Introduction to two-sided markets
Digital platforms, such as Amazon, Uber and electronic payment systems, rely on technology to link distinct groups of agents with each other, such as credit card holders with merchants, online buyers with sellers, advertisers with customers, and software developers with users. Digital platforms thus create multisided markets, in which the welfare of participants on each side of the platform is interdependent.
In multisided markets, no side can create economic value alone. Instead, economic value grows with the number of connections and choices available across the multisided platform. For simplicity, in this chapter we will limit our discussion to two-sided markets; however, unless otherwise noted, the principles discussed here apply equally to multisided platforms.
Many types of digital platforms can create two-sided markets.Online exchanges, for example, provide a marketplace for buyers and sellers to connect (e.g., eBay, Airbnb, hotel and travel booking aggregators). Advertising-supported media provides platforms that allow viewers to access content (often for free) while advertisers pay for access to viewers (e.g., advertising-supported TV programming, social networking sites). Software platforms are employed by both application developers and application users (e.g., iOS app developers and app users; Adobe PDF writers and PDF readers). More recently, a number of platforms (service-clearing platforms) have been bringing together buyers and sellers of services by combining algorithm pricing with usage tracking and geographic coordination (e.g., Uber, Lyft). Finally, payment services connect customers with merchants (e.g., credit cards, debit cards). In many cases, these services also involve banks that act as intermediaries to process payments from cardholders to merchants.
Digital platforms have helped technology-based companies such as Apple, Google, Amazon and Facebook grow into industry giants. It is, therefore, not surprising that competition authorities and analysts have been focusing on digital platforms, two-sided markets, their underlying economics and their impact on consumer welfare.
In fact, the economics of two-sided markets are different from those of one-sided markets, primarily because of the complexities introduced by network effects. Economists use the term ‘network effects’ to describe contexts in which a good or service offers increasing benefits or creates greater value the more users it has. For two-sided markets, network effects can be both direct and indirect. Direct network effects arise when users benefit from increased usage by peers. For example, a telephone becomes more useful when other people also use telephones, or Facebook becomes more valuable to a member as more of his or her friends use it.
Indirect network effects come about when the welfare of users on one side of the platform is dependent on the welfare of participants on another side of the platform. For example, a larger number of consumers using credit cards enhances the value of accepting the card platform for merchants and a larger number of merchants with credit card terminals increases the value of having a credit card to consumers. This creates a cascade effect, in which adding users on one side of a platform attracts more users on the second side, which in turn makes the platform more attractive for the first side, and so on.
Over time, this cascade effect may result in the emergence of a dominant supplier or even a ‘winner take all’ outcome. This phenomenon is called ‘tipping,’ where network effects result in most, if not all, users converging on one platform.
On the other hand, network effects can also send a platform into a ‘death spiral.’ For example, if the number of users on side 1 declines (say, because of an increase in price or a decrease in utility gained from the platform), the users on side 2 may value the platform less. The number of side 2 users also declines, further reducing the value of the platform for side 1, and so on.
In markets with competing platforms, tipping and death spirals may be interrelated. For example, the rise of Facebook to become the most popular social network platform encouraged users and advertisers to converge on it, tipping the market for social network platforms to Facebook. Simultaneously, as fewer users participated on Myspace, the platform became less useful for the remaining users on the platform, who then also moved away. The reduction in user base was a disincentive for advertisers and they too left the platform, leading to a death spiral for Myspace.
Network effects, then, significantly complicate competition analysis. All being equal, larger entities offering more connections, and therefore larger networks, offer greater consumer welfare. However, they simultaneously raise traditional concerns associated with supracompetitive pricing if tipping leads to winner-take-all platforms.
Using asymmetric pricing to create value for platforms
Many of the business models for digital platforms are driven by asymmetric pricing. As Rochet and Tirole noted in their ground-breaking 2003 paper,an important feature differentiating two-sided markets from traditional single-sided markets is that ‘platforms must choose a price structure and not only a price level for their service. It is not uncommon for a two-sided market to have asymmetric pricing [Emphasis added].’ Asymmetric pricing may mean that one side of the platform pays nothing or even pays a net negative price, while the other side faces positive prices. This is because the success of the platform requires a critical mass of participants on each side of the platform but may require different incentive structures on each side to achieve that outcome. The different sides may also exhibit different price elasticities and sensitivities.
