This is the second blog in a series on regulating digital platforms. A (less snaky) version first appeared on the blog of my employer, Public Knowledge.
In Part I, I explored the challenges of regulating digital platforms to promote competition, protect consumers, and encourage news production and civic engagement. Today, I plan to dive into the first set of challenges. First, I define what I mean when I talk about digital platforms. I will argue that platforms that (a) provide a two-sided or multi-sided market; (b) are accessed via the internet; and (c) have at least one side that is marketed as a “mass market” service, share a set of characteristics and raise a similar set of concerns so that we should consider them as a distinct set of businesses.
Let me stress at the outset something that I will repeat multiple times. First and foremost, describing the common attributes of platforms does not make value judgments about whether these attributes are bad or good. Indeed, many of the attributes I describe have enormous positive effects for consumers, competition, and civic discourse. At the same time, however, the implications of these specific attributes give rise to a number of unique concerns that we read about every day, ranging from companies using targeted advertising to stalk people to extremists using social media to radicalize and recruit.
Equally important, nothing in sector-specific regulation replaces antitrust or consumer protection laws of general applicability. Nor does it suggest that digital services that do not meet the definition of a “digital platform” do not need oversight. Rather, both the definitions I propose below and the sector-specific recommendations that flow from them (discussed in future blog posts) complement each other. The fact that many platform attributes complicate existing antitrust analysis does not mean that antitrust law has now lost its utility as an important tool for protecting competition. But even embracing a broader view of antitrust law and its goals, there remains an important role for sector-specific regulation to address concerns that arise from the unique nature of digital platforms (as unique from other sectors of the economy).
Finally, before diving in, I must caveat this with the recognition that this is a field very much in flux. I have identified what I think are the important elements which, taken together, make digital platforms different from other lines of business or even other “internet companies.” Nor is this the only potentially useful distinction. In the past, for example, I have argued that we should also distinguish between “public utility” concerns (services so important the government has an affirmative responsibility to ensure affordable access for everyone) and services that, while important, do not rise to this level. Deputy Director of Georgetown Law’s Center on Privacy and Technology Laura Moy, in testimony before the House Energy and Commerce Committee, provides an excellent distinction between “essential services” and “unavoidable services,” i.e., services so ubiquitous they are virtually impossible to avoid in one form or another. Others have different definitions of platforms, and/or different distinctions among them.
The definition I propose here is therefore not intended as a final conclusion, but an initial working definition to debate and refine over time.
With all that out of the way, lets move on to the good stuff . . .
What I Mean by “Digital Platform.”
When considering whether we need regulation, let alone what that regulation is, it helps to know what the heck we are talking about and what we are trying to accomplish. Or, as I like to say, the most important question in policy is “why do we care?” Are we talking about something that should apply to all businesses? For example, we all agree that no business should engage in deceptive advertising — even if we might disagree on what constitutes “deceptive.” For this kind of behavior, general laws of applicability suffice. But when a line of business or a particular set of products raise unique concerns by their very nature, then we want sector-specific regulation.
We Pause Briefly For A Hopefully Useful Analogy But Please God Don’t Get All Caught Up In The Details Because I’m Just Trying To Illustrate A Basic General Point About How The Above Statement Works.
For example, human beings have relied on various means of transportation since we domesticated the horse. But the invention of the automobile and its rapid adoption in the early 20th century required legislators to address entirely new problems created by cars. Some of these, like traffic control, were older problems magnified by the enormous speed of cars, which increased both the likelihood and potential consequences of collisions. Other regulations, such as regulations governing noise and parking (and later, fuel standards and emissions), were entirely new problems created by the nature of the technology. (If you are curious about the entirely different problems of managing horse poop, horse parking, and other fun aspects of urban horse traffic and how it differed from car traffic, see here.) Still other regulations, like “lemon laws,” were made necessary because of the economic and social patterns that arose organically with the adoption of the automobile.
