INTRODUCTION
Market definition belongs to the small group of concepts that are essential to all instruments of EU competition law:Footnote 1 the prohibition of anti-competitive coordinationFootnote 2 and abusive practices,Footnote 3 as well as merger control.Footnote 4 It is primarilyFootnote 5 employed as a first step to assess a firm's market power.Footnote 6 The definition of a market is meant to provide a framework to help identify the competitive constraints market players face. Thus, market definition is used to constrain the scope of the analysis of factors relevant for determining market power,Footnote 7 as well as to identify barriers to entry.Footnote 8
When, in 1997, the European Commission issued its current Market Definition Notice, the rise of the digital platform economy was still to come. Today, it is the digital platforms that appear to pose the greatest challenges to competition practice: search engines, social networks, online marketplaces, mobile operating systems and app stores, online travel agencies, video platforms, real-estate portals, and the like. What characterises these platforms is that they create and manage network effects,Footnote 9 particularlyFootnote 10 as they act as intermediaries between different user groups who are linked through cross-group network effects,Footnote 11 a feature commonly referred to as ‘two-sidedness’:Footnote 12 decisions of users in one group are materially dependent on the decisions of users in another group and the platform operator has the opportunity to significantly influence those decisions.
How these interdependencies between two or more user groups should be taken into account when defining markets raises a number of fundamental issues that were not addressed in the Market Definition Notice of 1997.Footnote 13 This is remarkable because two-sidedness is not a feature unique to digital platforms. Capturing the business models of newspapers, (ad-financed) TV, payment card systems, or shopping centres—to name but a few instances of ‘traditional’ two-sided platforms—poses essentially the same problems. However, digital platforms have increased the scale and scope of activities, and some control their own eco-systems. This has led to complaints by other market participants and become a general policy concern. It is against this background that, in the last two decades, the economic theory of two-sided markets has experienced considerable progress and its implications for competition policy have been widely discussed.
Adapting the Market Definition Notice to meet the realities and challenges posed by the rise of digital platforms will therefore be the central concern of the review of the Notice launched by the Commission in April 2020.Footnote 14 Thus, this evaluation must be seen as one of a series of measures by the Commission to adapt the EU competition law framework to cope with the (digital) platform economy. In this vein, the Commission has issued guidance on Article 22 of the EU Merger RegulationFootnote 15 and is considering further calibrations to the merger rules,Footnote 16 is also about to reform the Vertical Block Exemption Regulation,Footnote 17 and has proposed a ‘Digital Markets Act’.Footnote 18
The remainder of this article is structured as follows. In Part II we explain that the Commission, national competition authorities and courts are indeed well advised to pay attention to market definition precisely because, in the context of two-sided platforms, the concept is more cumbersome to implement. In Part III, we address fundamental challenges that come along with applying market definition on two-sided platforms. In Part IV, we consider two issues on the proper role of market definition that are conceptually intertwined and that underlie our position on the correct market definition in the context of two-sided platforms. First, we explain why in such a context the calculation of market shares is less informative as a proxy for market power. Second, we argue that market definition must not (conclusively) determine the scope of a competition analysis. In Part V we conclude.
I. WHY THE COMMISSION IS RIGHT TO ATTACH IMPORTANCE TO MARKET DEFINITION IN (DIGITAL) PLATFORM MARKETS
In their report on ‘Competition Policy for the Digital Era’, commissioned by Competition Commissioner Vestager, the authors Crémer, de Montjoye, and Schweitzer do not call into question the instrument of market definition as such but state that:
in the case of platforms, the interdependence of the markets becomes a crucial part of the analysis whereas the role of market definition traditionally has been to isolate problems. Therefore … less emphasis should be put on the market definition part of the analysis, and more importance attributed to the theories of harm and identification of anti-competitive strategies.Footnote 19
The premise of this statement should indeed not be controversial: how competition analysis can do justice to the interdependencies between the effects certain market conduct has on the different user groups of two-sided platforms is one of the key challenges the digital era poses to competition law. The questions, however, of how much emphasis should be placed on market definition and, particularly, how many resources should be devoted to it in competition enforcement by courts and authorities, has to be put in perspective.
First of all, as we will discuss more in detail in the next Part,Footnote 20 it is true that the established methodologies for defining markets need to be adapted when applied to two-sided platforms. Market definition in the context of two-sided platforms is more complex and cumbersome, and consumes more resources. Consequently, its results are more prone to error and, moreover, the simple statistics that are derived after defining the market become less informative for the competition analysis. In particular, market shares are less meaningful as an indicator of a two-sided platform's market power, the assessment of which is an essential element in merger control,Footnote 21 in abuse cases,Footnote 22 and when appraising the anti-competitive effects of an agreement.Footnote 23 This is because the said interdependencies between different user groups have to be taken into account and can, depending on the facts of the individual case, lead both to a situation where high market shares cannot be regarded as an indicator of a high degree of market power and to a situation where a platform should be regarded as having a high degree of market power even if it has only a relatively low market share on this ‘market side’.Footnote 24 These findings especially call into question intervention thresholds based on market share.
Furthermore, the rise of (two-sided) digital platforms poses the challenge for competition authorities of identifying cross-market anti-competitive strategies that are new or appear to be less relevant in other contexts. Just to name two examples: the Commission's Google Shopping decisionFootnote 25 could prove to be a catalyst for the development of a doctrine of abusive ‘self-preferencing’, the scope of which is mainly seen in the area of two-sided digital platforms. The concentration processes in the digital sector have fuelled the discussion over whether merger control should pay much closer attention to acquisitions that cut off potential competition.Footnote 26
Against this background it becomes clear that competition analysis in relation to two-sided (digital) platforms needs flexible thinking that must not be limited from the outset by conventional ideas of market definition and market power assessment. Thus, the introductory quote by Crémer, et al. is indeed convincing when taking the position of a competition authority that observes the potentially anti-competitive conduct of a platform or which has to assess a merger that involves one or several platforms. In such constellations, typically it does not seem reasonable to commence the investigation by devoting resources to a detailed and well-founded market definition. Market definition is inappropriate as a screening device or a first filter to identify competition problems with two-sided platforms. Instead, a competition authority is usually well advised to concentrate its resources on the analysis of theories of harm and an identification of possible anti-competitive strategies.
The picture changes, however, as soon as it is taken into account that market definition is in many cases mandatory for the application of EU competition law. Moreover, it is in particular the courts that are faced with competition cases—whether by way of judicial review of decisions of competition authorities or by way of private litigation—that benefit from a better transparency of the competition analysis and a clearer focus on the key arguments that go hand in hand with market definition. Viewed from this angle, the finding that market definition is more complex, error-prone and possibly less informative in the case of two-sided platforms may indeed lead to quite the opposite conclusion to the one stated above: the (correct) application and interpretation of market definition on two-sided platforms requires special diligence and attention.
