“Self-preferencing” refers to situations in which a firm favors, or “preferences,” its own products over those of rivals. Of course, it is literally “self preferencing” for a firm to sell nothing but its own product. For example, Farmer Jane might market her own cucumbers and refuse to sell those of her neighbors. But that type of self-preferencing has not provoked much concern outside of the law of refusal to deal, which applies only to monopolists. It may violate some proposed legislation. Rather, the idea is that harmful self-preferencing occurs when a firm sells one or more brands in addition to its own and gives its own brand favorable treatment, or else when it tries to steer customers of some primary product to its own brand of a secondary product or its own repair service.
Self-preferencing can come in many varieties, ranging from outright exclusion of competing alternatives to simple favored placement or promotion of the seller’s own version. Here are a few examples:
United States antitrust law is built on a common law tradition that every firm has a qualified right to choose its dealing partners and has no obligation to aid its competitors. The Supreme Court declared in 1919:
In 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.
That view accounts for the restrictive approach that antitrust law takes to unilateral dealing duties. The Supreme Court explicitly tied any duty to deal in a rivals’ products to a “purpose to create or maintain a monopoly.” Except for naked horizontal agreements, all limitations on self-preferencing that antitrust law currently recognizes incorporate a product-specific market power requirement.
Restrictions on self-preferencing have a surprisingly long history in American law, antedating the Sherman Act. Patent law is an important precursor. In Wilson v. Simpson (1850), the Supreme Court held that someone who purchased a patented wood planing machine could not be forced by a patent license to use the patentee’s own replacement blades. To this day, this doctrine of patent “exhaustion” protects the freedom of users of patented goods to select their own aftermarket or complementary goods unless they have contractually agreed to do otherwise.
In addition, patent law’s distinction between “reconstruction” and “repair” also protects the right of a patented good’s purchaser to install replacement parts provided by someone other than the patentee. For example, the Aro Manufacturing decision held that the owner of a traditional “ragtop” convertible could replace the entire fabric top when it wore out. That was a permissible “repair” of a patented product, rather than an infringing “reconstruction.”
Patent law was thus an important, although limited source of the “right to repair.” As the Court observed in 1863, someone who legally acquires a patented good may “use it until it is worn out, or he may repair or improve upon it as he pleases.” By contrast, third-party makers of aftermarket automobile parts have not had much luck preventing automobile manufacturers from using design patents to prevent the use of lookalike designs for such things as bumpers and mirrors. Rather, they have to provide a dissimilar-looking part by inventing around the patented design, and this has provoked consumer resistance.
The Supreme Court briefly relaxed the patent exhaustion rule in Henry v. A.B. Dick Co. (1912), sustaining an office equipment manufacturer’s suit against someone who purchased consumable supplies from a third party. Congress immediately responded by amending the antitrust laws, rather than patent law, to include Section 3 of the Clayton Act.
That statute became antitrust law’s first explicit “self-preferencing” provision, making it unlawful to sell a good, “whether patented or unpatented,” on the condition that the user of the good not deal in the goods of a competitor. One important limitation was that the refusal must threaten either to “substantially lessen competition” or “tend to create a monopoly.” Early decisions applied that provision to prevent IBM Corp. from mandating that users of its computational machines use its own data cards, and Standard Oil from requiring its franchised but independently-owned gasoline stations to pump its own gasoline exclusively. Section 3 might require a seller to (1) permit users of its projector to select rivals’ films; (2) permit a food franchisee to select single-use products of its choosing rather than selling those that are supplied or approved by the principal firm; or (3) permit the owner of a device such as a photocopier to select its repair technicians and parts, rather than those insisted on by the seller.
The sellers of goods in these patent and antitrust cases preferred that purchasers of their products procure aftermarket parts or complementary products only from themselves. By contrast, users wanted a choice. Tying law, in particular, went through a period of rapid expansion in the mid-twentieth century, and then contraction and even some legislative pushback in the 1970s and after. Tying law also has additional technical requirements, including market power in the tying product, and a requirement that the tying and tied good be “separate products.”
