This is the third blog post in a series on regulating digital platforms. A version of this first appeared on the blog of my employer, Public Knowledge.
In my last blog post, I explained my working definition for what constitutes a “digital platform.” Today, I focus on another concept that gets thrown around a lot: “dominant.” While many regulations promoting consumer protection and competition apply throughout a sector, some economic regulations apply to “dominant” firms or firms with “market power.” Behavior that is harmless, or potentially even positive when done by smaller companies or in a more competitive marketplace, can be anticompetitive or harmful to consumers when done by dominant firms — regardless of the firm’s actual intent.
For reasons discussed in my previous blog posts, defining what constitutes “dominant” (or even identifying a single market in which to make such a determination), presents many challenges using the traditional tools of analysis favored by antitrust enforcers and regulators. I therefore propose that we use the cost of exclusion (“COE,” because nothing in policy is taken seriously unless it has its own acronym) as the means of determining when we need to apply regulation to “dominant” firms. That is to say, the greater the cost to individuals and firms (whether as consumers or producers or any of the other roles they may play simultaneously on digital platforms), the greater the need for regulations to protect platform users from harm. If a firm is “too big to lose access to,” then we should treat that firm as dominant.
When Do We Distinguish Between General Sector Regulation vs. Regulation of Dominant Firms?
We must recognize at the outset that — contrary to the insistence of some of my Libertarian opposite numbers — we do not regulate solely to address market failure or market power. Rather, when a business becomes central to our economic or social well being, it requires some level of oversight to ensure that all members of the public are treated fairly and that unexpected or unanticipated problems don’t cause significant disruptions in commerce. Additionally, we may have other public policy goals, such as public safety, that require regulation.
But regulations are not binary. It’s not a question of simply on/off. Some regulations always need to apply. For example, we apply the health code to all restaurants, no matter how small. And yes, that does indeed drive up the cost of doing business, but since you are just as sick if you get food poisoning in a family-owned diner as you would be getting it at Chipotle, we as a society have decided it’s worth it. Likewise, if I am going to leave my money in a bank, or borrow money to buy a car or a house, it both facilitates commerce and protects consumers to have some basic rules in place governing what sort of institutions can do that and what default rules govern these transactions.
On the other hand, there are also a lot of rules that depend on things like size or market power. They are simply unnecessary, or even counterproductive, when applied to non-dominant firms. If a local bank goes under, or otherwise gets into trouble, that’s obviously a problem for the local community. But if a giant bank goes under, it can threaten the health of the economy as a whole. As my colleague John Bergmayer explained at greater length in his white paper on digital platforms and due process, we have a long history of applying special obligations to dominant firms.
Differentiating regulation between dominant and non-dominant firms particularly makes sense in industries that have a natural tendency towards concentration. From a consumer protection standpoint, dealings with a dominant firm are likely to be far more regular and consequential than with a non-dominant rival. On the competition side, we need regulation to counteract the specific economic forces in the sector that create the natural tendency towards concentration and the endurance of market power once achieved.
Defining “Dominant” Is Generally Tricky, and It’s Especially Difficult in the World of Digital Platforms.
As Bergmayer also pointed out in his white paper, there is no generally accepted definition of what “dominance” means. Just as regulators struggled with “how big is too big to fail” after the financial crisis 10 years ago, regulators and antitrust enforcers have repeatedly struggled with the question of what makes a firm “dominant” or “non-dominant.” In the past, regulators and antitrust enforcers have looked to things like “market share,” or “incumbency,” or being a “critical buyer,” or some other indicia of the ability to exert control over the behavior of others contrary to how we might otherwise expect them to behave in a competitive market. Sometimes, as with too big to fail, regulators look to balance the cost of regulation against the potential risk to the sector or economy as a whole.
Even if we could settle on a specific metric, what constitutes dominant is subject to considerable debate. At one time, antitrust law established a presumption that any entity with 30 percent market share would be considered “dominant.” This presumption, called the Philadelphia National Bank presumption, for the antitrust case that established it, is inconsistently applied in the U.S., but But 30 percent market share is still generally used in Europe for creating a rebuttable presumption of market power. The Federal Communications Commission declared AT&T a non-dominant long-distance carrier despite a 60 percent market share. On the other hand, in FTC v. Toys R Us, the court found that TRU had sufficient market power to support an antitrust enforcement action at 20 percent market share. What actual market share constitutes dominance varies depending on multiple factors including, but not limited to, the nature of the market, the purpose of the analysis, elasticity of demand, ease of entry, and possibly phases of the moon.
Furthermore, for reasons discussed at considerable length in my first and second previous blog posts, traditional economic measures of dominance and market power are particularly difficult to apply to digital platforms. It is a characteristic of these firms that they achieve dominance by a breadth and depth of services that makes traditional market definition and identification of actual or potential competitors challenging. In addition, because digital platforms have varied and novel business models, traditional tools of economic analysis have struggled to identify an appropriate approach to digital platforms, let alone create consensus around how to define market power or dominance among online platforms.
One of the advantages of sector-specific regulation over antitrust, however, is that we do not need to limit our concerns to firms considered dominant under classic antitrust analysis. Our touchstone lies in our overarching concern to promote the public interest with regard to the specific facts of the industry. As identified in my previous blog posts, our primary concerns are protecting consumers from harm, promoting the production and dissemination of news and differing perspectives in the marketplace of ideas, facilitating civic engagement and access to services, promoting competition, and public safety. These goals are consistent with both the fundamental values that have driven communications policy throughout U.S. history and with the public interest goals of the last century in regulating electronic media (see 47 U.S.C. § 257(b).