A. The Positivist Angle: EU Competition Law Requires to Define Markets
A competition law system might entirely dispense with defining markets. In particular, economic instruments make it possible to measure market power directly. Whether such a system is preferable in terms of administrability, implementation costs, error costs, legal certainty, etc.Footnote 27 appears to be doubtful but may remain open at this point. The practical relevance of market definition as a concept of EU competition law becomes obvious by the fact that it is not only widely used by courts and authorities but considered practically mandatory in various contexts,Footnote 28 in particular where market power has to be measured in order to determine whether or not:
-
– agreements between undertakings give rise to restrictive effects on competition pursuant to Article 101(1) TFEU,Footnote 29 and whether these effects are appreciable;Footnote 30
-
– agreements between undertakings afford these undertakings the possibility of eliminating competition in respect of a substantial part of the production in question and thus are prohibited even though they fulfil the other requirements of an exemption under Article 101(3) TFEU;Footnote 31
-
– an undertaking is below the market-share thresholds that define the scope of application of block exemption regulations;Footnote 32
-
– an undertaking is market-dominant pursuant to Article 102 TFEU;Footnote 33
-
– a concentration would (not) significantly impede effective competition, in the common market or in a substantial part of it, in particular as a result of the creation or a strengthening of a dominant position pursuant to Article 2(2) and (3) EC Merger Regulation.Footnote 34
Even if market definition is more complex and less informative in cases involving two-sided platforms, it seems rather unlikely that the European Court of Justice (‘ECJ’) will shift away from requiring it where it considers it mandatory in non-platform cases. Thus, while defining relevant markets is for good reasons not at the beginning of an investigation that concerns a (digital) two-sided platform, to make a decision watertight, courts and authorities will ultimately have to define the relevant markets at least in those contexts in which the ECJ has repeatedly considered it a necessary element of a competition law analysis.Footnote 35
B. On the Jurisprudential Function of Market Definition and Why the Commission Should Be Concerned about It
Including market definition as a formal step in a competition analysis, thereby combining an approach structured and defined by law with economic methods, forces those who invoke and enforce competition law to carefully consider, interpret and (verbally) explain when and to what extent substitution may restrict a firm's market power. This not only disciplines competition authorities or parties that carry out and put forward a competition analysis before a court but makes their arguments more transparent to the parties affected by a decision, the opposing parties in private litigation and, last but not least, the courts that have to review an authority's decision or to judge in disputes between private parties.Footnote 36 Thus, not only can the requirement to define markets make competition practice more transparent and predictable to the firms concerned; it also facilitates the work of the courts and, consequently, may help to avoid judicial errors and improve the effectiveness of judicial redress.
This is not to say that the direct application of economic methods to substantiate a theory of harm would necessarily lead to greater legal uncertainty.Footnote 37 But including market definition in an analysis may help courts to review cases more efficiently, because it may limit from the start the theories of harm and countervailing efficiency effects that might be put forward by the parties. Consequently, one may indeed assume that, insofar as market definition was postulated by the courts as a necessary prerequisite for the application of EU competition law, this was done in order to guarantee legal certainty and to facilitate judicial review.Footnote 38 This rationale would be missed if there were no legal clarification of those (fundamental) issues that market definition raises in the case of two-sided platforms.
This leads to the question of why the Commission in particular should take on this task. Since, after all, while the Commission itself, when exercising its discretion, is bound by the rules it has laid down in the Market Definition Notice,Footnote 39 it is the ECJ that has the last word on the interpretation of the EU competition provisions. One could therefore assume that the clarification of legal issues regarding market definition lies in any event in the hands of the European courts or, indirectly, the national courts as they can initiate preliminary references to the ECJ.Footnote 40
Analysis of the European courts’ adjudication on market definition has revealed that the Court of Justice and the General Court tend to attach great importance to the Commission's statements in the Market Definition Notice.Footnote 41 There should be no doubt that, as a matter of principle, the rules on market definition as developed and applied by the Commission should be subject to full judicial review. It is otherwise only for their actual implementation, insofarFootnote 42 as it involves complex economic assessments.Footnote 43 Yet, as one observer noted, ‘[o]ne is hard pressed to find a case, since the [General Court] was created, with any level of substantial analysis of [market definition] by the ECJ’.Footnote 44 This indicates that the Commission's 1997 Notice was a truthful codification of the court's preceding adjudication and that the Commission's positions on the legal issuesFootnote 45 of market definition were barely challenged by the parties and, thus, seem to have been essentially undisputed. This leaves us with the—admittedly somewhat trivial—insight that a careful analysis of the questions of law involved with market definition on part of the Commission provides valuable guidance to the European courts and saves judicial resources.
Further, it must not be ignored that the Commission enjoys a special position vis-à-vis the national courts and authorities: while the latter are not bound by guidelines, notices, and other soft law instruments,Footnote 46 they are bound by the Commission's decision-making practice in the circumstances covered by Article 16 of Regulation 1/2003. In fact, Member State courtsFootnote 47 and authoritiesFootnote 48 often seem to regard it as an obvious option to follow the Commission's view as expressed in its soft law instruments without further ado.Footnote 49 Thus, the distinguished position of the CommissionFootnote 50 should be seen also as entailing a particular responsibility to provide guidance on the legal rules that govern market definition in EU competition law.
Finally, it should not be forgotten that, comparatively, the Commission has great expertise and considerable resources at its disposal. Certainly, this also involves a wide range of tasks. But, when considering how to use its resources, it should not underestimate the crucial importance of the positive externalities it may create to the benefit of the national courts and authorities by elaborating and publishing its own legal position on market definition in the platform economy.
C. Interrelation with Pro-competitive Regulation of Digital Platforms
Uncertainties as to the scope of competition law has given rise to initiatives of regulating digital platforms. One aspect in this respect is the lack of clarity over market definition. For instance, in its initiative for a ‘Digital Markets Act’Footnote 51 the Commission expressed its belief that current EU competition law is inadequate to protect functioning competition in the era of large digital platforms acting as gatekeepers. This view was in part based on clear-cut limits to competition enforcement, but also on considerable uncertainties as to its actual scope and problems of implementation—including uncertainties concerning adequate application of the established methods of market definition.Footnote 52
Such uncertainties have also encouraged regulation at national level. Thus, for instance, the German legislature has established a right for payment service providers to access technical infrastructure that contributes to mobile and internet-based payment services (the so-called ‘Lex Apple Pay’).Footnote 53 One reason for this regulatory intervention—although certainly not the only oneFootnote 54—was that it appeared to be uncertain whether Apple, as the operator of mobile devices, could be regarded an addressee of Article 102 TFEU or the equivalent provision under German competition law. That depends in particular on the question of how the power that a platform derives in an environment with multi-homing on one side and single-homing on the other side (a ‘competitive bottleneck’Footnote 55) is considered within the market definition framework.Footnote 56
While new ex ante regulation may be necessary when current competition law fails to ensure competitive outcomes, this comes at a price. Apart from the immediate expenditure of legislation, it creates (new) legal uncertainty and is prone to error, in particular as it may overstep the mark. Moreover, when implemented at the national level such as in the case of the German ‘Lex Apple Pay’, regulation leads to fragmentation within the internal market and, therefore, will likely trigger a political debate on harmonising legislation at Union level.
By outlining how the concept of market definition has to be implemented on two-sided platforms, the Commission may contribute to clarifying the scope of EU competition law. This may help to tailor new ex ante regulation more precisely or, in exceptional cases, even to avoid it. Moreover, competition enforcement will maintain its significance alongside ex ante regulation only if the necessary adaptations of its essential concepts to the digital era have been clarified.
