The “Big Tech” Self‐​Preferencing Delusion

Critics deem it unjust, for example, for Google to bump up Google Maps results in its search rankings. Or for Amazon to launch cheap, generic Basics products that compete with third‐​party sellers on Amazon’s marketplace. Or for Apple iPhones to come preloaded with Apple apps with similar functionality to non‐​Apple apps sold on the App Store. Critics consider it illegitimate for tech businesses to restrict third‐​party platform access based on the host’s sales interests or to use platform‐​acquired data from third‐​party sales to alter the host’s product offerings.

Behind such claims lies an instinct that platform neutrality and openness is always desirable: that a Big Tech company being vertically integrated or using vertical restraints—running a platform and self‐​preferencing its own products on it—would have damaging, anti‐​competitive effects absent government conduct restrictions. Senator Elizabeth Warren (D‑MA) expressed this impulse when she claimed, “You can be the umpire, or you can be a player, but you can’t be both.”¹

That sentiment now risks becoming legislation. Senator Amy Klobuchar’s American Innovation and Choice Online Act (AICO), co‐​sponsored by 12 senators, would restrict much self‐​preferential conduct that has been essential to major firms’ product offerings.² The bill would broadly prevent platforms from preferencing their own products, using data gleaned from third‐​party sales to change their product offerings, or treating their own products more favorably in search or rankings.

This brief puts self‐​preferencing and the AICO bill into a broader historical and economic context by summarizing the bill, explaining how self‐​preferencing is a common form of business conduct, discussing how past efforts to limit self‐​preferencing wrought harmful consequences for customers, describing AICO’s economic flaws, examining its practical difficulties, and reviewing the bill’s political mischief. Overall, this brief shows that AICO would harm consumers to benefit competitor interests, potentially stifling innovation across the tech industry.

Senator Klobuchar’s AICO bill would fundamentally alter business conduct on covered platforms.

The AICO bill would generally make it unlawful for major online platforms to engage in conduct that has produced tech products that consumers enjoy. Taken at face value, the bill text suggests that Apple’s iPhones would need to come without FaceTime and iMessage apps preloaded, that Google would need to remove its Maps link when Google users search for a nearby restaurant, and that aspects of Amazon Prime—a service used by nearly half of Americans—would be de facto banned (among other consequences).³

As written, the bill would make it unlawful for a major online platform to preference its own goods, limit the ability of other platform business users to compete with the platform’s products, or discriminate in its terms of services in ways that would “materially harm competition.”⁴

As a default, affected platforms would not be able to restrict business users from interoperating with software open to the platform’s own products or require bundling the purchase of the platform’s own services to obtain access or preferred platform status. A covered platform would also be banned from using data from third‐​party sales to improve its own products or from treating its own products favorably in search rankings or functionality.

Companies could, in theory, negate liability by using a “preponderance of evidence” to prove that their restraint is narrow and necessary to comply with laws, to protect safety, privacy, or personal data, or to maintain or substantially enhance the “core functionality” of the platform. There is also an exemption for subscription services.⁵ But most of these exemptions put the onus on the platforms to prove their innocence, while the subscription service exemption could fundamentally change how services are delivered—to consumers’ detriment.

The bill primarily targets the Big Tech giants, for now at least. Covered platforms are limited to “a website, online or mobile application, operating system, digital assistant, or online service.”⁶ Restrictions only apply to public companies that are a “critical trading partner” with 50 million monthly active U.S. users or more than 100,000 monthly active business users and that have had either annual sales or a market capitalization above $550 billion or 1 billion monthly active global users. Private online platforms beyond these user thresholds are covered if the company generates more than $30 billion in annual revenue.⁷

It is unclear exactly which companies would be ensnared by these definitions. What constitutes an active platform user remains undefined, for example. The bill has clearly been developed with the bête noires of today’s trustbusting movement in mind—in particular, Apple, Meta, Google, Amazon, and Microsoft. But the financial and user thresholds mean that private companies in groceries, professional services, and agriculture could be covered if they operated online platforms. In time, given fluctuations in market capitalization and more business shifting online, payment providers and other large businesses might find themselves covered too.