Once it became clear that we needed specific regulations for cars, we found we needed to distinguish cars from other forms of transportation, and then further distinguish motorized transport based on things like size and purpose. Motorcycles, SUVs, compact sedans, and trucks are all motorized vehicles that share certain common characteristics, but they are also different from each other. In some ways they are regulated the same, but in other ways they are regulated quite differently. Sometimes the distinction is rational, sometimes arbitrary. But in general we can distinguish between a car and a horse-drawn buggy, or a car and an airplane, or a car and a train, even if certain edge cases (such as between a passenger car and an SUV) are somewhat more arbitrary.
Now Back To Platforms. Here’s The Juicy Definition Stuff.
Likewise, there is increasing consensus that a growing class of online companies operates in ways that neither network economics or standard industrial organization fully explain. Most of the attention focuses on the largest companies — notably Google, Facebook, and Amazon. But at the same time, we have increasingly recognized a broader group of “edge providers” or “platforms.” But what common thread ties together companies as diverse as Yelp, Twitter, and Vimeo? And what, if anything, distinguishes them from companies like Cloudflare and Comcast?
As my colleague John Bergmayer noted in his recent white paper, “Even Under Kind Masters,” the term “platform” is rather ambiguous. Bergmayer summarized various ways in which people have used “platform” as a forum for speech, as an operating system for development, or as a set of components around which users organize their activities.
Looking at commonalities of these uses, as well as what economic or other factors make sense to construct a set of things that defines “platforms” and excludes “non-platforms,” I propose the following definition. For purposes of this discussion, a digital platform:
1. Operates as a two-sided or multi-sided market;
2. the service is accessed via the internet; and,
3. at least one component of the platform is “open” and a mass market service.
These three factors combine to produce entities operating under broadly similar economic incentives and which raise a set of issues/concerns that are common to all such platforms (even if the services delivered are radically different) — and which are not wholly shared by other services. Put another way, Google and Amazon have much more in common with each other than they do with Netflix — despite the fact that all three stream video and that Google and Amazon have radically different businesses and business models from each other.
(No, Netflix is not a “digital platform,” it is a fancified cable programmer with more in common in terms of incentives with HBO than with Amazon or Youtube. Mind . . . Blown!)
Why Do These Features Matter More than Others?
Potentially low marginal cost, network effects (particularly the cost of exclusion), and the ability to scale rapidly to absorb millions of new customers make these platforms distinct from other types of businesses. The digital nature of the platform allows it to rapidly deploy new features, and integrate data across multiple apparently unrelated business lines or sources (especially if these are also digital). These factors allow platforms to avoid any of the traditional costs associated with rapid expansion, both vertically and horizontally. These features distinguish platforms from other traditional two-sided markets, and allow platforms to combine elements of traditional communications networks and mass media, as well as traditional retail market networks.
As Jean Tirole, who won the Nobel Prize for his work on two-sided markets, observed here, today’s dominant platforms began as niche segment vendors. Amazon, for example, began exclusively as an online bookstore. The features described above allowed it to expand relatively rapidly first from books, to other products, then to streaming, and finally to manufacturing of its own generic brands. Once a sufficiently large customer base began using Amazon for one purpose, it was much easier for Amazon to expand than it would have been for a traditional book chain such as Barnes & Noble or Borders. Its established distribution network (both the online access and the physical process of moving goods from one place to another) could be readily adapted for other goods, without any need to alter existing physical stores or dealing with what products to display in scarce shelf space. The relationship, algorithms for recommending related items, and the convenience of “one click” shopping were all readily and seamlessly expandable in a way that would be impossible for comparable brick-and-mortar retailers.