II. MAJOR CHALLENGES POSED BY THE TWO-SIDEDNESS OF PLATFORMS AND HOW TO COPE WITH THEM
To define markets, one has to identify the goods and services offered by an undertaking and to understand substitute offers. This is more challenging in the case of two-sided platforms as they typically offer complex and interrelated products. In the following, we address four key issues.
A. Single-Market Approach vs. Multi-Markets Approach
In the context of a two-sided platform, one approach is to define a market for each side. Consequently, each market can be analysed separately while taking into account that they are linked through cross-group network effects. This is referred to as the ‘multi-markets approach’. Alternatively, one may define a single market for an intermediation service offered to both sides of the market. This is referred to as the ‘single-market approach’.
1. Established EU Competition Practice: Multi-Markets Approach to Payment Card Systems
The Commission and the EU courts had to consider this issue with regard to payment card systems.Footnote 57 In the MasterCard case, the General Court rejected the applicants’ view that there was only one product market at issue, namely a market where the payment card systems provided a single service offered to both cardholders and merchants and where they competed against each other and against all other forms of payment. Thus, the General Court confirmed the Commission's view that the ‘issuing side’ and the ‘acquiring side’ should be considered separate marketsFootnote 58 and reinforced this position in a subsequent judgment involving the French payment card system Cartes Bancaires.Footnote 59
In Cartes Bancaires, the ECJ held that such a definition of separate issuing and acquiring markets must not, however, have the effect that interdependencies with the ‘acquiring side’ of the payment system had to be excluded from the analysis of agreements on the ‘issuer side’. The court stressed that network effects between the two user groups must be taken into account when assessing whether the payment system's restrictions on the issuing of cards should be regarded as a restriction of competition by object or effect under Article 101(1) TFEU.Footnote 60 To reach this conclusion, the ECJ did not take an explicit stand on the applicability of the multi-markets approach as such. The judgment is most crucial, however, because it clarifies in any case that doing justice to the interrelation between the different sides of a two-sided platform does not require adopting a single-market approach. Thus, the court implicitly accepted the multi-markets approach under the condition that cross-group network effects are considered for the definition of the markets on the two sides of the platform and at subsequent stages of a competition law analysis.
2. On the (In)Adequacy of a Single-Market Approach in the Case of Matching Platforms
While in recent years the Commission has dealt with quite a number of cases that involved two-sided digital platforms, it appears that in none of these cases did the Commission explicitly engage in a discussion on the question of the conditions under which a multi-markets approach or a single-market approach should be applied. Yet the Commission's practice indicates that it analyses these cases based on a multi-markets-approach throughout. First of all, in various cases the Commission defined separate markets for online advertising and user content.Footnote 61 In all those decisions, the application of a multi-markets approach does not seem to have been in dispute.Footnote 62 This approach is, however, considered more controversial for so-called ‘matching platforms’, ie platforms that facilitate transactions such as payment card systems, online marketplaces, hotel booking or real-estate platforms, or platforms that enable a different kind of interaction, for instance online dating platforms. Most prominently, the US Supreme Court in Ohio v American Express Co argued that ‘two-sided transaction platforms, like the credit-card market … facilitate a single, simultaneous transaction between parties’ and, consequently, adopted a single-market approach.Footnote 63 The UK's Competition and Markets Authority stipulated in a merger decision that involved two online food ordering platforms that ‘where the platform is “matching” or facilitating transactions … a single market definition is appropriate, which takes account of the competitive constraints on both sides of the market’.Footnote 64
The Commission has taken a different route in its practice on payment card systemsFootnote 65 and in particular its reasoning for the definition of a distinct market for comparison shopping services in Google Search (Shopping) also reveals the use of a multi-markets logic in the context of online platforms. Most notably, when separating comparison shopping services from merchant platforms from a demand-side perspective, the Commission did not presume a single intermediation service offered to two user groups but distinguished between the consumers’ perspective and the online retailers’ perspective, and gave separate reasons why it assumed that comparison shopping services and merchant platforms serve a different purpose.Footnote 66 Given this differentiated reasoning, it would have been desirable for the Commission to have also stated clearly that, correspondingly, two interrelated markets can be distinguished: one in which Google Shopping offers consumers a service to find products they may be interested in, and another one where it offers sellers a listing service that helps them to reach consumers.
A view on the practice of the Member States’ authorities and courts reveals a considerable heterogeneity as to the appropriate approach in case of matching platforms. After the Netherlands Competition Authority, in a 2007 merger case that involved two horticultural platforms, opted for a single-market approach,Footnote 67 it was most notably the Bundeskartellamt that explicitly considered the issue. As a matter of principle, the authority assumes that separate product markets for each side of a platform should be defined. However, in the case of matching platforms, the Bundeskartellamt regards a single-market approach as feasible if ‘user groups essentially have the same need for liaising with the respective other group, and therefore, the groups’ views regarding substitutability of function do not differ substantially’.Footnote 68 On that basis, the Bundeskartellamt assumed, for example, with a view on Amazon's online marketplace that the possibilities to substitute may be more limited for retailers than for consumers, who want to shop and who could switch to competing retailers (online and offline).Footnote 69 This suggests separate markets for the provision of online marketplace services to retailers and to consumers.Footnote 70 In the same vein, the National Commission of Markets and Competition (‘CNMC’) in Spain opted for a multi-markets approach in a merger case that involved online food ordering platforms. The CNMC assumed in particular different opportunities for consumers and restaurants to bypass the intermediation services offered by these platforms:Footnote 71 while the former could readily directly contact the restaurants that offered to deliver food, the latter were dependent on the intermediary services of the platforms to gain access to their customers. Consequently, the authority defined, first, a market where the online platforms offered their intermediation services to restaurants and, second, a market where consumers could order food with delivery service both via the ordering platforms and directly from restaurants.Footnote 72
Yet, the Bundeskartellamt either adopted a single-market approach or considered such an approach at least feasible in merger cases that involved online dating platforms,Footnote 73 online comparison platformsFootnote 74 and real-estate platforms.Footnote 75 The Autorité de la concurrence, the French competition authority, referred to this latter decision when it also applied a single-market approach in a merger decision that involved real-estate platforms.Footnote 76 It is remarkable that the adoption of a single-market approach in these cases does not seem to be the result of a differentiated analysis of potential substitutes on both sides of the platform.Footnote 77
Instead, the Bundeskartellamt took a shortcut: it put forward an ‘indivisibility’ argument, essentially claiming that an intermediation service offered to two (or more) ‘market sides’ could not be split up into two markets, as such a multi-markets approach could not fully do justice to the interdependencies of the ‘market sides’. Moreover, the authority referred to the nature of the ‘matching platform’ as it stipulated, for instance, in its decision involving online dating platforms, that the two user groups, ie men and women who are looking for a partner, would inevitably meet again if they switched to conceivable alternatives.Footnote 78 Such a shortcut in the reasoning, however, merely begs the question of the respective possibilities of men and women to do so without the use of an online dating platform. If, for example, men's and women's inclinations to use an online dating platform or user behaviour (eg the frequency of usage) is asymmetric, this will result in different possibilities of substitution. It is conceivable that there are specialised platforms where both groups are very asymmetrically distributed. Yet, when a user considers switching from a platform with an imbalanced gender ratio to one with a balanced gender ratio, this implies that the attractiveness of such a switch is likely to depend on the user's gender. Certainly, online dating platforms may have a self-interest in achieving a balanced gender ratio, and there are instruments available that may be used for this purpose, such as advertisements that specifically target one user group, or an adaptation of the price structure for using the service. But, then again, it seems rather doubtful to assume without hesitation that there are equal opportunities to substitute.
