1.1. Recent Regulatory and Legislative Scrutiny in the US and EU
The European Commission fined Google in 2017 for favoring its own products over competitors’ in the search results with Google Shopping. In the United States, as part of the 2020 “Investigation of Competition in Digital Markets,” lawmakers alleged that several large technology companies were engaging in various forms of self-preferencing and called on Congress to “consider establishing nondiscrimination rules to ensure fair competition and to promote innovation online.”
Lawmakers in both the United States and Europe have proposed legislation aimed at restricting the behavior of vertically integrated digital platforms. For instance, in March 2022, the European Union reached a provisional political agreement to approve the Digital Markets Act (“DMA”) to regulate so called “gatekeepers” (defined as large online platforms like Apple, Google, Facebook, and Amazon) and prevent, among other things, supposed self-preferencing practices such as exclusive use of platform-generated data and ranking of a platform’s own products higher than those of its competitors.,
In the United States, lawmakers have introduced, among other legislative proposals, a bipartisan bill called the American Innovation and Choice Online Act (“AICOA”). In its current form, the AICOA could make it unlawful for large online platforms to condition access to the platform on the purchase of their other services, to use non-public data they generated or collected through the platform to compete with third parties, or to rank their own products and services higher than competitors’. The bill has been met with significant criticism but has cleared the Judiciary Committees of both chambers and continues to make its way through the legislative process with varying degrees of bipartisan support. Further, the Department of Justice recently signaled its support for the proposed legislation and its concerns about self-preferencing practices in the digital economy.
1.2. Framing the Issue of Self-Preferencing, Fairness, and Pricing Practices
To a certain extent, the increased focus on alleged self-preferencing seems to reflect an increasingly critical view of large tech companies by regulators and lawmakers rather than an inherent aversion to the idea of self-preferencing practices more generally. But it is far from established that these challenged practices, without more, amount to exclusionary conduct under established antitrust law.
First, there is nothing new in companies, including platforms and other types of intermediaries, operating in “dual mode,” potentially favoring their own products over those of competitors. For example, supermarkets have a long history of selling their own private label products, which they can decide to place and price aggressively. Similarly, US health insurers have increasingly vertically integrated with healthcare providers as illustrated by Optum (a subsidiary of UnitedHealth Group) purchasing the DaVita Medical Group, and its offering of new plans that steer their members to their own doctors.
Second, while a common regulatory concern is that self-preferencing can create an uneven, or “unfair” playing field among competitors, some antitrust practitioners have questioned whether fairness itself is a relevant criterion. Even if a platform favored its products over those of competitors, therefore potentially harming those competitors, would it actually result in harm to competition through, for example, overall higher prices, lower output, or reduced innovation or product quality in the downstream market? Indeed, it is well established—at least under US federal law—that challenged conduct must “harm the competitive process and thereby harm consumers” and “harm to one or more competitors will not suffice.” “Even an act of pure malice by one business competitor against another does not, without more, state a claim under the federal antitrust laws,” explained the Supreme Court in Brooke Group, rejecting similar efforts to use antitrust law to enforce principles of “fairness.”
2. The Role of Commission Rates: Raising Rivals’ Costs or Benefits from Vertical Integration?
When platforms operate in a dual mode, their competitors may incur costs that they do not. Digital platforms that connect sellers and buyers typically charge a fee on transactions they facilitate, often expressed as a given percentage of the goods and services exchanged. However, when a platform also competes in a downstream market, it may decide whether, and how much, to charge for an internal commission fee.
Some have argued that charging competitors a fee that the platform does not pay on its own transactions is self-preferencing and could be anticompetitive. For instance, in 2019, Spotify complained to the European Commission about the App Store commission that Apple charges developers, but not itself.
2.1. Economics: Raising Rivals’ Cost and Eliminating Double Marginalization
Under the commission rate theory raised by Spotify and others, a platform would use high commission rates to drive the price of third-party sellers up, making the platform’ own products and services relatively cheaper and more competitive. Such concerns fall under the well-known antitrust nomenclature of “raising rivals’ costs.” Competitive concerns over raising rivals’ cost can arise, for instance, in the context of vertical mergers, when the vertical integration of an upstream and downstream player increases the incentive for the merged entity to increase upstream prices (i.e., input prices) to downstream competitors. By doing so, the merged entity loses some upstream profits but forces competitors to increase downstream prices, which makes the merged entity more competitive in the downstream market. It can increase downstream prices, sell more, or both and increase its profitability. In some cases, raising rivals’ costs could even lead a competitor to be forced out of the market entirely.