Only the Federal Trade Commission, the Department of Justice’s attorney general, and state attorneys general would be able to enforce the legislation’s provisions through civil litigation in federal district courts. Private‐​sector businesses would not be able to bring cases against other companies.

The potential fines for unlawful conduct are huge. The civil penalty for violations could be up to 15 percent of the company’s total U.S. revenue for the period when the violation occurred. Amazon’s annual net sales in North America in 2020 were $236 billion (excluding revenue from Amazon Web Services), implying maximum penalties could be tens of billions per year for unlawful self‐​preferencing conduct by the company in the United States.⁸

The bill treats self‐​preferencing as a nefarious, unusual activity. Yet self‐​preferencing is an extremely common business practice.

Supermarkets and other retailers, farmers’ markets, shopping malls, and sports stadiums regularly sell their own products alongside those of third‐​party businesses within their venues or marketplaces.⁹ They act like platforms (i.e., as intermediaries between buyers and sellers of goods, delivering services that reduce transaction costs or bring other benefits to their customer or business “users”).

Stores or marketplaces often self‐​preference their own products through differential product placement, by gleaning information from third‐​party sales, or by tying consumption of their services to third‐​party participation within the premises. They are profit‐​seeking businesses after all. How should a mall owner decide where stores are located without considering the implications for its profits?

Retail extensively features private‐​label products—goods manufactured and sold under the retailer’s name that then compete against other brand names within the host’s stores. Private Label Manufacturers Association data for 2020 showed that 18.1 percent of the dollar value of goods bought in supermarkets and 15.9 percent bought in drug stores were such goods.¹⁰ This vastly exceeds the share of Amazon’s own product sales on its marketplace—despite the company’s conduct being heavily criticized for this practice.¹¹

True, supermarkets technically act as goods resellers rather than platforms that facilitate direct seller‐​to‐​buyer sales. But this is a difference in contracting rather than a fundamental economic distinction. Your local grocery store engages in plenty of self‐​preferencing, including using in‐​store bakeries that produce pastries that compete against other brand equivalents on shelves. Supermarkets decide which third‐​party brands to stock, examine sales data to shape offerings, and place products to enhance the store’s long‐​run profitability.

And this benefits consumers. Bundling grocery and non‐​grocery services, alongside some in‐​house production, can deliver lower prices or more variety and reduce the time customers spend shopping. It’s generally not in the host’s interest to entirely foreclose competitors anyway, because a store selling popular goods increases customer flow. Why then, when it comes to large online platforms, does the AICO bill treat self‐​preferential conduct as highly damaging to competition?

Attacks on a firm’s ability to self‐​preference are nothing new. Through the 1960s, regulators heavily scrutinized vertical mergers and de facto prohibited self‐​preferential behavior such as “tying,” where use of a good or service is conditioned on the purchase of another of the company’s products.¹²

Then, hostility to vertical restraints was based on prevailing economic wisdom. Damaging market power was thought synonymous with high industry concentration. Restrictions that reduced concentration were deemed intrinsically desirable, even when larger firms benefited from economies of scale and so could sell cheaper goods.¹³ Under the dominant structure‐​conduct‐​performance model of the age, regulators worried that vertical practices that prevented competitors from accessing suppliers or consumers in one market could leverage a company’s market power into another sector.

Under the Department of Justice’s 1968 Merger Guidelines, vertical mergers accounting for 20 percent or more of market share, or acquisitions amounting to at least 10 percent of the market, were presumed anti‐​competitive and challenged. Claims that vertical mergers would yield efficiencies by eliminating the double mark‐​up problem (of monopoly manufacturers and monopoly distributors both seeking a price markup over marginal cost) were dismissed out of hand.¹⁴

Case law, meanwhile, made many vertical restraints effectively illegal. Manufacturers contracting for exclusive territories with distributors was effectively banned. Price discrimination by wholesalers to retailers was constrained by the Robinson‐​Patman Act.¹⁵ Antitrust policy sought to protect competition by protecting competitors rather than consumers.

