, 2021, to reflect feedback regarding the paper’s interpretation of certain relevant economic studies.

Continue reading
Telecommunications & Regulated Utilities

Response to Ramsi Woodcock’s The Hidden Rules of a Modern Antitrust

Scholarship Woodcock’s bold claims ignore or misconstrue several critical aspects of the modern antitrust apparatus. Chief among these is the uncertainty that underpins antitrust enforcement, and the rule of reason’s role in decreasing this uncertainty.

In The Hidden Rules of a Modern Antitrust, Ramsi Woodcock argues that courts’ systematic use of the rule of reason, which underpins most of contemporary antitrust law, effectively amounts to an unwarranted blanket exemption from liability for potentially egregious practices. According to Woodcock, this is due to the interaction between the exorbitant cost of prosecuting cases under this standard (compared to the cost of enforcing per se rules), the courts’ increasing application of the rule of reason, and the shrinking budgets of antitrust enforcement agencies.

As this Response discusses, Woodcock’s bold claims ignore or misconstrue several critical aspects of the modern antitrust apparatus. Chief among these is the uncertainty that underpins antitrust enforcement, and the rule of reason’s role in decreasing this uncertainty. It takes time and experience for courts to form an opinion about the value of certain forms of business conduct, and rule of reason litigation increases the accuracy of all subsequent litigation—and the ability of both economic actors and antitrust enforcers to predict judicial outcomes and adjust their practices accordingly. This stands in stark contrast to Woodcock’s model, which assumes that courts are unable to differentiate between forms of ambiguous conduct (and yet simultaneously well informed enough about enforcers’ budget constraints to know whether they can “afford” to litigate under the rule of reason).

Winston Churchill famously quipped that “it has been said that democracy is the worst form of Government except for all those other forms that have been tried from time to time . . . .”  Much of the same could be said about the rule of reason. While it is certainly not perfect, policymakers have yet to find another standard that provides the same flexibility to accommodate ever-evolving forms of conduct with initially ambiguous effects on consumer welfare. Woodcock’s paper underplays these important virtues, while his more pointed critiques often miss the mark.

Continue reading
Antitrust & Consumer Protection