The Cost of Exclusion From the Platform Is the Appropriate Metric.
One of the elements of network effects is that the network becomes more valuable to everyone on it when more people use it. The inverse is equally true. The larger the network, the greater the cost of exclusion from the network.
Consider as an example the traditional evolution of unregulated interconnection regimes from the telecom and internet transit worlds. Initially, no network is dominant, and so carriers have incentive to exchange traffic for free. Everyone needs everyone else and derives roughly equal value from interconnection. As some firms grow faster than others, the larger networks are much more valuable. Smaller carriers suffer much more from the inability to interconnect with larger carriers than larger carriers suffer from the inability to interconnect with smaller carriers. Larger carriers are therefore able to demand payment from smaller networks for reaching the customers on their larger networks. If the cycle continues and the size disparity increases, it becomes increasingly easier for the larger network to offer value to customers without the smaller network, and harder for the smaller network to offer value to customers without the larger network. In an extreme case, such as AT&T’s control over the “long lines” (national long-distance lines) at the beginning of the 20th Century, this network dynamic can create a monopoly.
But one does not need extreme cases to see how the cost of exclusion from the platform can allow a provider to drive up prices on either side of the platform, or both simultaneously. Consider credit cards. As merchants testified in the Department of Justice antitrust action against American Express, merchants felt they would suffer significant losses if they did not accede to Amex’s demands – even though Amex has an approximate 25 percent market share of credit card transactions in the United States. (The Supreme Court found for Amex on the theory that Amex’s contractual demands benefited consumers (albeit at the cost of merchants), but this should not obscure the larger point that the district court found that merchants felt they would suffer significantly from being excluded from the Amex network.)
COE Is Extremely Flexible and Focuses on the Central Reason Why We Care About Dominance.
An advantage of using COE is that it encompasses a wide range of potential costs and potential actors, while avoiding the arbitrary definitions that have plagued traditional efforts to determine market dominance. For example, it is clear that COE includes the loss of a substantial market for producers of goods and services, or loss of an important distribution network. At the same time, however, it takes into account the loss to consumers from being excluded from a specific platform. For example, whether or not we consider Twitter “dominant” in a traditional economic sense, it is clear that a business excluded from Twitter experiences some cost from its inability to communicate with Twitter subscribers. These costs include more than those associated with traditional advertising or direct sales: Companies use Twitter to respond to real-time events such as a blackout during the Superbowl or a Tweet from a celebrity, and companies monitor social media to address concerns and respond to criticism. These benefits won’t necessarily make or break a business, but loss of access to the platform would certainly carry the significant cost of losing a valuable channel of communication with the public.
We can equally apply this analysis to Twitter subscribers. In a case involving President Trump blocking critics from his Twitter feed, the district court observed that blocking the individuals in question deprived them of the ability to interact directly with the President’s statements, denying them the ability to engage in important and timely political discourse. Greg Norcie and L. Jean Camp have written an analysis examining the costs of abstaining from social media generally. As they demonstrate, exclusion from social media platforms can have significant costs to the individual that traditional metrics for measuring dominance do not address.
As an additional benefit, using COE directly addresses the reason we want to distinguish dominant platforms from non-dominant platforms in this context. Where the cost of exclusion is small, we are unlikely to have any particular concern about the practices of the platform distinct from whatever general concerns we may have about platforms more broadly.
It is important to note that COE does not tell what regulation to use, but rather what to regulate. Once COE shows us that a firm is dominant, that may indicate a need for some kind of action that only addresses this dominance indirectly. For example, if we determine that a platform such as Google is dominant and that the key to that dominance is high market share search, the remedy might involve actions to stimulate competition rather than directly regulating how Google manages its search engine. By contrast, if the primary harm in being excluded from Twitter is the more limited harm of losing one of several important conduits of reaching customers, the necessary regulation may be limited to an explanation and right to challenge arbitrary exclusion as suggested by Bergmayer in “Even Under Kind Masters.” Again, context matters enormously.
Finally, while the COE is the measure of dominance, that does not mean that exclusion is the only harm. Rather, COE works as a measure of dominance in this context because if the platform imposes some new rule or cost on a take-it-or-leave-it basis, the platform participant must decide whether the cost of acceptance outweighs the cost of abandoning the platform.
Finally, I stress that simply because exclusion may impose costs — perhaps substantial costs — that does not mean that exclusion is never permissible. Indeed, in many cases it may be warranted. Even public utilities, services so essential that we consider it the responsibility of government to make them accessible to everyone, have circumstances when they may terminate service. For example, although public utilities generally must provide customers with significant grace periods for late payments, and may have lengthy procedures to prevent consumers from being cut off, a utility may ultimately refuse to serve a customer who does not pay. The telephone network is a common carrier network, but it may refuse to allow a customer to connect a device that will do damage to the network.
Similarly, there may well be circumstances where dominant platforms can (or arguably even should) exclude certain kinds of speech or certain types of businesses or products. Again, the point of using COE to measure dominance is not to ensure that users of platforms never experience costs. The point of using COE as a proxy for dominance is to determine when the (potential) behavior of a digital platform potentially threatens the public interest. Determining what regulation, if any, is needed is an entirely separate exercise. Now that we have determined what sort of entities we are talking about, and the circumstances under which regulation may be appropriate, we are finally prepared to explore what about these platforms we may need to address to protect the public interest.
Stay tuned . . .