As Greg Werden has noted, the process of defining the relevant market in an antitrust case doesn’t just finger which part of the economy is allegedly affected by the challenged conduct, but it also “identifies the competitive process alleged to be harmed.” Unsurprisingly, plaintiffs in such proceedings (most commonly, antitrust enforcers) often seek to set exceedingly narrow parameters for relevant markets in order to bolster their case. In the extreme, these artificially constrained definitions sketch what can only be called “gerrymandered” markets—obscuring rather than illuminating the competitive processes at issue.
This unfortunate tendency is exemplified in the Federal Trade Commission’s (FTC) recent complaint against Amazon, which describes two relevant markets in which anticompetitive harm has allegedly occurred: (1) the “online superstore market” and (2) the “online marketplace services market.” Because both markets are exceedingly narrow, they grossly inflate Amazon’s apparent market share and minimize the true extent of competition. Moreover, by lumping together wildly different products and wildly different sellers into single “cluster markets,” the FTC misapprehends the nature of competition relating to the challenged conduct.
What follows is a distillation of my just-published ICLE Issue Brief analyzing these market definition problems in the FTC’s Amazon case, “Gerrymandered Market Definitions in FTC v. Amazon,” available here.
Defining the ‘Online Superstore Market’
According to the complaint, the online-superstore market is limited to online stores that have an “extensive breadth and depth” of products. This means online stores that carry virtually all categories of products, from sporting goods to consumer electronics, and that also have an extensive variety of brands within each category. In practice, this definition excludes leading brands’ private channels, such as Nike’s online store, as well as online stores that focus on a particular category of goods, such as Wayfair’s focus on furniture. It also excludes brick-and-mortar stores, which still account for the vast majority of retail transactions. Firms with significant online and brick-and-mortar sales might count, but only their online sales would be considered part of the market.
To see how this market definition tilts the balance, consider the FTC’s allegation that Amazon dominates the online-superstore market, with approximately 82% market share. To reach that figure, the FTC determined that Amazon’s offerings constitute 82% of the gross merchandise value (GMV) of U.S. online sales in a market that excludes perishables and includes only those goods sold online by Amazon, Walmart, Target, or eBay. While Amazon’s share of overall online retail is, indeed, substantial, it’s actually less than half the figure in the market the FTC has gerrymandered, at 37.6%. Indeed, if one were to count total retail sales, Walmart actually leads Amazon, not vice versa. And while e-commerce may be substantial and growing, it represents only about 15% of U.S. retail.
Not only does defining a relevant market with reference to a single retailer’s particular product offerings not “identify the competitive process alleged to be harmed,” as Werden put it, but it doesn’t actually identify a product at all. Instead, it ends up excluding a host of competing sellers that offer economic substitutes for the products that consumers actually buy. Consumers could prevent a hypothetical monopolist within the “online superstore market” from raising prices by switching to other online channels that don’t qualify as a “superstore,” as defined by the FTC. They could, for instance, switch to a brick-and-mortar retailer. How many might switch, and the extent to which that constrains the monopolist’s pricing, are empirical questions, but there is no question that some consumers might switch: retail multi-homing is common.
Further, despite its repeated emphasis on the “depth and breadth” offered by online superstores, the FTC’s complaint ignores e-commerce aggregators, which allow consumers to search products and pricing across an incredible variety of retailers. Google Shopping, the most notable example, is curiously absent from the complaint. Google Shopping and other aggregators allow consumers to browse extensive results in one place for almost any product, including across all categories and across many brands. Indeed, surveys find that roughly half of all shoppers said they use Google both to discover new items and to research their planned online purchases. And Google Shopping is not alone, as buying through social media has boomed. Instagram, for example, has become an online-shopping juggernaut.
Defining the ‘Online Marketplace Services Market’
The FTC’s complaint limits its definition of the online-marketplace-services market to those online platforms that provide access to a “significant base of shoppers”; 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. For instance, the complaint distinguishes online marketplaces from online retailers where the seller functions as a vendor 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.
This implies that current Amazon sellers can’t reach consumers through mechanisms that don’t incorporate all of these specific functions, even though consumers regularly use multiple services and third-party sites that accomplish the same thing, including Google Shopping, Shopify, and Instagram. Moreover, it implies that these myriad alternative channels do not constrain how Amazon prices its services.
The complaint alleges that neither operating as a vendor nor utilizing SaaS is “reasonably interchangeable” with online marketplace services—the key language from the U.S. Supreme Court’s 1962 decision in Brown Shoe Co. v. United States. But merely saying so doesn’t make it true. Differentiated competition exists in service markets, just as it does in product markets. Superficial differences among services do not establish that they are not competitors.
For example, if a hypothetical monopolist online marketplace increased prices or decreased quality for selling a product, why would, say, 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 did exactly this in 2019, when it stopped selling products to Amazon because it was dissatisfied with Amazon’s efforts to limit counterfeit products. Instead, Nike instead 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 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. It’s remarkable that it is completely absent from the FTC’s market definition.
For its part, Instagram allows sellers to use Meta’s “Checkout on Instagram” service to process orders directly on Instagram, and to use logistics services like Shopify or ShipBob to manage their supply chains and fulfill sales, replicating the core functionality of a vertically integrated storefront like Amazon.
Indeed, one thing that Amazon, SaaS providers, and other similar platforms have in common is that they invest significantly 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 disparate sellers all 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 Amazon’s “marketplace” or “superstore” aspects simultaneously facilitate that access while also impeding it by congesting it with other sellers and products. In this sense, a specialized outlet may, in fact, offer the optimal selling 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). There is little to 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 Problem with Cluster Markets
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. Indeed, despite their widespread use, the economic logic of cluster markets is, at best, poorly established.
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 entirely to group batteries and bedroom furniture into a single “market,” just because Amazon happens to facilitate sales of both.
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.”
A second basis for clustering is the “transactional-complements” theory, relabeled by the 6th Circuit as the “‘package-deal’ theory.” 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.”
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. But while the Court suggested some reasons why, in its own telling, “some commercial banking products or services” may be insulated from competition, 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.
Perhaps the best example of a 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. 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. But it is notable 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.” Moreover, 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.
While the Amazon case is only at the complaint stage, there is no evidence in the complaint that the FTC even considered the possibility that different products and different sellers would need to be considered separately. 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. Instead, the complaint appears to assume that, if Amazon offers a grouping of products, or offers services to different types of sellers, this constitutes an economically rigorous “relevant market.” (Spoiler alert: It does not.)
The implication of all this is that it seems highly dubious that furniture and batteries face sufficiently similar 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.
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 that no outcome can properly be determined.
Out-of-Market Effects
The relevant markets alleged in the FTC’s complaint draw a distinct line between the seller and buyer sides of Amazon’s platform, thereby implicitly rejecting cross-market effects as 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.
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.
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; 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.
The bifurcated market approach that the FTC appears to be pursuing here conflicts with the Supreme Court’s holding in Ohio v. American Express. 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).
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. (For a discussion of this problem in the context of mergers (though with relevance for Section 2 cases), see my TOTM post with my colleagues Dan Gilman and Brian Albrecht).
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 miss 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. 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. 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. Such an approach would make it much easier for the FTC to win its case, but would do nothing to ensure an accurate result.
Conclusion
Ultimately, what determines the proper scope of relevant markets is 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.