Google’s business model provides a widely recognised example of asymmetric pricing. Google provides many services to users for free (e.g., online search, email, social networking and maps) all of which are supported by advertising. Its success depends on its ability to attract and retain enough users to make its platforms sufficiently appealing to advertisers.Similarly, credit card companies employ different pricing models for different sides of their networks. They are multisided platforms that facilitate simultaneous transactions between cardholders and merchants. In the case of Visa and MasterCard, these operate as four-party systems (consumers, issuing banks, acquiring banks and merchants), in which issuing banks accept payments from cardholders and transfer those payments to acquiring banks for a fee (the multilateral interchange fee or MIF). The acquiring bank then transfers the payment to the merchant for an additional fee (the merchant fee).
Not only does the pricing across the sides of the platform differ but prices are also untethered from each side’s specific costs; instead, they reflect underlying customer elasticitiesand/or requirements to optimise the value of the platform to all sides. Cardholders may be charged an annual fee and interest by the issuing bank but they may also be offered free cards and reward programmes as incentives (a negative cost) to use a specific card. The issuing bank captures a percentage of every processed transaction paid by the acquiring bank in the MIF. The acquiring bank covers the cost of the MIF as well as its own margin through the fee charged to the merchant. The merchant pays the fee but also makes a profit from the original purchase. Finally, the credit card companies themselves charge fees to the banks for the use of the networks. Ultimately, the convenience of the credit card is provided to both consumers and merchants and paid for through a complicated combination of unequal payments by multiple parties on the different sides of the platform.
Credit card companies often maintain that these separate prices (positive and negative) are balanced to encourage the greatest number of matches between cardholders and merchants. As discussed in more detail below, however, the interchange fee has been a source of ongoing antitrust evaluation globally for more than 40 years, with plaintiffs and regulators claiming that the pricing structure and the associated acceptance rules for these networks result in interchange fees set above competitive levels.
In a two-sided market, the overall pricing structure can be assessed based on the combined pricing and corresponding but potentially unequal benefits to both sides. An assessment of competitive conditions requires a review of the complex interactions between competitive dynamics and consumer preferences, as well as weighing the long-run risks and benefits to innovation, prices, choice and, ultimately, consumer welfare to determine whether actions raise anticompetitive concerns.
Challenges in market definition with digital platforms
Assessing competitive conditions for two-sided markets requires economic analysis that considers competition on each side of the market as well as their interaction. While in single-sided markets only one price is relevant to competition reviews, in two-sided markets the combination of prices charged to each side, as well as any cross-subsidies and asymmetric pricing, are all relevant to any assessment of anticompetitive effects.
A standard market analysis in a competition review seeks to identify relevant product substitutes and the geographic boundaries over which customers might reasonably substitute those alternatives. In these analyses, regulators often use a hypothetical monopolist test (HMT) or an equivalent economic test to determine whether a hypothetical monopolist in a defined market could profitably inflict and sustain a small increase in price (a small but significant and non-transitory increase in price or SSNIP). If not, the set of products included in the hypothetical market is widened until the price increase is profitable. For the purposes of competition review, the relevant market comprises the smallest set of products over which a hypothetical monopolist could exert control.
For example, consider a candidate market for video content, in which viewers pay fees to subscribe to either of two content channels (A and B). If a firm acquired both channels A and B, it might plausibly seek to raise fees to consumers. In a competition review of a one-sided market, whether channels A and B form a relevant market depends on what content is offered by their competitors. Would a sufficient number of consumers switch to a third channel C if channel A-B raised its price? Would it be profitable for A-B to raise its prices? If so, then channel C is not part of the relevant market formed by channel A-B since consumers would not migrate to channel C when faced with a price increase from channel A-B.
However, channels A, B, and C may sell minutes of airtime to advertisers as well as charge a subscription fee to viewers. The channels now compete simultaneously in a two-sided market of viewers and advertisers.