Section 3 of the Clayton Act and the Sherman Act can additionally prohibit conditional discounts or rebates intended to encourage self-preferencing. For example, a seller who offers a discount on the condition that someone not deal in a competitor’s good would be covered. There is also some Sherman Act law of “quasi” exclusive dealing that occasionally reaches things such as market share discounts given to buyers who limit their purchases from rivals by a specified percentage. Alternatively, “slotting” allowances are discounts for favorable treatment given to retailers in exchange for preferred promotion, display, or shelf space. They are occasionally illegal under Section 2, or else under the Robinson-Patman Act.
Other than conditional discounts or rebates, tying and exclusive dealing rules do not usually reach mere preferential ordering of goods that falls short of prohibitions on dealing. There must be a “tie,” which is a coercive condition requiring purchase of the tied good. While the issue has not been fully adjudicated at this writing, default rules that tie only presumptively, but then permit the buyer to swap away from the seller’s product, are very likely not ties under existing law. They may sometimes violate Section 2 of the Sherman Act if they unreasonably create or maintain monopoly. There is some law indicating that even if a tying condition is not stated, it can be inferred from situations where a very high percentage of sales (say, 90%) are in fact tied. At this writing that rule has not yet been applied to defaults.
Even though consumers are free to opt away, being the default can be extremely valuable. For example, Alphabet reportedly pays Apple $12 billion annually for default search engine status on the iPhone. Nevertheless, an iPhone purchaser can easily and quickly download other search engines, and the iPhone Appstore even provides them at no cost. At this writing, a court has denied the defendant’s motion for summary judgment on the Government’s claim that various default provisions governing Google Search are unlawful under Section 2, and the case is in trial. One issue is whether a default amounts to the type of “foreclosure” that exclusive dealing is thought to require, particularly when the defendant has a dominant position in the affected market.
Another instance of Apple self-preferencing is its technological constraint limiting sales of applications (apps) for the iPhone to the phone’s own Appstore, where it charges a substantial commission. In Epic Games, the Ninth Circuit held that this limitation was permissible under Section 2 of the Sherman Act, mainly because the plaintiff failed to show sufficient market power. But it was found to violate California’s unfair competition law and the court issued an injunction that the 9th Circuit subsequently stayed pending a decision on a certiorari petition. Assuming the Supreme Court does not grant certiorari and reverse, questions remain about the scope of the injunction. In affirming, the Ninth Circuit noted that its scope was limited to Epic’s injuries, but also noted that “an injunction limited to Epic’s subsidiaries would fail to address the full harm caused by the anti-steering provision.” The injunction, which is entirely under state law, has no stated geographic limitation. While both Apple and Epic are headquartered in California, iPhones are used worldwide and many firms that would be affected by a broad injunction are located outside California.
Existing law does not ordinarily reach forms of self-preferencing that are limited to mere display or convenience. For example, a retailer might place a preferred brand at eye level or at the front of the store, while relegating others to less visible space. Netflix might display its internally produced video content in the top row of its website, while relegating third party offerings to a lower position, or an Amazon product search might rank its “Amazon Basics” house brand more highly than other brands. While these practices are literally a form of brand “preferencing,” they are ordinarily not reachable under current U.S. law, at least if they are unilateral.
The Robinson-Patman Act sometimes leads to liability for discriminatory promotional services. One important limitation, however, is that while the Robinson-Patman Act does not require market power, it does require discrimination between two independent purchasers. That is, the statute does not generally apply to “self” preferencing at all, but rather when a firm grants different terms to two independent dealers. Even if one of the resellers is an independent dealer while the other is a subsidiary, the statute does not apply. Indeed, one of the great ironies of the Robinson-Patman Act is that it very likely induced a great deal of vertical integration by firms wishing to avoid its “sale” requirement—not something its framers contemplated.
