What Makes a Common Carrier, and What Doesn’t
What Makes a Common Carrier, and What Doesn’t
What Makes a Common Carrier, and What Doesn’t

    Get Involved Today

    The effects that digital platforms like Facebook and Twitter play in our society, and our democracy, have been under increasingly heavy scrutiny. Now, people are rightly looking at the platforms’ role in last week’s violent insurrection at the U.S. Capitol: both at how the attack was organized and instigated and at the platforms’ reactions to it.

    For many reasons, we need regulation of dominant digital platforms to advance public interest goals. Dominant platforms should not be able to shut their competitors out of the market, and we should expect platforms to be forthright and transparent about their content moderation practices.

    This post explains why some form of “must carry” requirements for social networks are not necessary — whether you call it common carrier or utility regulation, or a form of net neutrality, or simply some requirement that social networks not be “biased” against certain points of view. Tools like this do have their place in media, communications, and tech regulation. This is not it.

    People’s reactions to Twitter and Facebook deplatforming President Trump, and to Amazon, Apple, and Google effectively shutting down Parler, a social network that welcomes (or welcomed) right-wing conspiracy theories and organizing of violent attacks due to its purposely lax moderation policies, have been all over the place. It is true that many Trump supporters (as well as conservatives with no fondness for Trump) have discovered a newfound love for regulating private companies once those companies began making politically inconvenient decisions. But there are people across the political spectrum who think that platforms should default to leaving content up, as long as it’s lawful, or that banning users, or at least prominent ones like the President, is a slippery slope. Many people come to this view honestly and I respect it. But I think it’s wrong: Not only should the law not require platforms to carry all user content indifferently, platforms that hope to attract a wide base of users should not do so, since unmoderated speech platforms that only take down content they might face legal exposure for do not promote the goals of free expression.

    With this, it is fair to ask why services that are, or should be, required to be common carriers, like broadband, telephone, and SMS text messaging, should play by a different set of rules. Common carriers must serve all customers and carry all lawful traffic. It’s a stricter set of rules than some other regulated communications services are under — for example, cable TV providers are forbidden from some forms of commercial discrimination, but do not have to make their channel capacity available to all, as they would under a common carriage regime. Since the consequences of treating a service as a common carrier are so significant, it would be nice if there was a simple way to determine whether a service should be treated as one. But there isn’t one single characteristic that demands that a service be treated as a common carrier, or prevents it from being one. The policy question is simply whether common carrier regulation would be socially beneficial with respect to a certain service, or whether there are alternative models of regulation that might work better. Traits that are characteristic of common carriers and the values and goals common carrier treatment seek to advance should be considered when making this determination — but no single one is determinative. (For a more detailed discussion of common carriage than this blog post provides, see Harold Feld’s “The Case for the Digital Platform Act.”)

    This is in part because common carriage is a concept with deep historical roots, and as a historical matter, private businesses of all kinds in the common law world (shopkeepers, doctors) were viewed as “common callings” and subject to various restrictions, notably, non-discrimination obligations.[1] Most of those requirements dropped away. They remained for those areas where they were most socially valuable: mostly transportation and communications services and infrastructure, where goods, people, or communications are put in someone’s care, and inns (hotels and motels) that are responsible for housing people overnight. You can certainly identify common themes with all of these services: They are important to culture and commerce, and they give their owners an opportunity to profit from or take advantage of their position in ways that are seen as unfair.[2] So, a shipper might be required to follow posted rates that are based on criteria like size and weight, instead of charging more to someone who is shipping more valuable goods. Or, someone with the only place to stay in a town might be restrained from taking advantage of travellers in a pinch.

    However, no one ever sat down and planned this out. It was a historic and gradual process. Further, attempts to find other common features among common carriers — such as whether they are essential infrastructure or natural monopolies — don’t give us hard and fast rules. There’s always a counterexample. That doesn’t mean looking at these features isn’t instructive, of course. Additionally, sometimes it might make sense to have a version of a service that operates as a common carrier alongside unregulated options, so that there is a universal, nondiscriminatory alternative. (I’d argue that this is why SMS should be common carriage, but private messaging services need not be. The existence of a fallback, universal option means it doesn’t matter as much if not everyone uses WhatsApp or iMessage.) Nevertheless, it is worth walking through some of these additional characteristics and seeing how they apply to social networks and app stores.