Since having more viewers increases the value of the channel to advertisers, a loss of viewers risks a death spiral for the channel as advertisers, seeing less value from lower viewer numbers, exit as well. An HMT must then consider not only the fees lost to channel C from the viewer side because of a price increase but also advertising revenue lost from the advertiser side. If the additional revenue lost from the advertising side makes a price increase unprofitable, channel A-B may no longer be considered the relevant product market, even though the degree of viewer substitutability with one another and with channel C is unchanged.
Note that the second side of the market (advertisers) must be considered even if channel C does not compete on both sides; for example, even if channel C were ‘subscription only’ and included no advertising revenue, the loss of viewers from Channel A-B due to a subscription price increase might still result in the loss of advertisers.
Alternatively, channels A, B and C may charge viewers a price of zero and generate all revenue via advertising. In this scenario, a hypothetical monopolist with control of channels A and B may leave viewers’ price at zero but raise the price to advertisers. Advertisers may then look to switch to channel C following the price increase, as in the previous example, suggesting that channel C should be in the relevant market for channel A-B. However, channel C may not wish to accept more advertisers if it believes too many advertisers would degrade the quality of the viewers’ experience. To the extent that this occurs, a hypothetical monopolist may still be able to profitably raise prices for channel A-B and channel A-B may remain a relevant market. In this case, the relevant market is dependent on the behaviour not only of users but of advertisers and channel C as well.
As a general matter, the correct definition of the relevant market requires careful attention to price changes and other competitive effects on all sides of the market, including any interactions among the sides of a multisided market.
The evaluation of potential anticompetitive conduct in platform markets also differs from a similar evaluation in traditional markets. Specifically, pricing, contracting and sales practices that may raise anticompetitive concerns in traditional markets may be procompetitive and welfare-enhancing (in a Pareto sense) in multisided markets because they increase value on one or more sides of the market without decreasing value to the other sides.
For example, pricing below marginal cost in one-sided markets may be predatory and anticompetitive. The same pricing practice in two-sided markets or even negative prices (offering incentives) may be procompetitive because it balances incentives on the two sides of the platform and maximises network effects; at the same time, it may also have the effect of excluding rivals. For this reason, determining whether pricing below marginal cost is anticompetitive in a two-sided market is more complicated than in a one-sided market.
Similarly, under some conditions tying by a dominant producer may restrict competition without providing benefits to consumers in single-sided markets.However, in two-sided markets, tying may benefit consumers because the other side of the market may reap benefits from the tying. For example, in the past credit card platforms have prevented merchants from accepting only a subset of cards (e.g., a merchant that accepted MasterCard credit cards would have to accept all types of cards from the MasterCard system; it could not refuse to accept MasterCard debit cards). While merchants may prefer to accept only certain cards (and negotiate lower prices from the accepted cards), consumers may benefit from having cards more widely available for use at merchants. Determining the overall impact on all users requires weighing the relative effects of the policy on the two sides of the market, namely, consumers and merchants.
Finally, when evaluating coordination and default pricing on a platform, one must also examine procompetitive benefits of those same actions. For example, network effects may hamper the creation of a new platform, since a platform is only valuable if a sufficiently large number of users are present on each side. Starting a new payment platform may require coordination across entities (issuing banks, acquiring banks, the scheme operator) and facilitating payments to encourage adoption by both sides of the platform. Absent these agreements, new platforms may struggle to emerge, consequently preventing consumers from obtaining benefits from newer and more innovative platforms.
As discussed in the next section, the legal framework provided by Article 101 of the Treaty on the Functioning of the European Union (TFEU) allows for an assessment that balances procompetitive and anticompetitive effects in the context of agreements between firms in European jurisdictions. An interesting question that has arisen in the context of multisided markets is whether benefits that accrue to one side of the market may be used to offset potential costs to other sides of the market. In other words, is it enough that potential anticompetitive restrictions raise aggregate net benefits above their level without the restriction or does the distribution of these benefits between sides matter? We consider this question next in more detail, with reference to a number of recent European cases on interchange fees.
