Discussions of antitrust and the FAGA companies–that is, Facebook, Amazon, Google, and Apple–often sound like a person with a hammer who just wants to hit something. Here\’s Peggy Noonan writing in the Wall Street Journal (June 6, 2019):
But the mood in America is anti-big-tech. Everyone knows they\’re too powerful, too arrogant, loom too large in public life. … Here\’s what they [Congress] should be thinking: Break them up. Break them in two, in three; regulate them. Declare them to be what they\’ve so successfully become: once a pleasure, now a utility. It all depends on Congress, which has been too stupid to move in the past and is too stupid to move competently now. That\’s what\’s slowed those of us who want reform, knowing how badly they\’d do it. Yet now I find myself thinking: I don\’t care. Do it incompetently, but do something.
When it comes to regulation of the big digital firms, we may end up with incompetence at the end of the day. But first, could we at least take a stab at thinking about what competent antitrust and regulation might look like? For guidance, I\’ve seen three reports so far this year on antitrust and competition in the digital economy:
- One is a report done by an expert panel for the UK government. Jason Furman, Diane Coyle, Amelia Fletcher, Philip Marsden, and Derek McAuley, Unlocking Digital Competition: Report of the Digital Competition Expert Panel (March 2019).
- One is a report done by outside advisers for the European Commission: Jacques Crémer, Yves-Alexandre de Montjoye, and Heike Schweitzer, Competition Policy for the Digital Era (April 2019).
- The remaining report was written by a committee composed of Fiona Scott Morton, Pascal Bouvier, Ariel Ezrachi, Bruno Jullien, Roberta Katz, Gene Kimmelman, A. Douglas Melamed, and Jamie Morgenstern, called Committee for the Study of Digital Platforms: Market Structure and AntitrustSubcommittee Report, published by the Stigler Center at the University of Chicago (May 2019)
The three report are all at least 100 pages in length. They are all written by prominent economists, and they have a lot of overlap with each other. Here, I\’ll just focus on some of the main points that caught my eye when reading them. But be aware that the topics and terms mentioned here typically come up in at least two of the three reports, and often in all three.
Antitrust Policy is Not a One-Size-All Solution for Everything
Antitrust policy is about attempting to ensure that competition arises. Pace Noonan, it\’s not about whether the firms or the people running them are powerful, arrogant, or large.
Digital technologies raise all sorts of broad social issues that are not directly related to competition. For example, the appropriate treatment of personal information would still be an issue if there were four Facebook-like firms and five Google-like firms all in furious competition with each other. Indeed, it\’s possible that a bunch of new firms competing in these markets might be more aggressive in hoovering up personal education and passing it along to advertisers and others. The question of if or when certain content should be blocked from the digital sites will still be an issue no matter how many firms are competing in these markets.
Digital firms use algorithms in many of their decisions about marketing and price, and such algorithms can end up building in various forms of bias. This is a problem that will exist with or without competition. The forces of corporate lobbying on the political process are a legitimate public issue, regardless of whether the lobbying comes from many smaller firms, or a larger firm, or an industry association.
In other words, the digital economy raises lots of issues of public interest. Responding to those issues doesn\’t just mean flailing around with an antitrust hammer, hoping to connect with a few targets, but instead thinks about what problems need to be addressed and what policy tools are likely to be useful in addressing them.
What is the Antitrust Issue with Digital Firms?
In the US economy, at least, being big and being profitable are not violations of existing antitrust law. If a firm earns profits and becomes large by providing consumers with goods and services that they desire, at a price the consumes are willing to pay, that\’s fine. But if a firm earns profits and becomes large by taking actions that hinder and block the competition, those anticompetitive actions can be violation of antitrust law. The challenge with big digital firms is where to draw the line.