These decades of crackdowns on vertical restraints undermined consumer welfare.¹⁶ In United States v. Arnold, Schwinn & Co., for example, the Supreme Court held that manufacturers specifying exclusive territories for goods sales was an unacceptable restraint of trade. Yet evidence since suggests that these restraints benefit consumers—not least by incentivizing distributors to provide information services without the threat that the consumer will absorb the information and then take their business elsewhere.

As consumer welfare was put at the center of antitrust policy, the Justice Department guidelines were rewritten in 1984, deprioritizing concentration and acknowledging the potential efficiencies of vertical restraints. The subsequent Vertical Restraints Guidelines were more permissive about most vertical conduct.¹⁷ Although the Clinton administration eventually rescinded these guidelines, authorities have approached both vertical integration and self‐​preferencing more pragmatically since, yielding lower prices and higher‐​quality goods and services for consumers.¹⁸

Extensive research of various industries including cement, beer, and tech, has found that vertical restraints can enhance economic efficiency, not least by improving logistical coordination.¹⁹ Other forms of self‐​preferencing have been found to improve outcomes for consumers by better distributing risk, incentivizing innovation, aligning manufacturer and distributor interests, or growing product demand by encouraging the provision of promotional services. In tech, research has concluded that Google’s entry into the camera app subsector benefited consumers by encouraging significant innovation in competitor apps on Google’s Android platform, for example.²⁰

There are theoretical cases where consumer welfare may be harmed by exclusionary vertical practices.²¹ Yet the empirical evidence suggests that where regulators cracked down on vertical mergers, the effects were harmful overall.²² Empirical surveys have overwhelmingly found vertical restraints also to be of little harm to consumers, with Federal Trade Commission economists concluding that they generally “appear to reduce price and/​or increase output.”²³

Given this evidence, a prudent antitrust policy should assume that most self‐​preferential conduct is benign. Yet Klobuchar’s bill is more evidence of policy shifting in the opposite direction. In September 2021, President Biden’s Federal Trade Commission rescinded the 2020 Vertical Merger Guidelines, claiming they represented a “flawed economic theory regarding purported pro‐​competitive benefits of mergers.”²⁴ The AICO bill would compound this trend by simply outlawing a range of vertical conduct, irrespective of the consumer consequences.

The AICO bill suffers from a host of economic flaws, including that it protects competitor interests over consumer interests, distorts competition, and limits consumer choice.

Tech analyst Benedict Evans has highlighted how when you installed a PC spreadsheet program in the 1980s, functionality was greatly limited compared with today.²⁵ Users had to buy separate programs to undertake what are now deemed simple tasks, such as printing spreadsheets in landscape.

That this can be done easily with a button click in Microsoft Excel or Google Sheets today is a boon for users. Nobody wants to have to pay for and install a completely different program for altering a single print setting. Bundling functions such as this within a program improves the customer experience.

Yet under the logic behind the AICO bill, Microsoft Excel’s built‐​in print option precludes competitor businesses from providing that service. The innovation is a vertical restraint against competition on Excel’s so‐​called platform. Under the caricatured view that says platforms should be nondiscriminatory and open, this self‐​preferencing could be deemed anti‐​competitive, as it closes off opportunities for third‐​party providers.

The AICO bill’s supporters would scoff that Excel isn’t an online platform. But the bill’s restrictions on self‐​preferencing could prevent or deter similarly useful product innovations on covered platforms that would otherwise improve life for consumers.

All production entails self‐​preferencing to a lesser or greater extent.²⁶ The question is what criteria should determine when self‐​preferencing is regarded as harmful as opposed to benign or beneficial?