    Natural Monopoly

    The most often cited justification for imposing common carriage is the “natural monopoly” argument. Natural monopolies exist where the cost structure of a market makes it impossible, or very difficult, to compete with an incumbent. If there is an established provider in a market with high fixed costs, but low marginal costs, it can be impossible for a potential new entrant to compete without public policy changes to the structure of the market. For example, an incumbent internet service provider (ISP) may have already paid for its network, so it only needs to charge each new customer the cost of extending service and some portion of overall operating costs. A new competitor has to pay to get its network up and running, but if it tries to pass those costs along to its customers, it can be easily undercut by the incumbent. Scenarios like this, or others where there are massive economies of scale, are likely to lead to natural monopolies.

    This dynamic does not apply to social networks that are free of monetary costs — but not necessarily of other costs — to end users. It does not apply to app stores, which enjoy their status due to technical design features of devices. Amazon’s cloud computing arm, Amazon Web Services (AWS) may be the closest example due to its economies of scale, but even AWS faces competition from well-capitalized companies like Microsoft, Google, and Oracle, and many large internet services operate their own data centers.

    Network Effects

    Social networks benefit greatly from network effects. Network effects occur when a network or platform becomes more valuable as more people use it. The classic example is the telephone system, which is valuable for all the people you can talk to, not its inherent technical characteristics. Some, but not all common carriers benefit from network effects. For example, all that matters is whether a package delivery company gets the package to its destination, not how many other people use that company.

    Network effects are ubiquitous in computing, and they can be indirect as well as direct. Software platforms (such as Windows and Android) often benefit from indirect network effects, since platforms with lots of users are likely to be better supported, and to have more applications, which in turn leads to more users.

    However, network effects do not have to lead to monopoly. To state the obvious, they accrue to networks, which do not have to be made up of just one company. Email benefits from network effects and so does the internet itself. A better policy option, that Public Knowledge has previously advocated for, is to increase competition through competition policy, regulatory requirements, and technical standards, so that users continue to benefit from broad networks but individual companies do not accrue excess power. My colleague Charlotte Slaiman has just written about this.

    Gatekeeper Power

    Social networks are gatekeepers, in that they control access to their user base, and as a practical matter using a social network might be important to actually being heard. But there are some differences between them and common carriers like telephone providers. An ISP can cut off a user from the entire internet, and even in markets where there might be more than one provider, users typically can only use one ISP at a time. This is not true today for social networks — there are many other ways of communicating and publishing online, and users don’t need to pick just one or the other. To the extent that we need to reduce their very real power, the better approach, as above, is to increase the number of competitors in the marketplace and the ease by which a user can choose an alternative.

    I’ve written extensively about app stores, which are definitely gatekeepers, and about how platforms at times abuse their control of them anticompetitively. At the same time, the best way to deal with them is to poke holes in their gatekeeper power, for example by platforms supporting or improving their support of software from outside platform app stores.

    Amazon’s AWS, again, may pose a more difficult question, if it turns out that it is not possible to operate an at-scale internet service without it in an economically viable way. But the existence of various alternatives would argue against AWS being considered a gatekeeper in cloud computing to the extent that common carrier regulation was necessary. It is true that getting suddenly cut off from a cloud computing provider and having to switch to a new one can be very disruptive for a business, just as having to switch ISPs with no notice would be disruptive for an end user, but it’s appropriate to have stronger safeguards for normal users.


    Finally, there is one last reason for common carrier rules, the cost to individuals if they cannot get service, or the inability of individuals to negotiate terms once they are locked in. For example, we long have regulated taxis, hotels, and delivery services as common carriers. Once the taxi driver has you in the cab, or you are staying in the hotel room, you have lost your ability to negotiate price or terms of service. Similarly, we have regulated power companies as utilities and telephone companies as common carriers because leaving someone without electric power, or refusing to connect the calls of their choice, makes it impossible for them to function, look for a job, or make doctor appointments.[3]

    There is therefore no legal machinery where you name a service, turn the crank, and spit out an answer as to how it should be regulated. We can identify lots of factors to consider — I’ve only named a few of them. But the facts of how each service operates, how economically and culturally important it is, and whether there may be superior alternatives to common carriage, have to be examined individually for each service. As a threshold matter, we need to understand how a service works, and who uses it, and for what. And we need to have the ability to cope with new situations. Simple rules like “Regulate natural monopolies as utilities or common carriers” are likely to be both under- and over-inclusive.