Most instances of product design that limit consumer repair access are permissible under current law, although the EU is moving toward greater oversight. One prominent exception under United States law is the C.R. Bard litigation, in which the Federal Circuit upheld a jury verdict against a firm who redesigned a biopsy “gun’s” connection point attaching it to single-use needles so that it could use only Bard’s needles. Prior to that a range of generic manufacturers had supplied the needles. The approved instruction entitled the jury to find that, not only did the revised gun work no better, but that Bard never intended for it to work better; it wished only to engineer incompatibility of third-party needles. Similar litigation involving Keurig, whose attempt to redesign its famous coffee maker to work only with its own coffee pods, ended in a settlement. In both C.R. Bard and Keurig, essential premises were substantial market power and a finding that the “improved” product was not seriously intended to be an improvement, but only to limit compatibility with a complementary product. In the previously referenced Epic Games case, by contrast, the court credited Apple’s defense that it needed to be able to control its Appstore in order to protect the integrity and security of its operating system.
The design problem also raises significant questions about the efficacy of ex ante regulation of product design and perhaps repair infrastructure. Consumers support technology evolution that makes products smaller and with more features. However, those same trends make repair more difficult and in many cases require that they be undertaken by trained technicians. As products are more technologically complex, responding effectively to possible government mandated right-to-repair obligations becomes more difficult as well. Some manufacturers may be able to respond only by reverting to older less appealing product designs. An EU Report indicates that it will also require greater investment in repair infrastructure, including investment in additional repair technicians. Finally, product manufacturers must have reasonable protection from the consequences of negligent third-party repairs.
The principal existing antitrust rules governing self-preferencing also include the Sherman Act’s highly restrictive law of unilateral refusal to deal, which may in a very few instances require a firm to deal in the goods of a rival. For example, under the narrow Aspen rule a market dominant ski company had a duty to sell a lift ticket combination that included its rival’s slopes as well as its own. But that was only because the plaintiff and defendant had previously done this by agreement, and then the defendant pulled out without a good explanation. In the Epic Games case, the Ninth Circuit dismissed for failure to show sufficiently durable market power the claim that Apple refused to share app sales on the iPhone. If durable power had been found, however, the case could very well have encompassed issues relevant to unilateral dealing obligations.
To generalize, while current United States antitrust law has many prohibitions on self-preferencing, they apply only when the firm in question has market power in the dominant good and competitive harm results from the refusal to give equal treatment. Mere harm to a competing seller is insufficient. Further, the refusal must be more than merely suggestive, although at this time default rules remain an open question. On the other side, however, these rules apply to the full range of activities and products covered by the antitrust laws, including both traditional and online sales.
In evaluating self-preferencing it is important to remember that a firm’s preferencing among the various products that it sells is less exclusionary than refusing to deal in them altogether. Further, unreasonably onerous restrictions on self-preferencing may induce a firm simply to drop a third party’s goods. That was the basis of Justice Douglas’ dissenting opinion in the Standard Oil exclusive dealing case. The majority condemned Standard’s rule that forbade its franchised dealers from selling gasoline supplied by third parties. Justice Douglas, one of the most pro-antitrust enforcement Justices ever on the Supreme Court, dissented. He presciently observed that requiring “split pump” stations would deprive Standard of its ability to control its gasoline sales. It would be forced to terminate its franchise agreements and sell gasoline directly. In some cases an effective rule against self-preferencing may also require a compulsory duty to deal, something that U.S. antitrust law has resisted. Otherwise firms might comply with a burdensome self-preferencing requirement by dropping rivals’ merchandise.
Every multibrand retailer continuously makes choices about which brands to handle and which should receive “preferred” attention in display or promotion. Retailers generally make these decisions by weighing such factors as customer evaluation, rate of product turnover, handling difficulties, supply availability and speed. For example, the United States has approximately 187 manufacturers or importers of ordinary kitchen cutting boards, in a wide variety of sizes, materials, and prices. Presumably, retailers will not stock and sell all of them. They select a few based on internal assessments of profitability. This forces the manufacturers to compete with one another for that retailer’s business. In the absence of evidence of monopoly or restraint of trade, the costs of antitrust control of a retailer’s choices seem to far outweigh any advantages. This is true even though some manufacturers will be excluded or will not get the favorable treatment that they want.
On the other hand, self-preferencing by a dominant firm in a dominant product, such as Google Search, might be anticompetitive to the extent that it unreasonably excludes rivals from a market. But the all-important ingredient is the presence of product-specific market dominance.