Recent decades’ antitrust policy in the United States has been notionally grounded to a consumer welfare lodestar that says vertical restraints and self‐​preferencing are fine so long as they do not raise consumer prices or harm innovation. But the AICO bill is more restrictive. In the spirit of today’s trustbusting movement, it bans certain restraints if they harm third‐​party businesses that compete with platform hosts, irrespective of how consumers or future innovation are affected.

It should not be understated how profound the shift in approach would be. The consumer welfare goal of sustained lower prices and greater innovation allows harm to competitors. When a firm develops a cheap and efficient product, this hurts companies providing more expensive rival products. So seeing certain vertical restraints as necessarily anti‐​competitive because they harm competitors represents abandoning the consumer welfare standard.

Third‐​party businesses might not like it when tech platforms self‐​preference in ways detrimental to their business’s profitability. But it’s a stretch to suggest that consumers are harmed when Google search highlights its Maps result to help users find a location faster, or when iPhones come with Apple’s pre‐​installed camera and music apps, or when Amazon sells its own Basics yoga mats.

A bill that explicitly rules out conduct that can make consumers better off can therefore only be understood as an attempt to overturn the consumer welfare standard.

But wouldn’t a more open and neutral platform be better for consumers? Not necessarily. A broader range of products is not the only thing consumers value. And restrictions that prevent vertical restraints might make it uneconomic to provide other services that consumers do want.

Even putting pricing aside, when customers browse online, they often do not want a neutral platform without preferencing. Users or customers seek out a bundle of services that provide the best user experience. This might include accessing a diverse choice of products, yes. But consumers want a responsive platform without glitches, trusted payment and review systems, accurate searches, speedy access to useful information, and in some cases, privacy protections.

As Sam Bowman of the International Center for Law and Economics has noted, platform openness and neutrality (as AICO seeks to enforce) brings consumer costs as well as benefits.²⁷ Microsoft Windows is more open than Apple’s macOS but also less stable and more susceptible to viruses. Amazon is less open to new sellers than eBay but bundles in useful services, such as Amazon handling sales payments and the option for sellers to use Amazon delivery. The downside of openness or easier interoperability, with fewer restraints on third‐​party conduct, can be risks to security, functionality, or platform usability.

Platforms must manage these competing interests, thinking through how best to attract third‐​party businesses and customers. Inevitably, providing popular platforms means requiring some in‐​house functionality or imposing restrictions on what third parties can do to protect the platform’s integrity or to enhance users’ experiences.

The AICO bill recognizes this tension, to an extent, in that it would permit businesses self‐​preferencing if the conduct maintains or enhances the platform’s core functionality or protects security or privacy. But platform innovation entails experimenting with different services. The threat of conduct being deemed unlawful will deter businesses innovating with novel self‐​preferencing services or vertical restrictions.

Rather than seeing government’s role as protecting the beneficial consumer effects of open competition within economic markets (i.e., a well‐​defined product market in a geographic region), the AICO bill sees the government’s job as enforcing a level playing field of competition in every single marketplace covered. In that sense, companies’ platforms would be treated like public utilities—critical infrastructure that must be equally open to all participants.

This is a misguided conception of the competitive environment. One reason is because the Big Tech platforms themselves compete against other platforms. Amazon’s Marketplace competes with more open rivals such as eBay as well as with other selective online platforms such as Walmart, Target, Wayfair, and Best Buy. E‑commerce company Shopify has grown quickly, too, with analyst Benedict Evans noting that businesses using its software to build their own online stores now derive sales equivalent to 40 percent of Amazon’s Marketplace.²⁸

Apple competes with Android in producing phone operating systems. Google competes with DuckDuckGo and Bing in general search and with Amazon, Yelp, and others in more specialized search. Facebook competes for user attention with TikTok, YouTube, Snapchat, Twitter, and many other apps.