The role of platform economics in recent UK interchange fee cases
Investigations and litigation surrounding the existence and level of the MIF have a long history in Europe. The European Commission’s Directorate-General for Competition (DG Comp), National Competition Authorities (NCAs) and national courts have all examined the extent to which the pricing models and in particular the setting of default MIFs,employed by the four-party credit card systems in Europe (Visa and MasterCard in particular), restrict competition under Article 101(1). Additionally, investigations and court rulings have examined whether default MIFs may be exempted under Article 101(3), despite their restriction of competition, because they provide positive benefits that outweigh the restrictions.
In what follows, we first provide a short account of recent judgments relating to interchange fees in Europe, touching on legal points only when we feel this is paramount in order to identify issues that have been settled by the courts versus those that are still open. Next, we set out how the issues that are still open hinge on the economic questions discussed above, namely, the assessment of benefits that stem from network effects.
In 2016 and 2017, a series of conflicting judgments in the United Kingdom from the High Court and the Competition Appeals Tribunal (CAT) examined restrictions from setting default MIFs under Article 101(1)and whether Article 101(3) may provide an exemption because of countervailing benefits. Put succinctly:
- In Sainsbury’s v. MasterCard the CAT found that Article 101(1) was infringed, and that default MIFs were not exempt under Article 101(3).
- In Asda v. MasterCard, however, Mr Justice Popplewell found both that setting default MIFs was not an infringement under Article 101(1) and that even if it was, it would have been exempt under Article 101(3).
- Finally, in Sainsbury’s v. Visa, Mr Justice Phillips found that setting default MIFs did not infringe Article 101(1) but also wrote that if it had infringed Article 101(1), it would not have been exempt under Article 101(3).
All three cases were given leave to appeal and were heard concurrently by the Court of Appeal for England and Wales. The court issued its judgment on 4 July 2018.
In line with the EU’s MasterCard jurisprudence,the Court of Appeal determined that setting default MIFs was restrictive of competition under Article 101(1) and that default MIFs could not be seen as an objective necessity for the functioning of the payment scheme. The court, however, did not rule on whether default MIFs, as implemented by Visa and MasterCard, were exempt under Article 101(3). It decided instead to remit the three cases to the CAT for determination of that issue. Importantly, the court provided strong guidance on how the tests for Article 101(3) need to be assessed in the presence of multisided markets.
The court recognised that any benefits arising from the restriction might be of a different magnitude for each side of the market (consumers and merchants). As such, net benefits arising from the restriction needed to be assessed for both sides of the market. Crucially, however, the court found that the market subject to the restriction (in this case, the merchants) must be no worse off than if the restriction was not in place. In other words, for the purposes of assessing whether default MIFs are exempt under Article 101(3), the CAT cannot ‘balance’ any positive effects to consumers against any net negative effects to merchants. The CAT must instead consider the effect of the restriction on both sides of the market and then ensure that the net benefits owing to the restriction are positive; and that the merchants are no worse off than they would have been without the restriction.
Furthermore, any benefits to each side of the market must be causally connected to the restriction and there must be empirical evidence supporting this connection. It is not sufficient for parties to rely solely on economic theory showing that the restriction will give rise to benefits to each side of the market.
In practice, this means that when assessing whether Article 101(3) applies to default MIFs, the CAT will first need empirical evidence showing that the MIFs directly caused issuing banks to increase benefits to card holders (e.g., in the form of increased card rewards, better fraud protection or faster adoption of technologies such as contactless payments). Second, the CAT will need empirical evidence showing that these increased benefits to consumers caused a net benefit to merchants (or at least that merchants were no worse off than they would be without default MIFs). As merchants benefit from an increased use of cards over cash,empirical evidence showing that an increase in card benefits to consumers increased card transactions, rather than redistributed transactions among different cards, will be increasingly relevant. This evidence is firmly rooted in the economics of multisided markets; causally linking the benefits from asymmetric pricing (the MIF) to each side of the market requires correctly accounting for network effects, and the competition within and between platforms for consumers, banks and merchants.