When the internet was young, two decades ago, there was a widespread belief that it would not be a hospitable place for big companies. [T]he economic literature of the beginning of the 21st century assumed that competition between online firms would arise as consumers hopped from site to site, easily comparing their offers. The reality however quickly turned out to be very different. Very early in the history of the Internet, a limited number of “gateways” emerged. With the benefit of hindsight, this might not be too surprising. Users have limited time and need curators to help them navigate the long tail of websites to find what they are looking for. These curators then developed a tendency to keep users on their platform, and by the end of the 1990s, it was common place to speak about AOL’s “walled garden”. AOL’s market power however rested in great part on its role as an Internet service provider and both competition in that domain and, according to some observers, strategic mistakes after its merger with Time Warner eroded its power.
Fast forwarding to today, a few ecosystems and large platforms have become the new gateways through which people use the Internet. Google is the primary means by which people in the Western world find information and contents on the Internet. Facebook/WhatsApp, with 2.6 billion users, is the primary means by which people connect and communicate with one another, while Amazon is the primary means for people to purchase goods on the Internet. Moreover, some of those platforms are embedded into ecosystems of services and, increasingly, devices that complement and integrate with one another. Finally, the influence of these gateways is not only economic but extends to social and political issues. For instance, the algorithms used by social media and video hosting services influence the types of political news that their users see while the algorithm of search engines determines the answers people receive to their questions.
As all of these reports note, it is now apparent that when a large digital company becomes established, entry can be hard. There is a \”network effect,\” where a system that has more users is also attractive to more users. There\’s a \”platform\” effect, where most buyers and sellers head for Amazon because the most buyers and sellers are already on Amazon. There are economies of scale, where the startup costs of a new platform or network can be fairly high, but adding additional users has a marginal cost of near-zero. There are issues involving the collection and analysis of data, where more users mean more attraction to advertisers and more data, to which can then be monetized.
None of these issues are new for antitrust. But the big digital firms bring these issues together in some new ways. The markets become prone to \”tipping,\” which means that when an established firm gets a certain critical mass, other firms can\’t attract enough users to survive. As economists sometimes say, it becomes a case where there is competition for the market, in the sense of which firm will become dominant in that market, but then there is little competition within the market.
One consequence is that when a big firm becomes established, entry is hard and future competition can be thwarted. Thus, there is reason to be concerned that consumers may suffer along various dimensions: price, quality innovation. In the areas of innovationstartup in that area slows down dramatically. From the Scott Morton et al. report:
By looking at the sub-industries associated with each firm—social platforms (Facebook), internet software (Google), and internet retail (Amazon)—a different trend emerges. Since 2009, change in startup investing in these sub-industries has fared poorly compared to the rest of software for Google and Facebook, the rest of retail for Amazon, and the rest of all VC for each of Google, Facebook, and Amazon. This suggests the existence of so-called “kill-zones,” that is, areas where venture capitalists are reluctant to enter due to small prospects of future profits. In a study of the mobile app market, Wen Wen and Feng Zhu come to a similar conclusion: Big tech platforms do dampen innovation at the margin. Their study analyzed how Android app developers adjust their innovation strategies in response to entry (or threat of entry) by Google …
Some Oddities of a Market with a Price of Zero
Big digital firms provide a remarkable array of services with a market price of zero to consumers. They make their money by attracting users and selling ads, or by charges to producers. Is this a case where the market price should actually be negative–that is, the big tech firms should be paying consumers. The Scott-Morton et al. report offers some interesting thoughts on these lines:
Barter is a common way in which consumers pay for digital services. They barter their privacy and information about what restaurants they would like to eat in and what goods they would like to buy in exchange for digital services. The platform then sells targeted advertising, which is made valuable by the bartered information. But, in principle, that information has a market price. It is not easy to see if the value of any one consumer’s information is exactly equal to the value of the services she receives from the platform. However, many digital platforms are enormously profitable, and have been for many years, which suggests that in aggregate we do know the answer: the information is more valuable than the cost of the services. …
Online platforms offer many services for zero monetary price while they try to raise participation in order to generate advertising revenue. Free services are prevalent on the internet in part because internet firms can harness multi-sided network externalities. While the low price can be a blessing for consumers, it has drawbacks for competition and market structure in a world where institutions have not arisen to manage negative prices. Because there is currently no convenient way to pay consumers with money, platforms are able to mark up the competitive price all the way to zero. This constraint can effectively eliminate price competition, shifting the competitive process to quality and the ability of each competitor to generate network externalities. Depending on the context this may favor or impede entry of new products. For example, entry will be encouraged when a price of zero leads to supra-competitive profits, and impeded when a zero price prevents entrants from building a customer base through low price. Moreover, unlike traditional markets where several quality layers may coexist at different price levels (provided that some consumers favor lower quality at low price), markets where goods are free will be dominated by the best quality firm and others may compete only in so far as they can differentiate their offers and target different customers. This strengthens the firm’s incentive to increase quality through increasing fixed costs in order to attract customers (known as the Sutton sunk cost effect) and further pushes the market toward a concentrated market structure. …
It is a puzzle that, to date, no entrepreneur or business has found a way to pay consumers for their data in money. For example, a consumer’s wireless carrier could aggregate micropayments across all manner of digital destinations and apply the credit to her bill each month. … Furthermore, a carrier that could bargain effectively with platforms on behalf of its subscribers for high payments would likely gain subscribers. Notice that an easy method to pay consumers, combined with price competition for those consumers, might significantly erode the high profits of many incumbent platforms. Platforms likely have no economic incentive to work diligently to operationalize negative prices.