This competition “for the platform” undermines the idea that eliminating self‐​preferential discrimination on every platform is necessary for pro‐​consumer competitive pressure. Consumers can choose what degree of openness and neutrality they want across platforms. This creates a discovery process in delivering what consumers desire. The problem is that if bills such as AICO preclude against certain conduct, it is possible that new or modified platforms might not arise that better serve customers’ needs.

Self‐​preferencing by these businesses can actually enhance competition in new markets for the benefit of customers or third‐​party sellers too. If Amazon bundles Prime Video into its Prime package or uses third‐​party data to advertise a new product, then the harnessing of its data to make a new product economically viable could widen the choice available to customers in the new market it enters.

Netflix used to be touted as the “video streaming monopoly” and had a pre‐​pandemic market share of 64.6 percent in video streaming services. Yet by April 2021, this fell to 48 percent, driven by the ascendant vertically integrated platform services of Apple+, Disney+, and Amazon Prime Video.²⁹ Would these other services have been able to leverage into this market without the self‐​preferencing facilitated via their existing platform services or products?

Likewise, the existence of a large marketplace with significant network effects (benefits to users from more people using the platform) can facilitate competition by allowing small businesses to access new geographic markets. For example, if consumers like the pre‐​installed features of Apple’s iOS ecosystem such that they help generate more iPhone sales, then that is good for most other app producers on the App Store. Google’s self‐​preferencing of its own search engine on Android operating systems likewise helps make it financially viable to offer a lot of the products on that operating system for free, again to the benefit of consumers and other third‐​party businesses.

It is simply misguided to assume that pro‐​consumer competition requires level playing fields between goods and services provided on every platform.

The AICO bill is being sold as a set of pro‐​competitive restrictions, but the bill’s scope and curbs on conduct risk distorting competition in product markets in several important ways.

First, the bill would create a two‐​tier system of legal standards for forms of business conduct common both online and offline. This would prevent major online platforms from engaging in some of the same conduct undertaken by offline competitors.

Second, the bill would allow some self‐​preferencing if the conduct pertains to a core functionality. But this means that new businesses with vertical restraints at their core could largely circumvent these restrictions, while existing businesses, especially today’s Big Tech firms, might struggle to innovate with these imposed limitations. Established platforms can hardly claim that new conduct is a core functionality. So this legislation would put platforms innovating using new vertical restraints at a competitive disadvantage to those set up with such conduct at their core.

Finally, the use of financial status and user thresholds to determine which platforms are covered under the bill creates boundaries that could distort competition further. Combined with other anti‐​Big Tech legislation such as the Open App Markets Act, companies like Apple would be bound by self‐​preferencing laws, while smaller competitors that they are then forced to host would not. Businesses operating as online platforms would have to try to estimate whether an innovation would propel their platform to the scale where the conduct might be deemed problematic under antitrust law too.

Suppose, for example, that a self‐​preferencing platform begins to approach the threshold beyond which its action will be covered. Would the company spin off into a smaller platform to avoid liability? Would the platform cap its user base or start charging for subscription services? Platforms might adjust their conduct in many ways unrelated to consumer welfare, actually harming the competitive process.

In time, all these inequities may create a political push to broaden the scope of these restrictions. Yet doing so would exacerbate the consumer harms of outlawing self‐​preferential conduct.

Aside from the blatant economic flaws, AICO’s ambiguous scope, vague wording, and questionable evidentiary standard turns it into a regulatory nightmare.

The bill’s murky statutes would bring significant uncertainty over what conduct would be deemed unlawful. Vague bans would force covered businesses to prove that their conduct is a core platform feature in court, unless they could prove that it is essential for security or privacy.

Proponents have claimed to the press that popular products such as Amazon Prime shipping, Apple’s native apps, and Google Maps appearing in search results would not be classed as violations.³⁰ But it’s unclear why given the bill’s current text. This sort of ad hoc assurance is not the basis for good policy, and attempts to carve out exemptions for popular services like Amazon Prime have been half‐​hearted.