Legislation proposed at this writing is both broader and narrower than existing antitrust rules governing self-preferencing. The most important of these is the American Innovation and Choice Online Act (AICOA). First, unlike existing antitrust rules, its coverage is limited to digital, or online, firms. Second, covered firms, or “gatekeepers,” are identified on the basis of overall size. Third, the proposed rules generally do not require an agreement; they can be triggered by purely unilateral conduct. Fourth, unlike the antitrust provisions the new rules do not generally require proof of market power or competitive harm. Fifth, the new self-preferencing rules reach not only absolute exclusion of rivals’ goods, but also apply to weaker forms of preferencing, including favorable treatment or default rules. Finally, in its present form the AICOA is not an antitrust law. It can be enforced by the federal agencies and state attorneys general, but not by private parties.
The AICOA was initially proposed and debated in the 117th Session of Congress and failed to pass. At this writing, it has been resubmitted to the 118th Session. While the provision has bipartisan support, sponsored by Senators Amy Klobuchar (Dem., Minn.) and Charles Grassley (Rep., Iowa), it has also faced stiff resistance.
The following discussion considers the most important features of the AICOA’s approach to self-preferencing and how they deviate from current antitrust law.
The proposed American Innovation and Choice Online Act (AICOA) applies only to large digital platforms, and not to traditional (“offline”) businesses. For example, while Amazon and Walmart are retailers of roughly equal size, Amazon is covered but not Walmart, whose online presence is much smaller. Both engage in self-preferencing behavior.
Singling out online firms for harsher treatment is inconsistent with good antitrust enforcement policy. Suitable targets for antitrust scrutiny are concentrated markets exhibiting stagnant growth, a propensity toward collusion, and lack of new entry or innovation. The large online firms identified in the AICOA fit none of these criteria. For example, data from the U.S. Bureau of Economic Statistics estimates that the economic growth rate in digital markets is roughly four times higher than in old economy markets. At least as measured by patent grants, large digital firms are also highly innovative. Among the top 300 utility patent recipients in 2022, Apple, Alphabet, Amazon, and Microsoft ranked #14, #15, #16, and #18, while Meta (Facebook) was #42. Most of the top patent-receiving entities in 2022 were in Tech.
While comparative data are scarce, digital markets have not been shown to be more concentrated or resistant to competition than offline markets. In most sectors of the digital economy entry is fairly easy. Digital firms often compete in larger geographic markets than offline firms, and therefore the effective range of consumer choice is typically larger—or at least, it has never been shown to be smaller.
Further, the presence of digital sellers often reduces concentration. For example, a community may have had two or three stores selling small appliances. When Amazon or other online retailers start making sales in that community the number of competitors increases, perhaps significantly. Movie streaming and online media selling have considerably reduced retail concentration in that market. Online book selling has reduced concentration there. One could go on with examples, but the idea should be clear: online sellers have considerably reduced retail concentration by increasing the number of competitive options available to buyers.
There is little evidence of traditional collusion among large digital players. “Algorithmic” collusion, which substitutes mathematical models for individual announcements or decisions, has been a hot topic of scholarship about digital markets, and it may become more significant in the future. Anecdotally, however, even the largest digital firms appear to be competing aggressively with each other. There is no reason for thinking that price fixing or other anticompetitive horizontal restraints are more prevalent on digital markets. Witness, for example, the competitive race featuring mainly Microsoft, Alphabet, Meta, and Amazon over incorporation of AI technology, and the fierce battle emerging between Twitter and Threads, which is Meta’s new social networking competitor. At this writing, big tech seems to be a reasonably competitive landscape. There might be counterexamples, but they would have to be proven.
For the majority of products and services, the internet offers a wide range of choices, with lower search costs than offline searching. To be sure, there are a few market-dominating products. Google Search steadily commands more than 90% of the consumer search market. Amazon has about 67% of a market for ebooks, although ebooks constitute only 36% of the number of books that are sold, and only 19% if measured by revenue.