Furthermore, one must not ignore that different people may use a matching platform for different purposes and with different intensity; thus, a platform seen by some as a good substitute may be seen by others as a bad substitute and there may be systematic differences between the two sides. Also, users on one side may typically be active on multiple platforms, while users on the other side may be active only on one; this will affect substitution possibilities. The intermediary offers fundamentally different services to the two sides.Footnote 79 It would be quite a coincidence if the substitution patterns were symmetric on both sides of the platform.Footnote 80 Moreover, competition analyses often draw on market prices or price changes, such as in the case of merger control by means of pricing pressure tests. But, if the prices differ between the respective user groups (as apparently in the case of online dating platforms), it remains unclear which is the single price that should be relied upon following the single-market approach.
This shows that, with regard to platforms that aim at brokering transactions between two user groups and even in the case of online dating platforms—which seem to many an obvious candidate for a single-market approachFootnote 81—it cannot simply be assumed that the competitive situation of the platform is symmetrical in relation to both user groups. But, even if an investigation concluded that the two sides were symmetrical (in terms of characteristics and the way the platform sets prices), the single-market approach does not offer any benefits over the multi-markets approach because the economic analysis of the latter could simply consider two identical markets characterised by symmetric cross-group network effects. However, symmetry may be observed in the status quo and disappear in a counterfactual, eg when evaluating the competitive effect of contractual restrictions imposed on users on one side. To study such effects meaningfully, it would be necessary to use the multi-markets approach.
3. Conclusions
Competition authorities and courts are well advised to use uniformly a multi-markets approach when they define markets in the context of two-sided platforms.Footnote 82 It would be desirable for the Commission to make this explicit with the reform of the Market Definition Notice.
The multi-markets approach is a logical and consistent application of demand-side substitutability to two-sided platforms as it naturally accounts for different substitution possibilities by the user groups on both sides of a platform. It is based on the economic primitives of the market and not on derived constructs such as an overall demand for an intermediation service, which depends on demand substitutability on each side of the platform as well as on the cross-group network effects linking the two.Footnote 83
While one might think of conditions under which a single-market approach could theoretically be feasible, the necessary conditions are so severe that it would only be applicable under specific circumstances. Moreover, recognising that a single-market approach might be applicable under certain conditions would create substantial risks that an authority or a court would adopt it erroneously. This does not mean that the interactions between the different user groups that are served by a two-sided platform are to be neglected; quite the contrary: they must be considered for the definition of the (multiple) markets as they may be crucial to evaluating demand-side substitutability on each side of the platform.
B. Markets without a Price
It is a widespread phenomenon that operators of two-sided platforms do not charge prices vis-à-vis one group of their customers: viewers may use free, ad-financed TV, consumers are not charged a (visible) price using e-commerce platforms such as Amazon Marketplace or Booking, and Google charges no search fee for using its search engine and Facebook no membership or usage fee.Footnote 84
1. Why Do Zero-Price Markets Exist?
A number of economic considerations can explain why this is a rational and sustainable business strategy. First, it is important to see that, in the case of transaction platforms, the platform may decide to levy the fee entirely on the merchant side. Thus, while the consumers do not pay a visible fee, they bear part of the fee or even the entire fee as the merchant will pass it onto them through higher prices. Second, a platform may want to subsidise one side, which may lead to negative prices.Footnote 85 However, such prices may be self-defeating if they attract unwanted types of users or are simply not feasible. In particular, ‘zero’ prices are often a feature of two-sided platforms on the side that exerts a positive cross-group effect and experiences a negative from the other. Many ad-financed platforms such as commercial TV or internet news portals have this feature. In this case, consumers incur an opportunity cost in the form of their scarce attention, which is partly diverted to advertising. Another instance is that consumers ‘pay’ with their data and the platform can use this data to improve services (which may even be a prerequisite for it to succeed) or offer alternative services that the platform or third parties can monetise (possibly with different consumers); in such instances it is possible but not always the case that consumers incur an opportunity cost for providing their data. Many ad-financed digital platforms benefit from the attention and data that consumers provide. The data they receive allows them to provide better-targeted ads, which is in the interest of advertisers and, sometimes, also consumers. Furthermore, some platforms choose a ‘freemium’ strategy, offering menus of contracts that include a base offer at ‘zero price’, or they offer a ‘zero price’ as part of a dynamic pricing strategy that builds up a sufficiently large user base to convince late arrivals that it is worth paying. Finally, ‘zero prices’ may be forced by regulation. For example, net neutrality regulationFootnote 86 can effectively restrict the freedom of internet service providers (ISPs)—which operate as two-sided platforms, enabling transactions between content providers and consumers—to charge content providers for the delivery of content.Footnote 87
2. Past EU Competition Practice: (Unfounded) Reluctance to Define ‘Viewers’ Markets’
A view on past EU competition practice reveals a certain reluctance on part of the European Commission to acknowledge that there can be a ‘market’ that deserves consideration even if the observable transactions do not involve a monetary price. In some merger cases from the 1990s involving TV broadcasters, the Commission left open the question of whether there is a ‘broadcasting’ or ‘viewers’ market’, arguing that:
all TV broadcasters compete against each other for audience shares. However, in view of the fact that there is no direct trade relationship between broadcasters of ‘free’ tv channels, on the ‘supply side’ and, viewers on the ‘demand side’, it might be argued that tv broadcasting does not constitute a market in the strict economic sense of this notion.Footnote 88
The Commission suggested that it might not be necessary for the purposes of competition law to consider the ‘viewers’ market’ a ‘market’ because, ‘[i]n any event, the audience shares in the TV broadcasting are a determinant factor for the success of the broadcasters in the TV advertising market and have, therefore, to be assessed at least in the context of this market’.Footnote 89 Thus, the Commission essentially reasoned that, as high shares in the market for TV viewers would translate into higher shares in the TV advertising market, it would in all probability not be decisive for the outcome of a merger case if the existence of a market for TV viewers were denied.
The underlying policy argument to legitimise this approach would seem to be that the interests that competition law is meant to protect are properly taken care of by focusing on those ‘sides’ of a platform market where the monetisation takes place. However, if we assume that competition law aims at protecting the economic interests of the consumers, ignoring the unpaid side is typically inappropriate. This can easily be illustrated with a view on commercial TV financed through advertising: if viewers dislike TV advertising (as evidence suggests), viewer demand will respond to changes in the level of advertising. A merger analysis would thus need to take this directly into account if it cared not only about advertiser surplus but also about viewer surplus, because the merger of two TV broadcasters that are close competitors on the ‘viewers’ market’ would give them leeway to increase their advertising volumes. The most straightforward and indeed necessaryFootnote 90 way to appropriately protect the economic interests of the consumers as viewers against potential negative cross-group network effects is to acknowledge the existence of a viewer market. Even if one finds instead that the viewers are neutral to advertising, a merger between two ad-financed TV channels is likely to affect the profitability of each viewer. This change in profitability per viewer affects the incentives of the merged entity to attract additional viewers. Therefore, a merger is likely to affect the content choice of the TV channels, which is likely to impact consumer welfare. Thus, ignoring the unpaid side in any case amounts to ignoring the economic interests of viewers.