It is important to note that, as with any of the characteristics that I describe, other successful (or even dominant) businesses will replicate some of the features described. Walmart, for example, likewise expanded its retail services to include pharmaceuticals, groceries, and even pre-paid cell phone service. It is the combination of being online, multi-sided, and open on at least one side (so as to capture a giant audience) that confer unique advantages, shape incentives, and raise concerns of enduring (rather than merely transitory) market power. In particular, the fact that platform users potentially play multiple roles simultaneously distinguishes digital platforms from other two-sided platforms or internet businesses that have clear distinctions between providers and consumers.
A Multi-Role User in a Multi-Sided Market.
Unlike in traditional two-sided markets, a single user may simultaneously engage in multiple roles on the platform. A subscriber to YouTube is potentially a producer of content and a consumer of content. A customer on Amazon may simultaneously be a reviewer, a buyer, and a publisher or retailer. This has several effects on the ability of the platform to extract value, avoid traditional costs, and maximize bargaining power over all platform users regardless of their comparative value or what role they play in the transaction.
First, this “multi-sided market” maximizes the “long tail” effect, which is where the true value of the platform lies. This is distinct from the more traditional “network effect,” which holds that an increase in the number of users of the network increases the value of the network to all users (although platforms also experience network effects). It is also different from economies of scale, which allow businesses to reduce marginal cost per unit due to increased scale (again, sufficiently large platforms may enjoy these as well). Rather, as popularized by Chris Anderson in his book of the same name, the idea of the “long tail” is that the value of a platform is derived from aggregating large numbers of niche products (the “tail”) rather than focusing on a few very successful products (the “hits”).
Consider, for example, a traditional cable package or an online streaming service such as Netflix. It is easy to divide the platform between subscribers/viewers and programmers. The value to the user derives chiefly from the availability of a suite of programming. If a major programmer withdraws its programming, the video provider may suffer as customers migrate to rival distributors of the programming. A package that lacks “must-have” programming (such as local live sports) will prove less able to attract subscribers than rivals who have the “must-have” programming. (Yeah, yeah. Judge Leon doesn’t believe it. I know.)
By contrast, Amazon does not particularly worry about any specific streaming content because its streaming service is merely part of its overall bundle. Streaming is simply one more product, like batteries or self-published novels, that attracts some portion of consumers. It is part of the overall “long tail” of goods and services Amazon offers. Similarly, there is no single must have content that attracts some significant number of YouTube’s customers. Even the most popular YouTube channel accounts for a tiny fraction of total YouTube views. Rather, it is the access to the platform, with the ability to upload or download literally billions of clips, that makes YouTube an attractive service (combined, of course, with it being free to end users). As a result, no single programmer, or even group of programmers, can effectively negotiate with YouTube. (Youtube’s recent forays into the traditional distribution business don’t change this any more than Amazon buying Wholefoods made Amazon a bricks-and-mortar business.)
Similarly, as news organizations keep discovering, any website can withdraw its content from Google’s search index. Doing so, however, will have little impact on the value of Google to users and will therefore have zero impact on Google’s revenue — which derives from targeted ads. It would require some huge portion of the internet to “go dark” to Google Search before it significantly impacted the value of Google Search to customers — and therefore to advertisers. This is simply not possible.
It is noteworthy that even the comparatively simple model of video distributor and subscriber becomes potentially more complicated when the distributor plays the additional role of content producer by generating its own content. This can create vertical integration concerns if the distribution platform is dominant (and therefore its ability to favor affiliated content threatens competition among video producers) or if the distributor controls “must-have” programming (and can therefore inhibit the ability of rival distributors to compete). We should therefore not be surprised that a platform that enables multiple users to play multiple roles simultaneously creates an entirely different set of incentives, potential benefits, and potential concerns — especially when coupled with the additional attributes described above and below.
Why This Potentially Creates Enduring Market Power in Ways That Challenge Modern Antitrust Analysis.