The extent to which online and offline products compete with one another varies immensely from one product to another. Digital search engines are so far superior to any traditional form of offline search that they are almost certainly a relevant market unto themselves. But in some cases the large consumer search engines may compete with more focused search engines used in specific markets. Aside from this, Google Search is very likely an antitrust monopoly. By contrast, online grocers have struggled. While Amazon and Walmart have roughly equal volumes of retail sales overall, Walmart has around 20% of the grocery market, and Amazon has roughly 1.3%. If one includes Amazon’s Whole Foods sales that number rises to 2.4%, but most of the Whole Foods sales are in physical stores, not online. Yet Amazon’s online grocery sales would be covered by the AICOA, because gatekeeper status under that act attaches to the overall size of the firm, not to the market share of any product. Walmart’s sales would not be covered. While it is as large as Amazon, it is not substantially an online firm.
In between are a host of products in which online and offline sellers compete with each other, but to various degrees. For example, Amazon and Walmart have roughly equal shares in the low twenties of small electric appliances. Amazon’s decisions about how to prioritize or rank brands could be challenged, while Walmart’s could not be.
As a result of these vast differences, there is no good substitute for product-specific inquiries into market power and competitive harm from any type of self-preferencing. Concededly, those inquiries are costly, but one reason for them is the extremely large number of false positives that more generalized tests will produce.
Looking from consumers’ perspective, online search costs are lower than in traditional markets. For example, someone unhappy with product choices in Walmart can get into her car and drive to a different store. Someone facing the same predicament on Amazon can escape with a mouse click, and generally to a great number of choices. Other things equal, as search costs are lower prices tend to move toward competitive levels.
To be sure, online commerce does invite problems, such as fraud, information security, or protection of vulnerable groups such as children. These are not antitrust problems under U.S. law, however, and are best addressed through more direct regulation of the particular behavior. Antitrust law is not a Swiss Army-knife directed at solving every problem, but only those that involve threats to competition.
Existing U.S. antitrust law imposes sharing requirements infrequently. When a firm is acting unilaterally, it must have substantial market power in the particular product at issue. The law of unilateral refusals to deal applies only to “monopolists,” which is a reference to a firm’s position in a particular product, not to the size of the overall firm. The law of tying and exclusive dealing require less market power than the law of monopolization, but they do requires some and it must be in the “tying” product. By contrast, the “gatekeeper” provisions in self-preferencing legislation such as AICOA apply to firms. Once a firm has been designated a gatekeeper, self-preferencing limitations can apply to any product it sells. There is no market power requirement.
For example, Microsoft very likely has a dominant position in its Windows operating system, but not in its search engine Bing (market share = 3%) or its internet browser Edge (market share = 5.4%). Amazon has a significant market position in the market for ebooks, but much less in the market for groceries. But self-preferencing duties under proposed statutes such as AICOA are attached to firms, who are selected on the basis of large overall size rather than the market share of any particular product. For any customer, however, the relevant power question is the range of realistic options available to that customer for that particular purchase. The overall size of the platform really does not matter that much.
Imposing self-preferencing duties on products whose sellers lack market power is particularly harmful if it is done selectively. The AICOA does not impose its rules on everyone, but only on the small number of firms designated as gatekeepers—very likely as few as five or six large internet platforms. A seller designated as a gatekeeper, but operating in a highly competitive market such as groceries, could be placed at an immense competitive disadvantage, perhaps enough to drive it out of that market altogether—and all for no obvious competitive benefit.
In contrast to AICOA, the European Digital Markets Act proceeds with Gatekeeper selection in two stages. First it selects Gatekeeper firms, which are Alphabet, Amazon, Apple, ByteDance, Meta, and Microsoft. Then for each of these it identifies “Core Platform Services,” which include some but not all of the products offered by each of these firms. For example, for Alphabet, the designated core services are Google Search, Maps, Play, Shopping, Android, Ads, Chrome, and YouTube, giving Alphabet the highest number. This is certainly an improvement over the AICOA approach of simply selecting covered platforms.
To see how a self-preferencing obligation might work under AICOA in its current formulation, consider Amazon.com, which sells a very large variety of products, most of them offered by multiple sellers. Some are pure commodities, but many are differentiated to some degree. When the customer does a search, Amazon selects a particular seller from the undifferentiated products that match the search criteria. It uses a ranking system that is based on price, customer satisfaction, delivery speed, and some other factors. The supplier who gets the highest rating goes into the “buy box” at the upper right corner of Amazon’s search results screen. By a large margin customers select the supplier that wins the buy box, even though they are free to choose an alternative. By contrast, if a customer does a more general product search in a differentiated market—say, for “bath towels”—then the search result will show up as a list. Because Amazon has a house brand—“Amazon Basics”—that will appear on the search result, although not necessarily at the top.