Even if the Commission has now clarified its position accordingly (as will be explained subsequently), the question of the existence of ‘viewers’ markets’, which has been kept open for quite a while, holds an important lesson: earlier clarification by the Commission, even if not strictly necessary in the context of the aforementioned merger decisions, could have prevented uncertainties among market operators and contributed to steer practice in the right direction at Member State level.Footnote 91
3. On the Acknowledgement of Zero-Price Markets in the World of Digital Platforms
With the advent of the digitised world, the European Commission did acknowledge without hesitation that a ‘market’ may also exist where a product is offered without monetary remuneration by the users. Thus, the EU Commission assumed, for example, in its abuse cases against Microsoft that there were markets for streaming media playersFootnote 92 and for web browsers,Footnote 93 even though these products were typically provided free of charge to final consumers. The (then) Court of First Instance confirmed this position.Footnote 94 In the same vein, the Commission subsequently took for granted in its merger decisions Microsoft/LinkedIn,Footnote 95 Microsoft/Skype,Footnote 96 and Facebook/WhatsApp Footnote 97 that remuneration is not an indispensable characteristic for the existence of a market. Moreover, in the Google Shopping case, the Commission concluded that the relevant product markets were the market for general search services and the market for comparison shopping services,Footnote 98 notwithstanding that the use of these services is offered free of charge to final consumers. Regardless of whether one finds the Commission's theory of harm in the Google Shopping decision convincing, the case nicely illustrates how certain conduct on a zero-price market might harm consumers in ‘paid markets’: if a search engine ‘manipulates’ its algorithm to give priority to its own affiliates, this may impede access to (paid-for) consumer markets and reduce the competitive pressure on those markets.
While the issue of zero-price markets on one side of a (digital) two-sided platformFootnote 99 has so far not been discussed by the ECJ in a competition case,Footnote 100 there should be no doubt that the court will approve the Commission's practice. Indeed, the court has acknowledged the economic rationalities of two-sided platforms that underly zero-price markets,Footnote 101 particularly as it considered the concept of a ‘service’ within the meaning of Article 57(1) TFEU, which explicitly requires that it is ‘normally provided for remuneration’. Thus, the court argued, for example, that amateur athletes who participate in sports events without being paid by the organisers receive consideration as their sponsors are provided with publicity.Footnote 102
4. Conclusions
To appreciate fully business activities in the context of two-sided platforms and to do justice to competition law's purpose of protecting consumer welfare, the legal concept of a ‘market’ should not be interpreted as requiring a (visible) price to be paid by one party to the other. It is not sufficient to consider the activities on the ‘unpaid side’ of the two-sided platform only by way of including them in the competition law analysis of the ‘paid side’ of the platform.Footnote 103 Such an approach would exclude certain activities and the ensuing positive or negative effects on consumer welfare altogether from the radar of competition law. Instead, competition practice should recognise straightforwardly that there can be ‘markets’ for products offered free of charge, ie without monetary consideration from those who receive the product. While it is well understood that the supply of personal data and/or the attention to the platform can be regarded as consideration because it can be monetised by the platform, it is neither necessaryFootnote 104 nor even beneficial to transform this insight into a legal concept of ‘remuneration’.Footnote 105
Consequently, a ‘market’ as a concept of competition law should be understood as consisting of transactions between two or more parties, of which at least one acts for economic purposes.Footnote 106 The latter is apparent in cases where a product is provided for remuneration. Moreover, where a product is offered free of charge, it suffices to demonstrate that the activity is part of a broad or a long-term strategy of the platform operator to generate revenue. This definition of a ‘market’ is meant to exclude essentially (only) activities that involve the exercise of power by public authorities and philanthropic activities. In the digital sphere, the latter category would include in particular Wikipedia, non-commercial blogs, and non-commercial donation-based crowdfunding platforms. The relevant provisions of EU competition law can straightforwardly be construed accordingly.Footnote 107 While this is in line with the Commission's current practice, it would be a helpful signal to firms, but also national authorities and courts, for the Commission to explain and refine its approach accordingly.
C. Market Delineation and Homing Decisions
If two-sided platforms offer intermediation services, it is important to take into account the users’ homing decision: if they make discrete choices between the offerings provided by platforms (and possibly non-platform providers of substitute services) they single-home, whereas they multi-home if they may decide to consume multiple offerings.Footnote 108 In a media context, viewers are single-homers if in the relevant period they pick one of the offerings. For example, a person may watch television only for the news and decide which news show to watch. Such a person is a single-homer.
Investigating whether and at which costs users may multi-home is essential for market definition and the assessment of market entry barriers in the context of two-sided platforms. This is above all because even strong positive network effects do not lead to consumer lock-ins on their own but only in conjunction with coordination problems.Footnote 109 However, the problem of miscoordination can be mitigated in the case of multi-homing because in this case there is no first-mover risk involved if a consumer uses a newly entered platform.Footnote 110
1. Identifying Homing Patterns
Taking a closer look at what counts as a single-homing and what as a multi-homing decision, consider platforms matching transport services to travellers. For example, if a traveller needs transport from an airport to the city centre, this is clearly a discrete-choice problem in terms of which product to consume. However, when we speak of single-homing versus multi-homing in a platform context, we ask whether a consumer has to decide on the particular platform or provider before actually choosing a particular consumption plan. If a traveller decides which mode of transport and associated provider of this transport to choose before deciding on the particular consumption, then we talk of a single-homing traveller. For example, if a traveller who has installed the Uber app takes Uber as the starting point of her transport decision and checks for available rides on Uber, but for alternative transport possibilities only if dissatisfied with the available offers, we may want to classify her as a single-homing user.Footnote 111 Correspondingly, somebody who first checks for the availability of a ride by local train and considers alternative means of transport only if dissatisfied with the local train offer is classified as a single-homer. By contrast, a traveller who checks offers, for instance on Uber and other transport platforms, before deciding what to do can be considered a multi-homer in the sense that she uses the information services by all information providers.
In our transport example, to the extent that individual providers of transportation services can list on alternative platforms, these transport providers can reach travellers through multiple channels (eg this holds if a van service can list on Uber as well as on other transport platforms). Hence, the degree of multi-homing by travellers affects the substitutability of platform listing services from the viewpoint of a provider of transport services and, as a consequence, also the latter's homing decision. It could be that a stable pattern of single-homing/multi-homing on the two sides emerges.Footnote 112
Moreover, a platform may affect the homing decision on one side through contractual clauses as, for example, in the case of exclusivity clauses that tie a seller to an e-commerce platform. If contracts enforce single-homing on one side, an immediate implication is that services provided by platforms (and other undertakings offering substitutes) to this group of users belong to one market.