At this point, it is worthwhile to point out (and repeat) several things. First, this feature of platforms is not intrinsically a bad thing. To the contrary, platforms empower consumers and producers to play multiple roles simultaneously, which creates many important benefits. Services like Patreon or Twitter make it easy for anyone to disintermediate traditional gatekeepers and leverage that platform to find other interested parties and engage in whatever joint, community related activities the platform supports. For example, “Black Twitter” describes how traditionally fragmented and marginalized African American activists and communities can bypass traditional bottlenecks to disseminate news, organize, and otherwise create a distinct cultural identity using the open Twitter platform. Teachers organizing for higher pay in West Virginia and elsewhere credit Facebook for providing them with the tools to communicate and organize. Millions of people are able to use platforms such as eBay or Etsy to supplement their income or create entirely new businesses without the need to negotiate individually with the platforms. The ability to create content and distribute it through platforms such as YouTube, Amazon, or Facebook allows individuals and organizations freedom to make their work accessible broadly whether or not they can prove to a traditional publisher it will be a commercial success.
But the ability of platforms to potentially put all this together creates a combination of user “stickiness” and a flexibility of revenue stream that, once enormous market share is achieved, is likely to become enduring. It creates a common set of incentives among platforms to engage in a strategy of taking long-term losses and cross-subsidizing services in order to defeat new entrants and maintain sufficient dominance across sufficient markets to hold monopsony power across a wide swath of related industries. It drives innovative startups to seek acquisition by dominant platforms rather than invest in competing services, and it drives dominant platforms to acquire potential competitors not merely because the acquisition of the potential competitor increases this depth of services, but because it neutralizes a potential rival.
This challenges existing antitrust jurisprudence in several ways. For example, ease of entry and low switching cost — features associated with platforms because of their digital nature and accessibility online — are usually mitigating factors against a finding of market power when considering potential mergers such as the acquisition of Instagram and WhatsApp by Facebook. This is particularly true where the service does not directly compete in a traditional sense (e.g., Instagram is designed for distribution of images, whereas Facebook at the time was primarily a “microblogging” site). But in the realm of digital platforms, this may eliminate a potential competitor. As described above, the online and digital nature of the service would have potentially allowed Instagram to expand quickly into services provided by Facebook. The more significant challenge than entering a new “market” is building a sufficiently large audience.
By focusing on acquisition of platforms that are experiencing high growth, even where they do not directly compete in a traditional sense, dominant platforms can dramatically delay, or even prevent, the emergence of future competitors. The digital and online nature of the dominant platform and the acquired platform reduce the cost of integration and increase the depth of service offered by the dominant platform, making it more difficult for firms to compete. Additionally, if the business model of the online platform depends on creating a detailed user profile for targeting ads or products (as pretty much all the dominant platforms do to some degree), the ease of integration noted above allows the acquisition to enhance the existing dominance of the platform by providing another source of personal user information.
Finally, the multiple roles/depth of service of platforms also stymies traditional antitrust analysis because there is no single, easily definable market. Facebook is not merely a “social network” competing with LinkedIn, Twitter, Reddit, and Livejournal. Facebook is a unique combination of services that includes a massive network of businesses, political speakers, and other social networks like WhatsApp and Instagram. This goes beyond traditional product and market differentiation, because the value to users on both sides of the platform is in part derived by the combination of services, not competition among services.
Again, we can find some analogies in other markets. For example, cable operators argued for decades that individual broadcast television stations, movies, and home video recordings were all competitors for “eyeballs” and thus part of the same market. Regulators rejected this argument because while each of these replicated some piece of what a cable subscription provided, the unique combination of multiple sources of programming distinguished cable (and later other “multichannel video programming distributors”) from these other providers of video.
Similarly, the attempt to define a new “attention economy” and concomitant “attention marketplace” falls short of the way in which this multifaceted combination creates value to the platform (and, to be fair, to users as well) and plays havoc with traditional market definitions. Because switching costs are extremely low, and because applications through which these services are accessed are generally non-rivalrous, the platform can continue quite nicely as users cycle from low engagement to high engagement. Certainly the incentive of the platform is to maximize engagement. But market power by dominant platforms proves more enduring than predicted because, in contrast to other markets where consumers buy one product or another, I can happily continue to consume several competing comments with virtually no effort. The ability of these platforms to form joint promotional partnerships further enhances the endurance of market power once established.