    Thus the fundamental question to ask is simply whether some form of must carry or common carriage rule would lead to good results for social networks, or app stores, or other online services.

    For social networks:

    Definitely not. An unmoderated network (particularly ones with public discussions) would be overwhelmed with low quality content, abusive users, spam, and so on. Or, as we’ve recently discovered, groups organizing themselves for mass violence. Unmoderated platforms like this hinder free expression, since loud, abusive users can drive everyone else away and dominate the conversation. Moderation allows voices to be heard, attracts a more diverse user base, and even allows different social networks to differentiate from each other. We need to regulate social networks and protect users in many ways, but they can be free to make the content moderation decisions they think best (while being subject to criticism for them, of course). The ACLU, which has expressed skepticism of Amazon’s move with AWS and Parler, has recognized that “speech communities” like social networks should be able to define their own rules.

    For app stores:

    Part of the value of an app store is curation — not just technical checks to make sure that software is not harmful, but ideally, a store should carry and highlight apps of decent quality, and help users make choices. An app store can put in place privacy requirements for apps it carries, or require that apps be accessible. But this only works if there are alternatives to monopolistic app stores — otherwise the risk of the abuse of gatekeeper power is too great. As I’ve written, there should be such alternatives. The approaches outlined in my app store paper allow app stores to be curated and manageable while ensuring that users can still access the apps and services of their choice.

    For other services:

    In general the lower level you get, the closer to being pure “infrastructure,” and the fewer the alternatives, the more likely it is to make sense to treat a service as a common carrier or something like it. Getting kicked off AWS, for example, was more devastating to Parler than being removed from the Apple and Google app stores — it rendered the entire service nonfunctional, on every platform. If there are no alternatives to a particular service, and/or the service is important enough, and serves a class of users that might need protection, and there are no likely policy options to create more competition, it might make sense to impose some kind of duty to deal with respect to lawful content. But you’d have to show that the facts bear this out. A harder question is what happens when the private choices of private companies do make it impossible for a service that, for the sake of argument, is perfectly lawful, to be on the internet at all. Do we want some kind of universal, basic fallback option?

    A lawyer’s favorite answer to any question is “it depends.” It’s not always satisfactory. But when it comes to digital platforms and services, it is difficult to even categorize the different kinds of service, much less to determine which rules should apply to which categories. (And it goes without saying that the choice is not between common carrier regulation and nothing! Consumer protection, competition, and even rules against commercial discrimination do not depend on a service operating in a completely “neutral” way.) That said, there are various conceptual tools to use when considering a service, and applying them, common carriage rules are not the appropriate tool for social networks or app stores, and probably aren’t for cloud computing services like AWS, either.

    [1] For more on this point allow me to refer you to one of my favorite law review articles, “The Origin of the Peculiar Duties of Public Service Companies, written by Charles K. Burdick, and published in 1911. Part one, part two, part three.

    [2] Historically and even in case law today there is a notion that common carriers “hold themselves out” as common carriers, and so undertake the obligations voluntarily. Under this theory, for example, an ISP that does not offer nondiscriminatory internet access is simply not upholding its end of the deal, since the very definition of providing internet access requires nondiscrimination. Common carrier rules, then, exist as a means of enforcing these obligations. There is much to this, but it also leads to tedious arguments about the scope a provider has to “hold out” a different service, and whether the government can simply order it to “hold out” a particular service or not. I would argue that, subject to constitutional constraints, the government can indeed order a service to operate as a common carrier, with the arguably more pressing question being when should it use this power.

    [3] It is also useful to note that even common carriers can refuse service, but the rules governing when a common carrier can refuse service vary from service to service. For example, common carriers are generally allowed to protect themselves from harm. But this means something very different for an electric utility than for a taxi or a hotel. Some common carriers have special obligations about when they can suspend service for lack of payment, whereas other common carriers can demand cash up front. Each case is different depending on the reason for imposing common carriage and the impact of losing service.