Practical Competition Policy Implications of Digital Platforms

Concern about the size and market power of large digital platforms has been on the increase, widely reflected in the media. The growing concern is manifested in official actions such as the European Commission’s €2.42 billion fine imposed on Google in 2017 for abusing its dominance in “demot[ing] rival comparison shopping services in its search results” to advantage its own service, a €4.34 billion fine on Google in 2018 for using restrictions on Android device manufacturers to “cement its dominant position in general internet search,” and a €1.5 billion fine in 2019 for abuse of its dominance in online advertising. It also manifests itself in policy proposals and investigations such as the European Commission’s ongoing Digital Single Market program and other interventions, including the General Data Protection Regulation (GDPR) and the wide-ranging Australian Competition and Consumer Commission inquiry into the market power of digital platforms.

However, without greater consensus in the academic and practitioner literature as to how to analyze digital platform market power and the competitive process for digital platforms, it will prove difficult for policymakers to address these concerns. There is a rapidly expanding theoretical literature on these models, but no consistent analytical framework and scant systematic empirical evidence about platform markets and consumer welfare effects (in other words, the benefits individuals derive from the consumption of digital platform services as assessed by themselves in light of their own preferences). As the earliest work in this literature dates back only to about 2003, it may be too early to expect research to have translated into practical competition policy tools.

Unfortunately, the need on the part of antitrust authorities is urgent. Since the arrival and rapid spread of smartphones after 2007, subsequent business model, and algorithmic innovations, the proliferation of digital platforms and growth of some of them has been striking. Some digital platforms have substantial scale and market share, such as Facebook with an estimated 63 percent of social media users globally in mid-2018 and Google with an estimated 90 percent share of the online search market. The need for some practical approaches is imperative.

While further rigorous theoretical and empirical research is needed, this essay offers immediate practical proposals for antitrust authorities. It focuses on three areas to consider in any competition assessment of digital markets:

• What are the incentives to invest and innovate, given that competitive dynamics in these markets takes the form of disruptive innovation? As major innovation is often due to new entrants rather than incumbents, have the incumbents erected barriers to entry at minimum commercially viable scale?
• What dynamic consumer harms offset static consumer benefits in digital platform markets, and what is the magnitude of each?
• What are the effects on the incentives to invest and innovate across the ecosystem? What is the impact on competition in adjacent and upstream markets from entry or acquisition by a large incumbent digital platform?

Even in the absence of standardized techniques comparable to familiar existing procedures (such as the SSNIP—“small but significant and non-transitory increase in price”—test), answers to these questions need to go beyond the conventional focus on firms’ pricing behavior and provide practical guidance for antitrust investigations and merger inquiries firmly rooted in the analysis of consumer welfare.

I. Why the Economics of Platforms Challenges Traditional Competition Assessment

An immediate difficulty in antitrust analysis is the lack of a clear, broadly agreed definition of a multi-sided platform (MSP), and even what terms to use to refer to MSPs. Although there is agreement about some of the core economic characteristics of MSPs, the economics literature contains several variants, of different degrees of precision. MSPs are agreed to have at least two “sides,” users and providers, or buyers and sellers. There are indirect network externalities such that participants on each side benefit, the more numerous are the participants on the other side. For example, travelers benefit from more hotels being on the platform, and hotels from there being more potential travelers, on an online travel agency platform. The opportunity for an MSP exists when the different sides cannot transact separately to capture the value of the indirect network externalities themselves. While MSPs have existed for a long time (e.g., bazaars, stock exchanges), new digital MSPs have greatly extended the scope of possible transactions due to matching algorithms and the technologies of broadband and smartphones.

More formally, “[A] market is two-sided if the platform can affect the volume of transactions by charging more to one side of the market and reducing the price paid by the other side by an equal amount; in other words, the price structure matters, and platforms must design it so as to bring both sides on board.” However, this description is not easy for competition authorities to translate into an operational definition as it is difficult to distinguish from cross-subsidies between product lines in non-platform businesses. The rapidly growing literature on MSPs also refers to MSPs as businesses, markets, or networks, as well as platforms, in a largely descriptive way without specific formal distinctions. The terminological variation in the literature is not surprising, as MSPs have features of both productive businesses, which produce products or services, and marketplaces, which enable exchanges to be made. For example, MSPs use various types of exchange coordinating mechanisms (such as matching algorithms, technical standards, or information capture and classification ) in place of the traditional business production coordination via time and place. In addition, some organizations operate both as one-sided and two-sided businesses. For instance, Amazon operates as a retailer while Amazon Marketplace operates as a platform. Many digital organizations are loosely considered as platforms, while some non-digital-era businesses are also considered with hindsight to be platforms. Netflix is an example of the former, as it has a conventional linear business model. Examples of the latter include nightclubs, or stock exchanges.

Dirk Auer and Nicolas Petit note that the scholarly literature has failed to converge on a definition because of a lack of conceptual clarity about which businesses merit classification as MSPs, resulting in disagreements in the literature. Operating as a platform is a business model choice—for instance, selecting advertising funding rather than a subscription model. Auer and Petit propose the importance of the Coaseian opportunity of reducing transactions costs and thus facilitating exchange, rather than network effects alone, as a defining feature: in other words, MSPs reduce transactions costs by enough to enable transactions that would otherwise not occur. In this essay, I use the term digital MSP for consistency, but my analysis is not confined to any specific form of a digital platform.