Platform business models are not new, and they have been faced with a robust history of antitrust lawsuits and enforcement over the past 40 years. However, as digital platforms become increasingly prevalent, disagreements remain on the circumstances in which antitrust intervention is warranted. It is clear that the analysis of competition in multisided markets is not a simple extension of single-sided market analysis. Specifically, pricing, contracting and business practices that have typically been considered detrimental to competition in single-sided markets cannot be assumed to be so in multisided settings, in which welfare implications are complex and interconnected across platform sides. Furthermore, differences in approach in the US and EU, especially when assessing benefits from potentially restrictive practices, add another layer of complexity when evaluating and regulating the conduct of global companies. Since market definition, market power and the evaluation of anticompetitive conduct all require special consideration when dealing with a digital platform, the economics of platforms have played, at least to some extent, a role in courts’ judgments thus far.
Many questions remain regarding how courts and regulators in the EU will address antitrust issues in platform markets going forward. However, the recent English Court of Appeal interchange fee ruling has provided some guidance on the nature of the test that should be applied. Specifically, it has held that, as multisided platforms implement pricing models or other features that may restrict competition, they must show empirical evidence demonstrating that these restrictions benefit the platform and specific sub-groups of users directly affected by the restrictions. This evidence is rooted in the economics of platforms, including asymmetric pricing, network effects and platform competition, and therefore it is likely that the economics of multisided markets will play an increasing role in these types of cases in the future.
 Joshua White is a vice president, Antoine Chapsal is a principal and Aaron Yeater is a managing principal of Analysis Group, an economics consulting firm.
 David S Evans and R Schmalensee, ‘The Industrial Organization of Markets with Two-Sided Platforms,’ Competition Policy International, Vol. 3 (1), 2007.
 While Myspace is still active, it has moved away from being a general social network site and has focused on music and streaming video. Furthermore, its Alexa ranking, a metric which assesses the popularity of a website, is approximately 4,300 while Facebook’s rank is 3. See https://www.alexa.com/siteinfo/facebook.com and https://www.alexa.com/siteinfo/myspace.com (Last accessed on 9 October 2018).
 J-C Rochet and J Tirole, ‘Platform competition in two-sided markets,’ Journal of the European Economic Association, 1(4): 990–1029 (2003).
 For example, a free newspaper may bundle admission tickets to a museum or a sample of a new musical artist. To the extent that these items have value to the consumer of the newspaper, the net price of the free newspaper is actually negative.
 Google’s efforts to develop and protect its user base have also led to a number of allegations of anticompetitive behaviour, such as tying its Search and Chrome apps with its app store (Google Play), paying manufacturers to exclusively pre-install its search engine, and preventing manufacturers that sell phones with pre-installed Google apps from selling any Android phones not approved by Google.
 Four-party systems are sometimes referred to as five-party systems, with the fifth party being the scheme itself (e.g., Visa or MasterCard). Note that American Express is a three-party system, in which it interacts directly with both the merchants and the cardholders to process transactions using its card.
 ‘Customer elasticities’ refers to the sensitivity of customer demand in relation to changes in the features or price of a good or service.
 European Commission, ‘Communication from the Commission - Guidance on the Commission’s enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings,’ Official Journal of the European Union, C 45, February 2009, pp. 7-20; ‘Tying the Sale of Two Products, Federal Trade Commission,’ Federal Trade Commission, https://www.ftc.gov/tips-advice/competition-guidance/guide-antitrust-laws/single-firm-conduct/tying-sale-two-products.
 David S Evans, ‘The Antitrust Economics of Multi-Sided Platform Markets,’ Yale Journal on Regulation, Vol. 20 (325), 2003.
 Article 10 of the EU’s Interchange Fee Regulation prohibits card networks from enforcing the so-called ‘Honour All Cards’ rule. In Europe, subject to certain restrictions, merchants are now able to selectively accept scheme cards. (‘Regulation (EU) 2015/751 of the European Parliament and of the Council of 29 April 2015 on interchange fees for card-based payment transactions,’ Official Journal of the European Union, L 123, 19.5.2015.)