Finally, it must not be overlooked that homing decisions on one side may to a great extent depend on homing decisions on the other side. Let us consider for illustration an area in which two rival ride-hailing apps are available. If most drivers offer their services at any point in time on both apps, there are likely to be few gains for travellers from using both apps simultaneously; thus, travellers have weak incentives to be multi-homers. By contrast, if most drivers single-home (in a given time interval), it is more beneficial for travellers to multi-home – that is, to check for availability and rates of transport offers on both apps. To acknowledge this interdependence is crucial as homing patterns on the two sides may change over time and, therefore, market delineations may have to respond.
2. The Multi-homing/Single-homing Framework
Where users on one side of competing two-sided platforms single-home, these platform services are substitutes belonging to the same market. If, however, users on the other side multi-home, each platform provides monopoly access to its set of users on the single-homing side. Thus, for given user behaviour on the single-homing side, each platform acts as a monopolist vis-à-vis users on the multi-homing side. This suggests that there is a market for each platform regarding the intermediation service provided to the multi-homing side.Footnote 113
Based on this consideration, the European Commission assumed the existence of a market for app stores available on the Android mobile operating system, which is dominated by Google's app store. This rests upon the assumption that consumers are single-homers as they make a discrete choice to use a device based on Android's, Apple's, or another firm's operating system, while app developers tend to be multi-homers.Footnote 114
It is important to see, however, that the monopoly power on each such market may be mitigated through interaction with the other user group.Footnote 115 More specifically, it may be the case that large parts of the revenues generated on the monopolised side are passed to the users on the other side.Footnote 116
Moreover, in the Commission's practice, the described multi-homing/single-homing framework has not always been considered in the context of market definition. In Travelport/Worldspan, the EU Commission recognised that platforms that offer electronic travel distribution services through a global distribution system (‘GDS’) to travel agents (‘TAs’) and travel service providers (‘TSPs’) typically face multi-homing on the side of the TSPs and single-homing on the side of the TAs.Footnote 117 The Commission stated this feature as part of its general description of the market for electronic travel distribution services through a GDS.Footnote 118 The Commission only referred to the multi-homing/single-homing frameworkFootnote 119 when discussing potential theories of harm through non-coordinated effects and, therefore, assessing inter alia whether ‘the merger would allow the merging undertakings to use their strong market position downstream vis à vis TAs in order to increase prices vis à vis TSPs upstream (“vertical cross market effects”)’.Footnote 120
This illustrates that it is (at least to some degree) functionally interchangeable whether the nature of the users’ homing decision is taken into account at the stage of market definition or (only) when considering the actual or potential effects on competition of a merger or any other relevant market conduct of the platform. In both ways, a competition analysis can consider that this framework enables a two-sided platform to exercise significant market power vis-à-vis the multi-homing side. While the former way, ie the recognition of a market for each platform, is a straightforward and consistent reaction to this framework, the latter approach is more flexible. Hence, it may be regarded as preferable in cases without a clear-cut multi-homing/single-homing setting but where the homing decisions of the different user groups are essentially in line with this scenario and therefore confer at least a significant degree of market power vis-à-vis the side on which multi-homing is prevalent. However, competition practice should not rashly discard the option of acknowledging a market for each platform to do justice to a multi-homing/single-homing framework, because otherwise there is a risk that this characteristic will not be given appropriate weight in the course of balancing the various factors.
D. The SSNIP Test
European competition authorities (among others) use the hypothetical monopolist test to identify the substitutes to a given product or service that are part of the relevant market. The General Court regarded the ‘small but significant and non-transitory increase in price’ (‘SSNIP’) test as ‘a recognized method for defining the market at issue’. Yet the court emphasised that the EU Commission may also rely on ‘other tools … such as market studies or an assessment of consumers’ and other competitors’ points of view’.Footnote 121 The Commission has to make an overall assessment of indicative factors without assuming a hierarchy between different types of available evidence. In particular, there is no requirement to prioritise technical evidence over non-technical evidence.Footnote 122
According to the SSNIP test, the relevant market is defined as the smallest product group such that a hypothetical monopolist in control of this product group could profitably sustain a small but significant non-transitory increase in price.Footnote 123 To answer the question of whether a hypothetical monopolist could impose a price increase, one needs to ask whether such a price increase would be profitable. Hence, the issue is whether selling a smaller quantity at a higher price would be more profitable than selling the initial quantity at the initial price. Whether this is the case depends on how sensitively demand reacts to a price change, ie on the elasticity of demand.Footnote 124
1. Implementation Challenges: Feedback Effects and Demand-Side Substitutability with Two-Sided Platforms
In its Booking.com decision, the French Autorité de la concurrence relied on the SSNIP testFootnote 125 to define the product market on the side of the platform where intermediation services are supplied to the hotels. The competition authority started with the hypothesis that the relevant product market comprises the reservation of overnight stays in French hotels via online reservation platforms and online travel agencies.Footnote 126 The authority then, first, argued that the hotel-keepers did not consider other distribution channels such as the hotels’ websites, meta-search engines (including hotel comparison sitesFootnote 127) and search engines to be substitutes for the use of an internet reservation platform.Footnote 128 Second, the authority referred to a statement by Booking.com, which had argued that most hotels would not have sufficient means to ensure their visibility on the internet through meta-search engines and search engines.Footnote 129 Based on these considerations by the hotels and Booking.com, the competition authority concluded that a small but significant increase of the commissions charged by a hypothetical monopolist platform would not result in such a significant shift of demand to other distribution channels as to make the price increase unprofitable.Footnote 130 Thus, the authority saw its hypothesis on the definition of the product market confirmed.
It is remarkable that the Autorité de la concurrence applied the SSNIP test to the market on one side of the platform without (explicitly) considering interrelations with the other side of the platform, ie the market for intermediation services offered to consumers looking for a hotel. Since consumers on Booking.com are not directly charged by the platform, their decision is only affected by the hotel prices that prevail on the different distribution channels (taking service provision by the platform as given). The question then is whether an increase in commissions affects consumer choice through prices charged by hotels. Under wide price parity clauses, hotels may increase their prices in response to commissions by some amount, but they have to do so on all distribution channels on which they are active. Thus, in the present case it is unlikely that consumers’ substitution patterns are affected significantly (there may be fewer total bookings but the relative shares of the distribution channels are unlikely to change much). While, therefore, for the application of the SSNIP test the interdependencies between price effects on the two sides of the platform actually appear negligible, it would be desirable for a competition authority to state such an assumption explicitly. Moreover, the case nicely illustrates that the relevance of the said interdependencies may be tied to the particular contracting environment:Footnote 131 absent price parity clauses, the market may have to be defined more widely as consumers may substitute away from higher prices on Booking.com after an increase of commissions. Thus, in general, when applying the SSNIP test, taking into account the interrelation with the other side may lead to a different definition of the market.
When applying the SSNIP test to two-sided platforms charging both user groups, different ways of increasing prices can be considered: the hypothetical monopoly intermediary could be thought of as raising (1) the sum of prices while optimally adjusting the price structure; (2) all prices together while keeping the price structure fixed; (3) each of the prices separately, allowing the other prices to be adjusted optimally; or (4) each of the various prices while keeping the other prices fixed.Footnote 132 Using (4) as the default option,Footnote 133 if there are mutual positive cross-group effects, then a drop in user participation as a result of a price increase on one side generates feedback that further reduces participation. To apply a one-sided SSNIP test without taking such feedback into account effects risks defining markets too narrowly.Footnote 134
When the Higher Regional Court in Düsseldorf upheld a Bundeskartellamt decisionFootnote 135 to block a merger involving platforms for ticketing services, the court formulated a guiding principle that the SSNIP test was not ‘sufficiently conclusive in case of two-sided markets because it cannot adequately capture the feedback effects between different market sides’.Footnote 136 But this does not hit the nail on the head—the problem of capturing the cross-group network effects is not a problem specific to the application of the SSNIP test but arises in general when assessing demand-side substitutability in the context of two-sided platforms. While it is correct that the implementation of the SSNIP test needs adaptation and is significantly more difficult in the context of two-sided platforms, this is only an expression of the general challenges for an assessment of demand-side substitutability in the context of two-sided platforms.