Of course, this idea of a market in which consumers barter attention for zero-marginal-price services isn\’t new. Television and before that radio operate on the same basic business model.
Categories of Data
The ability to collect data from users, and then to collate it with other information and pass it along to advertisers, is clearly a central part of the business model for digital firms. There\’s a lot of quick-and-easy talk about \”my\” data and what \”they\” should or shouldn\’t be able to do with it. The Cremer et al. report offers some useful drilling down into different ways of acquiring data and different ways in which data might be used. Sensible rules will take these kinds of distinctions into account. They note:
Data is acquired through three main channels. First, some data is volunteered, i.e. intentionally contributed by the user of a product. A name, email, image/video, calendar information, review, or a post on social media would qualify as volunteered data. Similarly, more structured data—directly generated by an individual—like a movie rating, or liking a song or post would also fall in the volunteered data category.
Second, some data is observed. In the modern era, many activities leave a digital trace, and “observed data” refers to more behavioural data obtained automatically from a user’s or a machine’s activity. The movement of individuals is traced by their mobile phone; telematic data records the roads taken by a vehicle and the behaviour of its driver; every click on a page web can be logged by the website and third party software monitors the way in which its visitors are behaving. In manufacturing, the development of the Internet of Things means that every machine produces reams of data on how it functions, what its sensors are recording, and what it is currently doing or producing.
Finally, some data is inferred, that is obtained by transforming in a non-trivial manner volunteered and/or observed data while still related to a specific individual or machine. This will include a shopper’s or music fan’s profiles, e.g. categories resulting from clustering algorithms or predictions about a person’s propensity to buy a product, or credit ratings. The distinction between volunteered, observed and inferred data is not always clear. …
[W]e will also consider how data is used. We will define four categories of uses: non-anonymous use of individual-level data, anonymous use of individual level data, aggregated data, and contextual data.
The first category, non-anonymous use of individual-level data, would be any individual-level data (volunteered, observed, or inferred) that was used to provide a service to the individual. For instance, a music app uses data about the songs a user has listened to in order to provide recommendations for new artists he or she might enjoy. Similarly, a sowing app uses data from farm equipment to monitor the evolution of the soil. Access to individual-level data can often be essential to switch service or to offer a complementary service.
The second category, anonymous use of individual-level data, would include all cases when individual-level data was used anonymously. Access to the individual-level data is necessary but the goal is not to directly provide a service to the individual who generated the data in the first place. These would typically include cases of data being used to train machine-learning algorithms and/or data used for purposes unrelated to the original purposes for which the data has been collected. An example of this would be the use of skin image data to train a deep learning (Convolutional Neural Network) algorithm to recognise skin lesions or the use of location data for trading purposes. In specific cases, the information extracted, e.g. the trained algorithm, can then be used to provide a better service to some of the individuals who contributed data. For instance, film reviews are used collectively to provide every individual with better recommendations (collaborative filtering). For the anonymous use of individual-level data, access to a large dataset may be essential to compete.