 With respect to the cases at hand, the scheme rules established a default level of the MIF charged by the issuing bank to the acquiring bank, when the two were not the same institution. The default level of MIF would then be included in the acquiring bank’s charges to the end merchant for providing card acquisition services.
 Article 101(1) of the TFEU ‘prohibits agreements, decisions by associations of undertakings and concerted practices that are restrictive of competition.’ For further discussion of Article 101(1), see, R Whish and D Bailey, Competition Law, 9th Edition, 2018, pp 82-156.
 The three judgments also made claims under Chapter 1 of the UK Competition Act 1998 and Article 53 of the Agreement on the European Economic Area. However, for ease of discussion in the remainder of the article we refer to the claims as made under Article 101(1). This is in line with the CAT’s decision in Sainsbury’s Supermarkets Ltd v. MasterCard Inc, in which it stated, ‘[t]he distinctions between Article 101 TFEU, Chapter I of the Competition Act 1998 and Article 53 of the EEA Agreement do not – on the facts of the present case – appear to us to be material’, and ‘[s]ave where the contrary is stated, or the context otherwise requires, when we refer to the provisions of Article 101 TFEU, we also intend to refer to the provisions of Chapter I of the Competition Act 1998.’ See, Approved Judgment, Sainsbury’s Supermarkets Ltd v. MasterCard Inc (Rev 1)  EWCA Civ 1536, Court of Appeal (Civil Division), July 2018 at ¶2.
 The European Commission’s guidance for the application of Article 81(3), now Article 101(3), lays out four key tests that need to be met when considering whether a particular agreement is exempt from enforcement under 101(1). ‘The application of the exception rule of Article 81(3) is subject to four cumulative conditions, two positive and two negative: (a) The agreement must contribute to improving the production or distribution of goods or contribute to promoting technical or economic progress, (b) Consumers must receive a fair share of the resulting benefits, (c) The restrictions must be indispensable to the attainment of these objectives, and finally (d) The agreement must not afford the parties the possibility of eliminating competition in respect of a substantial part of the products in question.’ (European Commission, ‘Communication from the Commission – Notice – Guidelines on the application of Article 81(3) of the Treaty,’ Official Journal of the European Union, C 101, April 2004, p. 101). For further discussion of the application of Article 101(3), see Whish and Bailey, Competition Law, pp 157-179. It was generally accepted by the CAT, the High Court and the Court of Appeal that in the case of the default MIF cases, only the first three conditions were potentially at issue. Sainsbury’s v. MasterCard, ¶76.
 Sainsbury’s Supermarkets Ltd v. MasterCard Inc  CAT 11, Competition Appeal Tribunal, August 2016, ¶4, 288.
 Asda v. MasterCard  EWHC 93 (Comm), High Court of Justice (Queen’s Bench Division), January 2017, ¶181, 255.
 Sainsbury’s Supermarkets Ltd v. Visa  EWHC 355 (Comm), High Court of Justice (Queen’s Bench Division), February 2018, ¶1, 67.
 In 2007, the European Commission determined that MasterCard’s default MIFs were a violation of 101(1) and not exempt under 101(3) (European Commission, Commission Decision COMP 34.579). This decision was appealed by MasterCard, first to the General Court which, in 2012, upheld the European Commission’s decision (Judgment of the General Court (Seventh Chamber), MasterCard, Inc v. European Commission, T-111/08, EU General Court, May 2012) and then to the Court of Justice for the European Union, which dismissed MasterCard’s appeal in 2014. (Judgment of the Court (Third Chamber), MasterCard Inc v. European Commission, C-382/12 P, European Court of Justice, September 2014.)
 The court relied, inter alia, on the existence of other card schemes elsewhere in the world that exist without MIFs to conclude that setting a default MIF was not an objective necessity to the survival of the scheme. (Sainsbury’s v. MasterCard, ¶10.)
 This approach seems to deviate from the position taken recently by the US Supreme Court in Ohio v. American Express, where the court expressed the view that all sides of a multisided market must be considered simultaneously without considering the impact on each individually.
 Merchants may prefer to take card payments instead of cash due to transaction costs associated with handling cash. These costs may include needing increased security, frequent bank deposits, increased chance of theft, etc.