Another Example For Illustration Purposes. I Know It’s Anecdotal and Don’t Get Too Hung Up On It.
To illustrate using a personal example. I had no interest in Twitch.tv until I saw an advertisement on YouTube that TwitchPresents was running a marathon of all classic Doctor Who episodes. I accessed Twitch through “Twitch Prime,” a service of Amazon Prime which allows an Amazon Prime subscriber to subscribe to one Twitch Channel as a premium viewer. I usually keep open my YouTube browser viewer while watching Doctor Who on Twitch because during the commercials at the end of each Doctor Who episode, I bounce back to YouTube to watch something else. Meanwhile, since I am accessing Twitch via Prime, I assume that Prime derives some value and gathers some information from my Twitch activities. Meanwhile, because Twitch has an entirely different business model, its revenue stream is completely unimpacted by my shifting attention from one platform to another. (I, by the way, am a pathetic fuddy-duddy when it comes to this sort of multitasking entertainment. I have only to glance over at my son playing Twitch in the background while playing multiple games and chats across three different devices.)
Contrast this with my traditional voice/broadband/video subscription package. I used to subscribe to Verizon FIOS. I switched to RCN. FIOS lost me as a customer, while RCN gained me as a customer. Zero-sum game. By contrast, although Twitch, Prime, and Youtube are theoretical competitors in the video streaming market using classic antitrust analysis, none of them have lost me as a customer. Even my shifting attention may not impact significantly my value to the service as a subscriber.
Perfect Information Asymmetry.
Finally, the combination of features puts the platform in a unique position with regard to platform users and control of information. The platform enjoys essentially perfect information with regard to the activities of users on the platform. Importantly, this includes not simply information about consumers, but also information about content producers, advertisers, or anyone else using the platform for any purpose. By contrast, the user will only have access to the information that the platform enables the user to collect. Additionally, the platform can make different levels of information available to different users on an individualized basis – although sophisticated users may also find ways to reverse engineer data and exploit the platform in potentially harmful or even dangerous ways.
This has implications well beyond privacy and surveillance (although these are obviously enormous concerns). This blog post is already too long to even scratch the surface on the literature of information asymmetry and its effects on competition and consumer protection. But it is worth pointing out some of the concerns raised recently with regard to platforms. For example, Amazon reportedly uses the information about sales by third-party vendors through its platform as market research to develop its own line of competing products. Google has been accused of manipulating search results to favor its own products. Facebook has admitted to conducting secret experiments on its users to influence their moods. Advocates have raised concerns that the ability to understand users and their behavior to an unprecedented degree facilitates “design for addiction.”
In particular, it is the opacity of the algorithm that platforms use to make recommendations and order the presentation of products, news, or services that can create concerns in ways even the platforms cannot anticipate. The ability of the platforms to analyze user behavior drives the recommendations of Google’s search algorithms, Facebook’s news feeds, and Amazon’s product recommendations. But a user — whether a consumer or a content producer — cannot easily determine what factors drive the recommendations. Even advertisers who specify particular attributes they desire for targeted placement have tremendous difficulty confirming that these advertisements are being placed appropriately beyond the tools provided by the platform.
To repeat a now familiar caveat, this ability of the platform to potentially control the information flow is not, in itself, a good or bad thing. It is a feature of the digital nature of the platform, combined with the integration of the component parts via the internet. Consumers enjoy enormous benefits from recommendations tailored to their needs or tastes. Search tools and tools for organizing the proliferating deluge of information depend on absorbing and processing vast amounts of information, and the ability of the platform to limit dissemination of that information plays an important role in protecting user privacy.