However, compared to traditional one-sided businesses where downstream firms buy inputs from upstream sellers, a platform may have more limited incentives to raise rivals’ costs. Digital marketplaces create value by facilitating the interactions between sellers and buyers. They encourage interactions and benefit when more transactions are facilitated by the platform. Additionally, they are typically characterized by strong indirect network effects, meaning that more or higher-quality sellers increase the attractiveness of the marketplace to buyers, and more or more profligate buyers increase the attractiveness of the marketplace to sellers.
Therefore, making sellers on its platform less attractive to buyers goes directly against the fundamental business model of a platform and its general objective of encouraging interactions on the platform. This reduces the value of the platform itself and risks generating a negative feedback loop where fewer or less attractive sellers on the platform make the platform less attractive to users, which, in turn make the platform less attractive to sellers, and so on. Given that the platform’s goal is to attract users on both sides of the market and leverage indirect network effects, self-preferencing behavior that raises rivals’ costs goes directly against a platform’s business model and can endanger the profitability of the platform.
In addition, economists have long recognized that vertical integration creates the incentives for firms to lower their price to final consumers, through the “elimination of double marginalization.” The elimination of double marginalization has often been identified by economists, regulators, and courts alike as a pro-competitive effect of vertical integration. From that perspective, platforms offering their own products to consumers, possibly at lower prices, may eliminate double marginalization and benefit consumers through lower prices.
Is it unfair for a platform not to charge itself fees, as argued by Spotify, or is it a natural and common result of vertical integration that benefits consumers through the introduction of an additional option at possibly a lower price?
As part of this debate, it may be informative to consider whether a platform charges different fees to sellers on its platform if it competes with them. Under a theory of raising rivals’ cost, a platform would be expected to increase fees only to sellers it competes with. Uniform fees across sellers regardless of their competitive positioning would therefore be inconsistent with claims of unfairness or anticompetitive conduct.
2.2. Proposed Remedies
Although the discussion above raises significant doubt about the anticompetitive nature of so-called self-preferencing behavior related to commission rates, even ignoring that, finding an adequate remedy would be far from trivial. Neutrality remedies would require platforms to charge their own subsidiaries the same commission rate they charge third parties. However, implementing and enforcing such a policy is not obvious. For example, requiring platforms to create internal walls and charge internal rates equal to their downstream subsidiary would still leave open the potential for other financial interactions between the upstream entity and its downstream subsidiary that may undo at least in part the rate setting. The rate setting may also, if it can effectively be implemented, directly increase prices paid by consumers on the platform’s product.
Another alternative that has been suggested is to prevent platforms from selling downstream altogether, as US Senator Elizabeth Warren proposed in 2020. However, recent economic research suggests that banning “dual mode” would harm consumers. Three recent economics papers, Hagiu et al. (2020), Hao Lee, Musolff (2021), and Kang and Muir (2021), all find that such a drastic policy could result in lower consumer surplus because of its deleterious effect on choice and lower search costs—which are two raisons d’être of platforms in the first place.
3. Using Marketplace Data to Price Products
Platforms that are intermediaries between small sellers and buyers may collect data about interactions on their platforms. However, if the platform also competes in a downstream market, is the collection and use of such data by the platform just another form of anticompetitive self-preferencing?
Such allegations have been made against Amazon, which is both a marketplace through which sellers sell their products to consumers and a retailer that sells its own products. In the United States, some have accused Amazon of using data originating from interactions between buyers and third-party sellers for private-label decision making. These generalized accusations broadly target how Amazon prices and markets its products, and allegedly uses data to research successful products it may want to compete against. The European Commission, which is currently investigating Amazon’s data collection practices and trying to determine whether it constitutes an abuse of its dominant position as a retail platform, has stated in a preliminary fact-finding phase that “Amazon appears to use competitively sensitive information – about marketplace sellers, their products and transactions on the marketplace.”
While the allegations of self-preferencing regarding the use of data are wide-ranging, a subset of them relates to how platforms allegedly use the data they collect on third-party vendors to price their own products. For instance, in her paper titled “Amazon’s Antitrust Paradox,” Lina Khan, Chair of the Federal Trade Commission, alluded to press reporting on how “Amazon uses sales data from outside merchants to make purchasing decisions in order to undercut them on price,” concluding, “[t[his dual role also enables a platform to exploit information collected on companies using its services to undermine them as competitors.”