The SSNIP test can be seen as a thought experimentFootnote 137 that helps to gain clarity in the application of demand-side substitutability; it may serve ‘as a framework … onto which qualitative evidence is applied (for example views on substitutability from consumer groups, industry analysts or firms that are informed by verified observations on previous experience)’.Footnote 138 Thus, the Bundeskartellamt concluded, with regard to the applicability of the SSNIP test to two-sided platforms, ‘[w]hat would be conceivable are surveys on the switching behaviour of platform users under certain modified overall conditions based on the SSNIP test's fundamental idea’.Footnote 139 In Facebook/WhatsApp, the Commission referred to a survey among advertisers, most of whom submitted that they were not likely to switch from search ads to non-search ads in the event of a 5 to 10 per cent price increase, and concluded that the market investigation supported a corresponding sub-segmentation of the online advertising market.Footnote 140
A look at the practical application of the SSNIP test logic by the European Commission demonstrates how it may be useful even without considering empirical evidence. In Visa International, the Commission defined the inter-system market based on an analysis of demand for payment instruments from both merchants and cardholders.Footnote 141 Thereafter, the Commission assumed, inter alia, that neither cash nor cheques should be considered substitutable for payment cards. To justify this assumption with regard to merchants, the Commission argued that:
such non-card payment instruments are not at all substitutable with cards, since the loss of revenue for merchants from ceasing to accept all cards would be far greater than the loss of revenue from increasing their general level of prices by the amount of any small but sustained increase in merchant fees for all cards.Footnote 142
This application of the SSNIP test logic shows, albeit only implicitly, the interrelation with the general considerations on demand-side substitutability in two-sided platforms and, more particularly, serves as an example of how demand-side substitutability may be asymmetric between two user groups of a (matching) platform.Footnote 143
2. Zero-Price Markets and Markets with Diverging Monetisation Models
Further considerations are needed in zero-price markets and in markets in which platforms employ different business models with respect to monetisation. If there is a zero price on one side of the market, the SSNIP test would need to consider an increase of the price in absolute terms (a percentage increase clearly will not do) or would require additional modifications according to which not price but product characteristics such as quality are varied to understand substitution patterns. In particular, if users are expected to react very sensitively to price, this does not necessarily imply that a broader market has to be defined. The reason is that a platform may not even consider charging consumers but rather aims at getting their attention by making attractive offers. Such a strategy may be highly profitable if a platform is able to convert attention into revenues made from a different user group.
Purely ad-financed platforms are a case in point. For example, applied to an ad-financed social network with an alleged dominant position, to understand substitution patterns the question is: by decreasing the quality of service offered by this social network, how much traffic would it lose (eg to other social networks or different offerings such as video streaming or online games)? Such an alternative test has been called SSNDQ (small but significant decreases in quality). In Google Android it was applied by the European Commission to consider the indirect constraint exercised by non-licensable smart mobile operating systems on Android and, therefore, to determine whether a separate market for licensable operating systems can be defined.Footnote 144 However, a quantification is challenging, as it is often unclear how to operationalise a certain quality reduction.Footnote 145
It also seems conceivable to substitute the relevant costs to be borne by the users (personal data and/or attention) for prices, an approach that has been referred to as SSNIC (small but significant non-transitory increase in (exchange) costs). Thus, with a view to ad-financed search platforms, it was suggested to consider ‘whether a market-wide five percent increase in the amount (or length, duration, etc.) of advertisements would cause search customers to substitute away to a different product’.Footnote 146 However, possible feedback on price and demand on the advertising market must not be ignored. A SSNIC test is therefore only meaningful without further complications if demand on the advertising market is almost perfectly elastic, ie if many customers on the advertising market are willing to buy advertising space at the same price.
Such implementation problems are arguably the reason why the Commission in the Google Shopping case did not consider the option to adapt the SSNIP test (at least not explicitly) but simply stated that ‘the SSNIP test would not have been appropriate in the present case because Google provides its search services for free to users’.Footnote 147 The Commission could rely in this matter on the General Court's adjudication since the latter had held that the ‘SSNIP test may also prove unsuitable in certain cases, for example … where there are free goods or goods the cost of which is not borne by those determining the demand’.Footnote 148 However, the SSNIP test applied to quality changes can be used as a thought experiment.
If undertakings with different business models offer substitute services, the application of the logic of the SSNIP test becomes particularly challenging. As an example, we consider dating apps for heterosexual users. These apps possibly offer substitute services (the degree of substitutability depends on user behaviour). Some platforms charge subscription prices for male and female users, while others offer matching services at zero prices and run an ad-financed business model with advertisers as the third side. Depending on the identity of the platform, a price increase or an increase of advertising volume for one group of users (male or female users) could be considered.
3. Conclusion
The SSNIP test serves conceptual clarity in the application of demand-side substitutability. Therefore, although it is often difficult to empirically implement the test in the context of two-sided platforms, it is a useful instrument for competition practice even if only applied as a thought experiment.
III. CONSIDERING CONCEPTUAL INTERDEPENDENCIES: MARKET DEFINITION, MARKET POWER AND CROSS-GROUP NETWORK EFFECTS
In order to assess the competitive position of two-sided (digital) platforms, the cross-group network effects between their different user groups must in particular be appreciated. Against this background, two aspects are especially significant to correctly understand and implement the role that market definition should play as an instrument to facilitate competition practice and, thus, were already assumed in the above analysis. First, market shares are less informative as a proxy for market power. Second, market definition must not conclusively determine the scope of a competition analysis.
A. Market Shares as a Proxy for Market Power: Mitigating and Aggravating Factors
With a view on two-sided platforms, market shares have a relatively low significance for the assessment of market power. Given a firm with a certain market share, network effects may aggravate or mitigate the market power that follows therefrom.Footnote 149
Network effects can mean that the ‘coordinated’ decisions of the economic agents have the consequence that it is not the platform with the highest-quality offer that dominates the market, but a different platform. If the latter is the incumbent platform and a higher-quality platform enters the market, the former may still prevail. As Shapiro and Varian nicely put it from the viewpoint of the incumbent platform, ‘[p]recisely because various users find it so difficult to coordinate to switch to an incompatible technology, control over a large installed base of users can be the greatest asset you can have’.Footnote 150
The entrant platform must overcome the problem that users have no incentive to switch if they expect most of the remaining users to remain on the established platform. If all users remain in the status quo unless unilateral switching to the new platform is more attractive, barriers to entry will arise owing to user miscoordination.Footnote 151 For one thing, a necessity for coordination may arise because users would otherwise forgo the strong positive direct network effects from which they want to benefit, for example when using social networking services. For another thing, the need for coordination may refer to the users on different sides of a two-sided platform who are linked through mutual positive cross-group network effects, as, for example, in the case of matching platforms. In order to convince users to switch in such a situation, the new platform may need to subsidise early users (in the case of a two-sided platform, on at least one side). The extent of such subsidisation represents the level of entry barriers.