Nevertheless, the fact that near-perfect control of information is both a natural artifact of the platform and in some cases a necessary (or socially desirable) feature in providing the service does not eliminate concerns. To the contrary, it highlights the need for targeted rules and protections that carefully analyze both the dangers and the benefits and tries to arrive at a reasonable trade-off between enabling the positive and mitigating the negative.
Sector-Specific Policy Should Direct Itself to Those Attributes That Make the Sector Different. What About These Differences Raise Concerns?
Again, the point of the definition is not simply to identify problems. It is to understand what (if anything) makes digital platforms distinct as actors as compared to other business sectors. This combination of enabling producers and consumers to simultaneously play multiple roles, the openness of the platform to attract a critical mass of participants that produce and consume the content, and the advantages of online distribution create these distinctions.
If this were simply a matter of antitrust and traditional concerns about dominance, we could resolve these problems simply by enforcing the antitrust law. But one of the advantages of sector-specific regulation is that we can use policy not merely to protect existing competition, but to promote competition. By understanding what distinguishes digital platforms from other sectors of commerce, we can carefully target any needed regulation to supplement antitrust enforcement to achieve this goal.
Similarly, we should target consumer protection concerns to supplement existing state and federal consumer protection laws of general applicability based on the unique features of digital platforms and the incentives they create both for the platforms and for actors using the platform. Two examples illustrate this principle. We have specific laws targeting robocalls enforced by both the Federal Communications Commission and the Federal Trade Commission. This reflects the fact that while robocalling is a generically bad practice by any business, it is the use of the telephone and telephone technology that enables robocalls. Similarly, although privacy is a broad concern, we have specific laws governing privacy in the health profession as a consequence of the unique sensitivity of medical records and the need to ensure that patients can trust that their medical records will not be exposed without their express consent.
Who Should Enforce Sector-Specific Regulation of Digital Platforms?
As is often the case, people in Washington often start out with an answer and then work backward rather than starting with the question. Even before defining what digital platforms are, people have been eager to argue that they properly belong under the purview of the FTC or the FCC. For purposes of our ongoing discussion, let’s set that question aside until we determine what actual policy we need. Determining what regulation we need to protect the public interest will instruct us on which existing agency is best suited to enforce these policies. We may divide enforcement among several agencies. We may even determine we need a new specialized agency to address the unique concerns raised by digital platforms.
In other words, we should let regulatory form follow function rather than force function to follow form. We should start by considering what policies we need to protect consumers, promote free expression and civic discourse, and encourage competition. Once we have some idea of what we need to do, we can draft the job description for who we want to do it.
We can say with confidence that digital platforms are a distinct set of businesses, with their own set of capabilities and incentives that set them apart from other online businesses. This does not make laws of general applicability such as antitrust inapposite. Nor are these distinct capabilities and incentives intrinsically bad or good. Identifying the nature of digital platforms and understanding the implications of this increasingly important sector of the economy is critical to understanding both how to update antitrust and other generally applicable laws, and when sector-specific regulation is required to promote the public interest.
In particular, this analysis should make clear that regulation based on surface similarities is not merely unlikely to be helpful, but may be downright harmful. Facebook is not the same as Apple, which is not the same as Microsoft, which is not the same as AT&T. To treat all of these as “platforms” makes as much sense as insisting that we treat motorcycles, 18-wheeler trucks, trains, and bicycles the same because they are all “vehicles.” Even where digital platforms share some attributes with other lines of business, understanding the differences (and whether they matter in the particular instance) is essential to good policy.
It is a common cliché that “if it looks like a duck and quacks like a duck, it’s a duck.” Unless, of course, it is actually a clever hunting decoy. For ducks, distinguishing between these two possibilities is rather essential. Similarly, for the formulation of good policy — whether through updating antitrust law or applying sector-specific regulation — understanding what makes digital platforms different is critical to protecting and promoting the public interest.
In my next installment, I will discuss what actually makes a digital firm “dominant.”
Stay tuned . . .