On the other hand, other antitrust practitioners have noted that the at-issue behaviors are not new, are typical of many intermediaries, and have been demonstrated to benefit consumers. For instance, supermarkets have long developed their private labels, with products that imitate competitors’ higher end products and are often offered at lower prices. Due to their position as intermediaries, like digital marketplaces, supermarkets have access to third-party sales data and they choose how they want to display products in their store; unlike digital marketplaces, they also typically decide how to price them. When supermarkets use their sales data to shape their product offerings, consumers often benefit from lower prices and an increase in the variety of products available. Similarly, a digital platform’s use of data to develop a new product or cheaper imitation product increases choice for consumers.
Private label products are successful with consumers: In 2020, private labels were estimated to account for around 18% of product sales in US supermarkets. Meanwhile, product innovation does not seem to have been limited. As a result, some antitrust practitioners, such as Aurelien Portuese, Director, Antitrust and Innovation Policy at the Information Technology and Innovation Foundation, have argued that it would be inconsistent to consider offline forms of self-preferencing procompetitive and consider online forms anticompetitive.
Other antitrust practitioners have argued that the scale of data gathered by digital platforms, coupled with the level of sophistication of these digital marketplaces, would allow them to use data differently from brick-and-mortar stores, such that similar self-preferencing practices could become anticompetitive in the case of digital platforms. For example, Lina Khan has argued that “Amazon … has expanded into a suite of additional businesses and amassed significant troves of data on users. This data enables it both to extend its tug over customers through highly tailored personal shopping experiences, and, potentially, to institute forms of price discrimination.” But it is not clear what data digital platforms are actually using and how.
Lambrecht & Tucker (2015) argue that “big data is not inimitable or rare, that substitutes exist, and that by itself big data is unlikely to be valuable,” and that therefore big data is not itself a source of competitive advantage. Specifically: Data are not rare, but are available cheaply; and data are not inherently valuable; they require strong management and operational capabilities to make it so. Data are unstructured and it is difficult (if not impossible) to establish causal relationships with data alone. Firms must build algorithms and business processes to take advantage of these large, unstructured data. It is these algorithms and processes that provide value to the firm, not the data itself.
3.1. Proposed Remedies
In the context of concerns associated with data use for self-preferencing, two notable types of remedies have been used or envisioned by regulators to reduce the asymmetrical access to data (in a more general context than pricing only). The first would require the platform not to use certain data at all. For example, in the Google/Fitbit acquisition (a case that also involves the use of health and fitness data, although in the context of ad targeting rather than pricing), the European Commission approved the merger, but only on the condition that Google not use data from Fitbit to target search ads, requiring Google to “silo” the Fitbit data. In Europe, under the DMA, large platforms would also be prohibited from using data about their competitors generated on their marketplaces to compete. Similarly, in the United States, the current text of the AICOA could make it unlawful for a platform to use non-public data about its business users to compete with those business users.
The second type of proposed remedy would require data to be made available more broadly. For example, the DMA contains a data-sharing obligation that could require that large digital firms share customer data collected through their platforms with third-party sellers for free upon request.
Whether and how these policies would lower prices or increase choice, hence benefitting consumers, is an open question. The premise of such policies is also unclear. Do they imply that data impart a monopoly or that they constitute a digital essential facility at some stage in the distribution or competitive process? Before putting a remedy in place, a fact-based and economically sound analysis should consider whether economics conditions are such that these policies are likely to actually promote lower prices, more choices, and greater innovation for consumers (e.g., are data the core of the issue?) or whether the underlying costs to implement them (e.g., technologically) and distortions they would create would end up having the unintended consequence of interfering with innovation.
4. Conclusion
To conclude, allegations of self-preferencing in digital settings are likely to remain a focus for lawmakers, regulators, and courts, as well as an object of controversy. As in many areas of competition, a careful statement of the issue is an important first step. This should include a thorough understanding of whether and how the digital setting informs the potential impact of supposed self-preferencing practices and, as always, a focus on harm to competition rather than competitors. Furthermore, a careful analysis of the short-term and long-term consequences, as well as the technical complexities of implementation and privacy concerns, of any proposed remedy should precede regulatory or legal intervention. Specifically, the raison d'être of digital platforms is to provide increased choice through seamless interactions and to reduce transaction costs. It seems therefore prudent to examine how proposed remedies would affect choice, transaction costs, and ultimately prices.