The third category, aggregated data, refers to more standardised data that has been irreversibly aggregated. This is the case for e.g. sales data, national statistics information, and companies’ profit and loss statements. Compared to anonymous use of individual-level data, the aggregation is standard enough that access to the individual-level data is not necessary.
Finally, contextual data refers to data that does not derive from individual-level data. This category typically includes data such as road network information, satellite data and mapping data.
It\’s interesting to consider whether people\’s objections about use of their data are rooted purely in principle, or are not about being paid. If firms used your observed data on location, shopping, and so on, but only sold aggregated versions of that data and paid you for the amount of data you contributed to the aggregate, would you still complain?
Some Policy Steps
After this probably over-long mention of what seemed to me like interesting points, what are the most useful potential margins for action in this area–at least if we want to take competent action? Here are two main categories of policies to consider: those related to mergers and anticompetitive behavior, and those related to data.
1) Big dominant firms deserve heightened scrutiny for actions that might affect entry and competition. This is especially true when a firm has a \”bottleneck\” position when everyone (or almost everyone) needs to go through that firm to access certain service. One particular concern here is that big dominant firms buy up smaller companies that might, if they had remained independent, offered a form of competition or a new platform.
Furman et al. write:
There is nothing inherently wrong about being a large company or a monopoly and, in fact, in many cases this may reflect efficiencies and benefits for consumers or businesses. But dominant companies have a particular responsibility not to abuse their position by unfairly protecting, extending or exploiting it. Existing antitrust enforcement, however, can often be slow, cumbersome, and unpredictable. …
Acquisitions have included buying businesses that could have become competitors to the acquiring company (for example Facebook’s acquisition of Instagram), businesses that have given a platform a strong position in a related market (for example Google’s acquisition of DoubleClick, the advertising technology business), and data-driven businesses in related markets which may cement the acquirer’s strong position in both markets (Google/YouTube, Facebook/WhatsApp). Over the last 10 years the 5 largest firms have made over 400 acquisitions globally. None has been blocked and very few have had conditions attached to approval, in the UK or elsewhere, or even been scrutinised by competition authorities.
But along with mergers, there are a variety of other potentially actions which, when used by a dominant firm, may be potentially anticompetitive. Another concern is when big dominant firms start offering a range of other services on their own, and then use their dominance in one market to favor their own services in other markets. Yet another issue is when big dominant firms choose a certain technological standard that seems more about blocking competition than advancing its own business. Some dominant platform firms use \”best-price clauses,\” which guarantee that they will receive the lowest possible price from any provider. Such clauses also mean that if a new platform firm starts up, it cannot offer to undercut the original provider on price.
In other words, if a large dominant firm keeps your business (and your attention) by providing you with the quality and price of services you want, and earns high profits as a result, so be it. But if that firm is earning high profits by seeking out innovative ways to hamstring the potential competitors, it\’s a legitimate antitrust problem.
2) Data openness, portability, and interoperability.
Data openness just means that companies need to be open with you about what data of yours they have on-hand–perhaps especially if that data was collected in some way that didn\’t involve you openly handing it over to them. Portability refers to the ability to move your data easily from one digital firm to another. For example, you might be more willing to try a different search engine, or email program, or a different bank or health care provider if all your past records could be ported easily to the new company. Interoperability refer to when technical standards allow your (email, shopping, health, financial or other) data to be used directly by two different providers, if you desire.
Again, the underlying theme here is that if a big digital firm gets your attention with an attractive package of goods and services, that\’s fine; but if it hold you in place because it would be a mortal pain to transfer the data they have accumulated on you, that\’s a legitimate concern.
Finally, I would add that it\’s easy to focus on the actions of big digital firms that one sees in the headlines, and as a result to pay less attention to other digital economy issues of equal or greater importance. For example, most of us face a situation of very limited competition when it comes to getting high-speed internet access. In many ways, the growth of big digital firms doesn\’t raise new topics, but it should push some new thinking about those topics.