If either new ex ante regulations or existing competition policy analytical tools are to be applied to digital MSPs, then a clearer operational definition is necessary. The firms, collectively known as GAFAM (Google, Amazon, Facebook, Apple, Microsoft), as well as other large or fast-growing businesses, such as Uber, Airbnb, Booking.com, and Deliveroo, are involved in a wide range of different activities, only some parts of which can be characterized as digital MSPs. There is growing concern about the scale and perceived power of these firms.

From the perspective of competition authorities, digital MSPs pose several challenges specific to the economic characteristics of digital MSPs.

First, to what extent should sheer size be a concern?

Second, the price on each side of the market is not reflective of marginal cost on that side, as the average price level across all sides and the price
structure (i.e., the price on each side) are set separately in digital MSPs. This makes traditional SSNIP tests non-operational as the prices set by the platform on each of its “sides” cannot be considered in isolation, so that looking at the high price side alone as a potential indicator of market power could give a misleading result. It is always the case with digital MSPs that the least priceelastic side cross-subsidizes the other (usually, although not always, the consumer side is subsidized).

Third, the standard market definition exercise can often be misleading because of the feedback links between the two (or more) sides of the MSP. As a result of these feedback links, envelopment is one form of competition between MSPs. Envelopment refers to the practice of adding another group of users or “side” of the platform, to gain additional revenues from another existing side. MSPs often adopt an envelopment strategy once it has a large user group “on board.” For instance, Uber moved into food delivery with UberEats, adding food providers in addition to drivers to take advantage of its large consumer “side.” Similarly, Google moved beyond its original search business to add a wide range of others such as payments, email, productivity software, mapping (not all platform-type models). MSPs may also adopt the bundling or tying of services to cross-subsidise between different groups of users when the MSP is unable to set a negative price to subsidize one side directly. Envelopment and bundling strategies have the potential to build barriers to entry in the “enveloped” markets because it will be harder for smaller MSPs, without so many groups on board, to match the prices or services of the bigger incumbent that is able to cross-subsidize.

Fourth, with any digital MSP, while price is clearly relevant to the welfare assessment, the products or services are highly differentiated, making price a poor metric. Variety and better matching of supply and demand features create economic welfare. Antitrust enforcement has for a considerable time applied a consumer welfare standard, often attributed to the influence of Robert Bork’s book The Antitrust Paradox and subsequent Chicago School work. In practice, prices have been taken as the indicator of consumer welfare, but in any technology market, product characteristics will be at least as important. When the consumer-side price is zero, as on many digital MSPs, all the direct
competitive pressure is exerted through service quality and innovative features. Although competition guidelines often refer to quality and other characteristics as features of competition, in practice the focus is on price because it is easier to define and easier to measure. For example, the UK’s Competition and Markets Authority Merger Assessment Guidelines state that competition “creates incentives for firms to cut price, increase output, improve quality, enhance efficiency, or introduce new and better products,” but subsequent amplification in the Guidelines refers almost entirely to prices.

Fifth, the phenomenon of “multi-homing” may have counterintuitive implications. Multi-homing refers to a user signing up to several MSPs. On the consumer side, the presence of multi-homing is evidence of competitive pressure on MSPs because if the MSPs delivers an inferior service or price, the MSP is likely to lose consumers on that side of its business. On the supplier side, however, strong competition among suppliers could lead to them avoiding multi-homing because if rival suppliers select different MSPs, the suppliers may dilute the competitive pressures they face, to the extent that buyers on the other side do not monitor all MSPs. In this case there would be multiple similar MSPs charging high fees on the supplier side. If there is less competitive pressure among suppliers, the MSP network effects will lead suppliers to join a dominant MSP. The MSPs will compete for the market and charge low fees to users on the supplier side. The correlation between price and digital MSP concentration is the opposite of what intuition suggests; in this case, higher concentration is correlated with lower fees.

Sixth, MSPs face what the literature refers to as the “chicken and egg” problem. The “chicken and egg” problem refers to the need to expand both sides of the digital MSP in balance. The reason is that MSPs are characterized by indirect network effects. Direct network effects refer to the positive externality or spillover one user derives from other members of a network; for instance, a telephone is more valuable the more other people have telephones. Indirect network effects are positive externalities due to the presence of other groups. For example, a restaurant booking platform needs enough diners to make it worthwhile to restaurants paying the commission to the platform, and there need to be enough restaurants signed up for consumers to use the platform to search for a place to eat. The indirect network effects attracting users
on each side to the digital MSP mean that the digital MSP has to be sure to have enough of each kind of user.

Until the digital MSP reaches critical mass, it is likely to be loss-making. However, once the digital MSP reaches the critical point, the platform can quickly grow to a large scale as a result of indirect network effects. This dynamic makes it hard to interpret profitability in terms of competition and market dynamics. While, many MSPs fail without ever having made a profit, when an MSP does reach profitability, investors have a reasonable expectation of a return commensurate with the risk. Some digital businesses, such as Amazon, report negative to low (economic and accounting) profits for a long period of time. A long period of no or low profits is sometimes seen as a cause for concern in terms of competition, as the motivation is taken to be the desire to grow to a dominant position in one or more markets. In addition, a long period of no or low profits is an inversion of the normal economic intuition, in which one would interpret low profits as a lack of competitive concern, not heightened competitive concern.

Continue reading the full text of this article in PDF format.