If, on the other hand, a new platform (for example, based on its reputation acquired in other markets) influences users’ expectations in such a way that all potential users assume that the status quo will be replaced, there are no barriers to entry.Footnote 152 The challenge is, thus, to identify the cases where network effects work in favour of an incumbent firm (and thus constitute an entry barrier) and those cases where they work in favour of entrant firms.Footnote 153
Moreover, digital platforms frequently operate on emerging and dynamic markets characterised by rapid growth (in which many unattached users arrive) and short innovation cycles. As was acknowledged by the General Court in Cisco, in such contexts ‘high market shares are not necessarily indicative of market power’.Footnote 154 However, as this cautious statement rightly points out, one must beware of generalisations and cannot avoid a case-by-case analysis of whether the characteristics of a market such as new product developments, the presence of new consumers and the ‘timing’ of new product introductionFootnote 155 actually point to relatively low entry barriers.
Aggravating and mitigating factors can easily be taken into consideration where the law does not foresee specified thresholds of market shares that are meant to function as indicators for a certain degree of market power. Most prominently, Article 102 TFEU applies to ‘undertakings of a dominant position within the internal market’, and pursuant to Article 2(2) and (3) of the EU Merger Regulation the compatibility of a merger with the common market depends in particular on the question of whether the merger will result in the ‘creation or strengthening of a dominant position’. While market shares have for a long time been used by authorities and courts to implement the concept of ‘dominance’,Footnote 156 and the ECJ even established a presumption of dominance applicable to undertakings with a market share of 50 per cent or more,Footnote 157 it has always been clear that market shares are only one of a multitude of factors that have to be considered to substantiate that a firm dominates a market.Footnote 158
Where the law does, however, provide for specified thresholds of market shares that are used as an indicator of market power, the importance of cross-group network effects makes it clear once again that such provisions should not contain ‘hard’ thresholds. The law should always provide for ways—whether through the structuring of substantive law or through procedural rules—to prevent ‘false positives’ as well as ‘false negatives’.
The EU block exemption regulations are a case in point. Their application typically requires that the market share of an undertaking does not exceed a certain threshold.Footnote 159 But, for one thing, an undertaking that exceeds this threshold may still rely on an exemption pursuant to Article 101(3) TFEU, which, thus, functions as a catch-all provision for scenarios where the market share of a firm erroneously signals a certain degree of market power. For another thing, pursuant to Article 29 of Regulation 1/2003, the European Commission or the national competition authorities of the Member States may withdraw the benefit of an exemption regulation when they find that measures by undertakings that are covered by it have ‘certain effects which are incompatible’ with Article 101(3) TFEU. This, therefore, forms a procedural mechanism to remedy the over-inclusiveness of a market-share threshold.
Instead of a focus on market share, competition practice concerned about persistent market power should focus their efforts on detailed case-by-case analyses of current and future barriers to entry.Footnote 160
B. Market Definition Must Not (Conclusively) Determine the Scope of the Competition Analysis
In the application of EU competition law, no clear and consistent position has been evolved as to whether or to what extent EU competition law requires or allows the balancing of pro- and anti-competitive effects or the netting of welfare effects on different groups of market participants. The main underlying reason for this is arguably the difficulties—quantification problems and distributional appraisals—that come along with such balancing exercises and inter-group comparisons of welfare.Footnote 161 To give an illustrative example, in a leading decision concerning a system of collective resale price maintenance for the trade of books, the European Commission rejected the argument that those vertical restraints enabled the booksellers to provide ancillary services and, hence, benefitted the consumers. In its justification, the Commission argued that the system denied the consumer the chance to decide whether or not she liked to pay a ‘service charge’Footnote 162 and, thus, refused to engage in a discussion on the aggregate effects that such a coordination might have on consumer welfare. In line with this, the Commission also rejected the proposition that price maintenance would allow publishers to cross-subsidise books published in small print runs.Footnote 163
In the context of a competition analysis of two-sided platforms, the important point to note is that, when it is acknowledged that a weighing of different and diverging effects is allowed or even required where these effects relate to a single market, then it must be allowed or required in just the same way where it concerns cross-group network effects on different user sides of a two-sided platform that, following a multi-markets approach, belong to different markets. While it may in many instances be reasonable to focus an impact analysis on a defined market,Footnote 164 this does not mean that any effects outside the boundaries of a defined market must be ignored for the purposes of a competition analysis. Such an understanding would assign a legal meaning to market definition that is not appropriate to it. As mentioned above, the ECJ has stated this with desirable clarity in Cartes Bancaires, rejecting the General Court's view that the analysis of whether a restriction of competition within the meaning of Article 101(1) TFEU had occurred were to be restricted to the market defined on the ‘issuing side’ of the payment card system, neglecting effects on the ‘acquiring side’.Footnote 165 Moreover, in its adjudications concerning both Article 101(3) TFEUFootnote 166 and Article 2 of the EC Merger Regulation,Footnote 167 the General Court took the view that market boundaries must not define the effects that may be taken into account for the necessary efficiency analysis.Footnote 168 This confirms that the appropriate consideration of cross-group network effects between the different sides of a two-sided platform does not require the (inadequate) adoption of a single-market approach.
IV. CONCLUDING REMARKS
The thinking and practice of EU competition law is at various points structured by defining the relevant market. The gist of the approach is to identify the products and services offered by an undertaking and to understand substitute offers. While the concept of market definition may, as a matter of principle, be transferred to competition analyses of two-sided platforms, it needs to be amended and extended. Since two-sided platforms offer interrelated and often complex products, and since the interdependencies between their different user groups have to be considered, market definition is more challenging and prone to error or even abuse.
Therefore, market definition in the context of competition practice on two-sided platforms must combine adequate economic reasoning with flexible legal thinking. It is crucial to understand in each case the nature of the offered products and services as well as the interrelations between the multiple connected sides and markets. These interrelations must, however, be considered not only for market definition but also for the analyses of the effects a certain practice may have on competition in the defined markets. Thus, competition practice in relation to two-sided platforms also requires understanding the specific limits to the significance of market definition. In this context, it is important to see that calculated market shares are less informative as an indicator of a two-sided platform's market power.
Because of the objective of market definition to structure and discipline competition analyses by authorities and courts, the true challenge is not only to adapt the conventional ideas and tools in a formally correct and consistent way but also to provide clear statements and practical guidance that makes market definition foreseeable and manageable also in the context of two-sided platforms. While one may argue about implementation details in individual cases, the Commission has so far found convincing answers to the conceptual challenges posed by the rise of digital platforms. However, it would help to see more explicit and clearer statements on the necessary conceptual adjustments. With the reform of the Market Definition Notice, the Commission has, first and foremost, the chance and the responsibility to provide market players with more legal certainty about its own thinking in this respect. In addition, the reform presents an opportunity to guide the Member States’ courts and competition authorities, and possibly even to influence the future case law of the ECJ.