A change in the manager’s amendment to the bill is instructive. It permits a “fee for subscription service” that benefits platform users—a clear attempt to console critics claiming that the law would gut Amazon Prime. But because Amazon also charges Prime merchants fees to cover the logistics and funding of its free and fast shipping—a form of “preferred status” that would still be deemed unlawful—Prime’s essential funding and operating mechanism would be demolished, potentially forcing Amazon to close Prime services or face daunting penalties.³¹ Numerous other examples abound.

Several of AICO’s ambiguous key terms pave the way for uncertainty, gaming, or abuse.

As noted, AICO’s “Affirmative Defenses” section would permit self‐​preferential behavior in the name of maintaining or substantially improving a platform’s core functionality.³² But how is core functionality defined?

Amazon, for example, operates the most popular e‑commerce site, but with a Prime subscription, users can also enjoy access to other bundled goods such as Amazon’s streaming platform. Is Prime Video, which competes with the likes of Netflix and Hulu, also a platform core functionality?

The markets in which firms operate are rarely demarcated by solid lines and commonly overlap. AICO would do little to clarify these boundaries. Large platforms’ activities tend to innovate as technologies and consumer preferences shift. The whole “core functionality” concept risks becoming a highly uncertain target subject to the whims of regulators and haphazard decisionmaking.

Under Section 3 of the bill, firms would also be barred from conditioning access to their platform on the purchase of products “not part of or intrinsic to the covered platform itself.”³³ But what, again, does “intrinsic” mean? Is a pre‐​programmed Apple camera app intrinsic to an iPhone? Halide, a competing camera app, might argue not. What about Find My iPhone, a built‐​in feature that locates lost devices? Here, other security apps such as Tile might disagree. But integration of many of Apple’s native apps into an iPhone comes as a key selling point for many of its customers regardless.

With so many Big Tech firms creating native software and hardware ecosystems, defining what is intrinsic becomes nigh impossible. And often it’s parts of those ecosystems that do not initially seem core that make the platform’s services financially viable.

Finally, AICO’s “Unlawful Conduct” section also would prohibit self‐​preferential behavior that would “materially harm competition.”³⁴ But what does “materially harm” mean? Delivering innovations that lower prices inevitably harms one’s competitors. In fact, any behavior in which a firm develops a better, cheaper good can be construed as materially harming competition, if that is synonymous with competitor interests. How will antitrust regulators distinguish so‐​called material harm from pro‐​consumer healthy competition?

When filing legal claims under the bill, prosecutors would be beholden to a “preponderance of the evidence” standard of proof.³⁵ This means that they would have to demonstrate that the probability that a certain company’s behavior had violated AICO is at least greater than 50 percent.³⁶

Using this standard poses several problems, not least because the benchmark is lower than other common prosecutorial standards such as “clear and convincing evidence.”³⁷ The language therefore heightens the risk for identification errors—where normal business conduct that promotes consumer welfare is mistakenly prosecuted as a violation. Even if identification errors could be reversed, the damage to consumers and the threat of future court appeals or lawsuits would greatly diminish the incentive to bring back a canceled product or service.

Finally, it must be noted that AICO reeks of cronyism. In particular, the bill’s scope would ensnare Amazon, Apple, Google, and (possibly) TikTok—the bête noires of the new trustbusters and a company feared for its Chinese origins. But it would not cover, say, Walmart or Target, which also have very large online marketplaces and self‐​preference extensively within their online stores.

The Washington Examiner’s Tim Carney notes that these latter two retailers are headquartered in Arkansas and Minnesota, states whose senators are signatories to this bill.³⁸ Whether this is coincidental or not, the contours of which platforms are covered seem arbitrary and more about the current political zeitgeist than any economic rationale. The aforementioned ambiguities, moreover, would feed a frenzy of lobbying to convince policymakers that their clients and constituents are free from liability while competitors should be prosecuted.