Because Amazon has no duty to deal in a rival’s products under current law, the way it ranks choices or its selection of a particular vendor for the “buy box” is up to Amazon, although a more complex practice such as tying might be unlawful. At least one court has sustained a complaint that Amazon unlawfully tied access to the Buy Box to an agreement to sign a most-favored-nation clause, which prevented the seller from offering better terms on a non-Amazon site. There has also been some non-antitrust litigation involving claims that third-party sellers have manipulated information in order to win the buy box.
For many products sold on Amazon, its selection of a vendor is simply a choice to deal through a third party rather than sell directly. That is, Amazon could take care of purchase, inventory, resale, credit, and shipping of a product itself, or it could engage a third party to do it in Amazon’s behalf. Nevertheless, because Amazon is likely to be a covered platform under the AICOA, its ranking choices could be subject to review.
While the AOCOA’s text is not final at this writing, the current proposals would make it unlawful for any firm to “advantage” the operator’s “own products, services, or lines of business over those of another business user.” Another makes it unlawful for Amazon to “discriminate” between multiple business users. An overlapping provision makes it unlawful to “exclude[] or disadvantage[] the products services, or lines of business of another business user relative to the covered platform operator’s own products…” The term “exclude” is both ambiguous and loaded. Does “exclude” imply a global duty to deal? If Amazon decides to sell its own house brand of scissors does that mean it must carry scissors for every one of the other 177 listed manufacturers?
Further, how do the AICOA’s nondiscrimination requirements apply to differentiated products? Amazon’s house brands are mainly manufactured products with at least modest differentiation from those of rivals. For example, an Amazon search for “electric toasters” (July 2023) reveals a large variety, with many different features. Prices on just the first three pages of the search result range from $13.49 to $399.95. Further, they are not listed in any readily comprehensible sequence. The first one on a search conducted in late July, 2023, was a $349 model with a touchscreen. It is made by Revcook, Inc., a small manufacturer that specializes in high end toasters. The fourth one was a simple 2-slice Proctor Silex model for $26.99. Procter-Silex is a brand of electric appliances owned by Hamilton Beach, which is a large firm. The first listing for an “Amazon Basics” toaster came in at $24.00 and is in the middle of the first page.
In some cases, a search is limited to a particular model of a particular brand, and there could be multiple vendors. In that case, there might be more objective criteria for ranking them, such as price, shipping speed, customer satisfaction, and so on. But what should be the ranking order for a $349 and a $13.49 toaster which differ significantly in features. And what if Amazon, responding to low sales, removes a particular model from its search results?
The text of the AICOA does provide an affirmative defense, with the proof burden on the defendant and a “clear and convincing” evidence standard that the conduct “would not result in harm to the competitive process. . . .” Why the statute uses this milquetoast definition of antitrust “harm to the competitive process” is unclear. It adds to the statute’s ambiguity because it says absolutely nothing. A “competitive process” standard does not identify violating conduct in terms of higher prices, lower output, or even something like equal access. In any event, however, satisfying this burden of proof by clear and convincing evidence is going to be a major challenge, particularly if the only evidence of discrimination is that the defendant did not include a particular supplier among the dozens that were available, or perhaps because it ranked a particular firm’s product fourth on the list instead of second.
There are pretty good reasons for leaving things like product selection and display to individual firms’ market selections. They involve a large number of choices about price, quality, performance, features, brand recognition, to name just a few. The choices generally reflect seller experience. For example, a firm will give prominent shelf space to items that sell quickly. But the important thing is that competition and the wish to maximize sales drives these choices, and it is very hard to believe that a government agency can make them better by substituting its own judgment. It forces firms to compete with one another for a firm’s attention.
The literature on self-preferencing is not all in agreement about its merits, but these conclusions appear to be tentatively justified:
