Google, Google Shopping and Amazon: The Importance of Competing Business Models and Two-Sided Intermediaries in Defining Relevant Markets. Sébastien Broos* & Jorge Marcos Ramos† ABSTRACT We show that the Commission intends to adopt a flawed definition of the relevant market in the Google Shopping case. Using two-sided intermediaries’ theory, we show that only one market linking the two sides of Google, consumers and advertisers, should be defined. Therefore, competitive constraints should be evaluated on both sides and not only on consumers’ side. Moreover, we show that, because a single market must be defined for Google, competing business models such as those of Amazon also pose significant competitive constraints to Google. In sum, it is not clear that Google enjoys a dominant position if the market is properly defined. We propose that business model competition and two-sided intermediaries’ theory should be included in the market definition toolbox in new technology and dynamic markets. JEL Codes: D42, K21, L12, L41 Keywords: Competition Policy; Abuse of Dominance; Two-Sided Intermediaries; Business Models; Relevant Market
1. INTRODUCTION In November 2010, the European Commission (“Commission”) opened an investigation into Google’s alleged abuse of a dominant position in the “online search” market.1 Five years later the investigation is still ongoing. The original view that the case would be closed under the commitment procedure (Article 9) has been swept away by the issuance of a Statement of Objections (“SO”) indicating that the case will most probably2 end up with an infringement decision (Article 7).3 In return, the Commission’s allegations have shrunk. The SO, issued in April 2015, drops (or leaves in the drawer) the concerns regarding exclusivity obligations on advertising partners and portability restrictions of online advertising campaign data, and focuses on the more favourable treatment that Google allegedly gives to its price comparison service: Google Shopping. In a nutshell, the Commission, in this “Google Shopping case”, considers that Google is dominant in the market of “general online search” and leverages its position into the “comparison shopping market” by treating Google Shopping in a more favourable way * Liege Competition and Innovation Institute (lcii.eu); HEC-ULg; University of Liege. Email:
[email protected] † Liege Competition and Innovation Institute (lcii.eu); School of Law; University of Liege. Email:
[email protected] We are thankful to Andrei Hagiu and Julian Wright for letting us use one of their figures. We also thank Emilio Calvano and Michele Polo for the permission to cite their preliminary paper, Nicolas Petit and Axel Gautier for useful comments and the participants of the 2015 PhD Forum on Law and Governance in the Digital Era in Amsterdam. This research was funded through the ARC grant for Concerted Research Actions, financed by the French-speaking Community of Belgium. 1 See Press Release, European Commission – IP/10/1624, “The European Commission has decided to open an antitrust investigation into allegations that Google Inc. has abused a dominant position in online search” Available at http://europa.eu/rapid/press-release_IP-10-1624_en.htm 2 Commissioner Vestager has hinted that a commitment decision is not completely excluded. See Jean Leymarie, Margrethe Vestager: ‘Un accord avec Google est encore possible’, Oct. 22, 2015, FRANCE INFO. 3 Council Regulation (EC) No 1/2003 of 16 December 2002 on the implementation of the rules on competition laid down in Articles 81 and 82 of the Treaty OJ L 1, 4.1.2003, p. 1–25.
1
than Google Shopping’s competitors. According to the Commission, Google’s own algorithm mechanism should treat its service and those of its rivals in the same way. The aim of this paper is to highlight a number of market and economic realities that seem to suggest that the definition of the relevant market(s) proposed by the Commission is imperfect at best.4 Although defining the relevant market is nothing but a tool, the correct assessment of any anticompetitive effects of the conduct investigated depends on a proper market definition. To that end, this paper contends that a correct market definition should address the two-sided nature of Internet platforms5 as well as the breadth of each side. Further, such definition should not neglect the existence of different business models that can be evidence of healthy rather than absent competition. We believe that if the relevant market is correctly established, the dominance of Google is not apparent. Defining the relevant market of certain online services is not –a priori– self-evident. This is even more the case in the presence of multiple business models that can be one-sided or two-sided and that are employed simultaneously and within the same market by a single firm. Many would intuitively agree that UBER, a two-sided platform, operates in the same “city transportation market” as regular taxi drivers, irrespective of how these businesses are organized. Nevertheless, the idea that Google competes face to face with Amazon is less intuitively appealing. This paper tries to explain how the definition of a relevant market should be approached in such cases. To that end, we confront the business model of Google with those of Amazon and consider the competitive constraints that they pose to each other. We do not define an exact relevant market. This would require data to which we do not have access. Neither we claim that Google has not degraded rivals. What we do claim is that even if Google has “misbehaved” it does not have the dominant position required to trigger the application of Article 102 TFEU. . Indeed, what we contend is that where Google has the technical ability to degrade rivals that is, when Shopping ads appear, it does not have a dominant position. The reason is that, when market dynamics are properly understood, the role of Google is not that of an advertiser or a purveyor of search results but, more simply, that of a matchmaker. In the case at hand, Google’s main competitors are not necessarily only traditional advertisers or search engines but also other matchmakers. In other words, it may be that Google has a dominant position somewhere, but not in the case at hand. This paper thus provides the tools and benchmarks that the Commission should apply to correctly assess the relevant market in the Google Shopping case. The paper is structured as follows. First, we explain how Google, Google Shopping and Amazon work (II). Second, we clarify a terminological confusion surrounding two-sided intermediaries and explore the two-sided nature of Google’s services (III.A.). Third, we propose that in the present case two-sidedness requires a single market definition and that a market for “online search” does not exist (III.B.). Fourth, we show that, if Shopping ads are displayed, then “search” and Shopping are in the same relevant market (III.C.) and that Amazon, even operating under a different business model, also belongs to that market (III.D.). Fifth, we relate these considerations to the alleged bottleneck role that Google plays in providing consumers with access to online services (IV). Finally, we conclude (V).
4
The SO is not public. The factual information relating to the Commission’s probe is mainly retrieved from the press release, available at http://europa.eu/rapid/press-release_MEMO-15-4781_en.htm. 5 We use the terms “platform” and “intermediary” interchangeably.
2
2. HOW GOOGLE SEARCH, GOOGLE SHOPPING AND AMAZON WORK This section first provides reasons for our focus on Amazon. It then gives an overview of the business models of Google Search, Google Shopping and Amazon Services. Note that by “business models” we understand a choice that firms face in the way they organize their interactions on the market. Essentially, it is the choice between four forms of organization: vertical integration, two-sided platform, input supplier or reseller. See Figure 3 for a very clear diagram of these different forms. 2.1. Why Amazon? We could have chosen other competitors of Google –eBay for instance– to illustrate this article. As we suggested in the Introduction, Google has more competitors than may appear at first glance. In the first place, we chose Amazon for its economic strength and value;6 if it is indeed proven that it belongs to the same relevant market as Google then, clearly, Google does not enjoy a dominant position. Secondly, Amazon has in place a number of complex business strategies that allow for a relatively complete comparison of different competitive constraints: competition among platforms and competition between a platform and a reseller. Thirdly, and perhaps more on the anecdotic side, Google has explicitly identified Amazon as one of its main competitors.7 We believe that competition enforcement, although focused on consumer demand, should not disregard the constraints that firms themselves identify, even though they might be biased for obvious reasons. 2.2. Google Search and Google Shopping Google Search is a two-sided platform –a statement often asserted but rarely proven– and is the most famous product in Google’s portfolio. It provides links to consumers. There are two types of links, organic and sponsored. “Organic links” are provided for free, neither consumers nor websites pay for them. “Sponsored links”, on the other hand, are paid for by advertisers. More precisely, each time a consumer clicks on a sponsored link, the advertiser pays Google. The fee is determined using auctions. Advertisers bid for a keyword and the one bidding the most wins the auction and the sponsored link.8 A sponsored search ad mostly contains the name of the website and a brief description of it. Sponsored links and organic links are visually different (see Figure 1).
6
In 2014 Amazon’s global revenue was about 89 billion dollars. See http://finance.yahoo.com/q/is?s=AMZN&annual. 7 Eric Schmidt, The New Gründergeist, available at http://googlepolicyeurope.blogspot.be/2014/10/the-newgrundergeist.html,“But, really, our biggest search competitor is Amazon”. 8 The process is far more complex. For instance, the quality of the link plays an important role and the highest bidder will not necessarily win the auction if Google deems its quality is poor. Also, the highest bidder does not pay its bid but that of the second-highest bidder. For more information on sponsored search auctions, see Hal R. Varian, Position auctions, 25 INT. J. IND. ORGAN. 1163 (2007); Benjamin Edelman, Michael Ostrovsky & Michael Schwarz, Internet Advertising and the Generalized Second-Price Auction: Selling Billions of Dollars Worth of Keywords, 97(1) AM. ECON. REV. 242 (2007).
3
Figure 1. Screenshot of the results from a query on google.co.uk Google Shopping is also a two-sided platform operating a vertical search engine. As opposed to a general or core search engine such as Google Search, a vertical search engine only provides links for certain types of requests. Prominent examples of these are kayak.com for flights or booking.com for hotels. Google Shopping is, as its name states, specialized in online shopping. All links provided are paid for by advertisers who, as in Google Search, set prices through auctions. Advertisers pay Google per click on their ad. Importantly, Google Shopping ads display more information than typical sponsored search ads. A picture of the good is usually shown as well as its price. Google Shopping can therefore be seen as a premium version of sponsored ads.9 Finally, let us highlight that Google does not own any of the goods sold or listed on Google Shopping. Moreover, no sale actually takes place on Google Shopping. Rather, consumers buy on the merchant’s website after having clicked on its ad. Figure 1 shows a screenshot where the two products have been highlighted. To make the terminology as clear as possible, we will refer to Google Shopping as “Shopping”, to Google Search as “Search” and to both as “Google”. 2.3. Amazon Amazon.com (and its national equivalents) is a vertical search engine specialized in online shopping. There are roughly two types of products on Amazon.10 The first kind is products that Amazon sells directly to consumers. Here Amazon acts as a reseller: it buys goods from producers, sets prices and sells to consumers. The control over the main strategic variables will be crucial for our section on two-sidedness. The second kind of products are those owned by third-party sellers but displayed and sold through Amazon. This two-sided platform service is called Amazon Marketplace. Depending on the size of the seller, it either pays a fixed fee and a commission or only a commission. Sellers set the prices of their products. Figure 2 displays a screenshot of a product search on Amazon.
9
According to a report by Searchmetrics, in the second quarter of 2014, average cost-per-click is $1.63 for Shopping and $1.07 for sponsored links. Click-through rates were respectively 5.91% and 4.91%. SEARCHMETRICS, White paper: Google Shopping & Adwords. The 10 most important players, (2015). 10 There is actually a third type of product: ads. The analysis we perform can also be applied to this side of Amazon’s business. As of now, this is not a very important part of Amazon’s revenues and we ignore it. However, it would not affect our main conclusions.
4
We will refer to “Marketplace” when we write about Amazon Marketplace and to “Amazon” either when we refer to both or when we only refer to the reseller part. If the context is unclear, we will specify which one we are referring to. 3. SHOPPING AND SEARCH ARE IN THE SAME RELEVANT MARKET Article 102 TFEU prohibits the abuse of a dominant position. As the Courts have consistently reminded,11 the holding of a dominant position is not prohibited in itself, rather it is a prerequirement for the finding of an abuse. It is thus the abuse of that dominant position that it is outlawed.12 Foremost, the assessment of anticompetitive foreclosure requires a market to be defined: a relevant antitrust market in which the undertaking under investigation holds a dominant position. The purpose of market definition is thus to identify in a systematic way the competitive constraints that the undertaking under scrutiny faces.13 To identify the competitors that are capable of constraining the undertaking involved is fundamental to assess the degree of market power. According to the Relevant Market Notice, defining the boundaries of the relevant market requires an immediate recourse to demand side substitution. Supply side substitution plays a marginal role in market definition14 and its “proxied” through reference to the demand substitution effect.15 We propose a shift from the analysis of the Market Notice, which dates from 1997, to capture the dynamic nature of the industry under scrutiny. Supply substitution should play a more vivid role in defining the boundaries of the relevant market and not only play a competitive constraint role once market shares have been calculated on the basis of demandconsumer substitution. Otherwise, the character of convergence, where online shopping and search become almost indistinguishable, is lost. Importantly, we do not consider situations where Shopping ads do not appear on Google. This is a major point: if they do not appear, there is, by definition, no scope for degradation. We do not say that these situations do not matter or that Google does have a dominant position therein, but rather that such a scenario does not appertain to the case at hand. Therefore, we do not propose that Google is dominant nowhere but that it is not in the situation that occupies us here. In this section, we first provide what we understand is the correct departure to properly define the relevant market in the Google case: its two-sided nature. We then review the approach that the Commission has followed in defining markets with two-sided intermediaries. In the second place, following the most recent economic literature, we explain why only one market should be defined. It is out of the scope of this paper to address the geographical extent of the relevant market. Whether a European or national market(s) is defined does not change the nature of our conclusions. 11
See, among others, Case 322/81, Michelin v Commission [1983] ECR 3461, at §57 and Joined Cases C-395/96 P and C-396/96 P, Compagnie Maritime Belge and others [2000] ECR I-1365, at §85. 12 Unlike Article 102 TFEU, Section 2, the sibling provision in US antitrust law, also catches anticompetitive attempts of monopolization. Although the list of abuses is open ended, the Commission has prioritized its enforcement activity to tackling abuses that are exclusionary. These have the effect of foreclosing competitors out of the market or preventing their expansion therein. The Commission refers to this as an “anticompetitive foreclosure” effect, which can be the result of a number of practices: predatory pricing, refusal to supply, tying, rebates, margin squeeze, etc. 13 Commission Notice on the definition of relevant market for the purposes of Community competition law, at §2, (“Relevant Market Notice”). 14 Id., at §13. 15 Id., at §20.
5
3.1 Ambiguities related to two-sided platforms Inquiring first about the two-sidedness nature of different intermediaries is not a circumstantial consideration. To understand the competitive constraints that Google and Amazon pose to each other through the implementation of different “business models” requires understanding about how these work. It is necessary thus to identify which are twosided intermediaries, which are not, and then explain what implications this classification has on the definition of the relevant market. But first some conceptual questions are in order. 3.1.1 Terminology The careful reader will have noticed that in section 2, we defined Google Search as a twosided platform and not as a two-sided market. This may seem like a “how many angels can dance on the head of a pin?” type of question but this terminological ambiguity, which we are not the first to notice,16 can lead to confusion when one tries to define the relevant market if some firms are two-sided. First, using the expression “two-sided market” in relation to a single company implies that this company is a market in itself. For instance, asking “Is Google a two-sided market?”17 already presupposes that Google is a market. In turn, this misjudgement would place Google under severe scrutiny of the antitrust authorities who would -prima facie- consider Google dominant. The term “market” also implies that two-sidedness is a feature of the market (“‘Twosided’ markets have two different groups of customers that businesses have to get on board to succeed”)18 while in reality, two-sidedness is a characteristic of the firms or of their business models. A simple proof of this is the coexistence of two-sided and one-sided intermediaries in
Figure 2. Screenshot of the results from a query on amazon.co.uk
16
For instance “The term “market” was meant loosely and does not refer to how that term is often used in antitrust.” See David S. Evans and Richard Schmalensee, The Industrial Organization of Markets with TwoSided Platforms, COMPETITION POL’Y. INT’L. 150 (2007). 17 Giacomo Luchetta, Is the Google platform a two-sided market? 10 J. COMPETITION L. & ECON. 185 (2014). 18 David S. Evans, Antitrust Economics of Multi-Sided Platform Markets, 20 YALE J. ON REG. 325 (2003).
6
many markets, if not in the same firm. This was already noted by Rysman:19 “Of course, one-sided markets have intermediaries, too. […]. Strikingly, one-sided and two-sided selling strategies exist side-by-side at Amazon.com. For some products, like certain new books, Amazon (basically) buys at a wholesale price and sells for a retail price, which is a one-sided model. But for many other products, Amazon provides a web portal for a producer that sets the retail price that a consumer would see. As this distinction often depends on the decisions of the intermediary rather than on purely technological features of the market, it may be better to use the term ‘twosided strategies’ rather than ‘two-sided markets’.” The same holds true for other canonical examples of two-sided intermediaries. For instance, shopping malls,20 could compete with single-brand stores, which are not two-sided intermediaries but rather vertically integrated structures. Free-TV21 (“Free-to-air TV” or “FTA”) might compete with video on demand on movie nights but the VOD service is only a one-sided reseller. Software producers22 may be two-sided – OpenOffice needs to attract both consumers and developers – or one-sided – Microsoft employs the developers. The Commission has not been immunized against this terminological confusion. In the Microsoft / Yahoo! Search Business decision, the Commission correctly identified the Yahoo, Bing and Google search engines as two-sided platforms.23 However, in the most recent Access/PLG decision, the Commission considered whether the recorded music market should be viewed as a “two-sided market”.24 To avoid confusion and a biased approach to the definition of what constitutes a relevant market, we consider that competition authorities should use the terminology “two-sided intermediaries” rather than “two-sided markets”. This correct departure in the analysis will prove of help when explaining why, in the Google Shopping case, Google and Amazon should be considered as being in the same relevant market. 3.1.2 Defining nature of two-sided intermediaries These examples lead us to the second ambiguity: the contentious definition25 of two-sided intermediaries. Rochet and Tirole26 argue that the defining feature is the non-neutrality of the price structure on interactions. For instance, if an intermediary decreases its price by 10% on one side and raises it by the same proportion on the other side, the number of interactions between the two sides should change. This implies that Coasian bargaining is not possible and that the side whose price has increased cannot pass it through (completely) to the other side.27 Another strand of the literature28 highlights the importance of indirect network effects between the different sides interacting on the platform. 19
Marc Rysman, The economics of two-sided markets, 23 J. ECON. PERSPECTIVES 125 (2009). Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 1 J. EUR. ECON. ASS’N. 990 (2003). 21 Mark Armstrong, Competition in Two-Sided Markets, 37 RAND J. ECON. 668 (2006). 22 Jean-Charles Rochet & Jean Tirole, Two-Sided Markets: A Progress Report, 37 RAND J. ECON. 645 (2006). 23 Case COMP/M.5727 - Microsoft/Yahoo! Search Business, at §100. 24 Case COMP/M.6884 - ACCESS/PLG, at §21. Also parties to the proceedings seem to fall in this confusion. In Case COMP/M.6967 - BNP Paribas Fortis/Belgacom/Belgian Mobile Wallet JV the notifying parties claimed to operate in a two-sided market, at §15. 25 For an extended discussion, see Dirk Auer & Nicolas Petit, Two-Sided Markets and the Challenge of Turning Economic Theory into Antitrust Policy, 60 ANTITRUST BULL. 426 (2015). 26 See Rochet & Tirole supra note 22. 27 Lapo Filistrucchi, Damien Geradin, Eric van Damme & Pauline Affeldt, Market definition in two-sided markets: theory and practice, 10 J. COMPETITION L. & ECON. 293 (2014). 28 See Evans, supra note 18 and Armstrong supra note 21. 20
7
The multiplicity and breadth of these definitions has made the task of distinguishing onesided from two-sided platforms arduous. For instance, one would not consider shopping malls as two-sided under the definition of Rochet and Tirole29 but Rysman30 classifies them as such. Newspapers are two-sided according to Luchetta31 but not according to Anderson and Gabszewicz.32 We use the definition developed by Andrei Hagiu and Julian Wright in a series of recent papers.33 In accordance with them, the first characteristic of two-sided intermediaries is that they enable direct interaction between the two sides. It is the sides, not the intermediary, which control the most important variables of the interaction such as pricing, marketing efforts, etc. Secondly, it must be that both sides affiliate through the platform, that is, the interaction requires that each side makes some costly investment to join the platform. It can be explicit –a membership fee or the purchase of a required equipment– or implicit –an opportunity cost (going to mall A instead of mall B) or a transport cost (driving to the supermarket). This definition allows us to distinguish easily between two-sided intermediaries and more traditional business models (reseller, vertically integrated firm and input supplier, see figure 3). 4 A. Hagiu, J. Wright / International Journal of Industrial Organization xxx (2015) xxx–xxx Direct interactions
“Side” A “affiliation”
“Side” A
“Side” B “affiliation”
MSP
Sale of goods or services
Re-seller
Sale of goods or services
“Ownership” of side A (vertical integration)
V.I. firm
“Side” A
Input supplier
Sale of input
Sale of goods or services
“Side” A
Sale of goods or services
“Side” B
“Side” B
“Side” B
Fig. 1. MSPs vs. alternative business models.
Figure 3. Figure 1 from Hagiu and Wright (2015b). Different business models.34
optimistic expectations, and then extend the analysis to the case with pessimistic expectations. In each case, we consider what happens when the VI firm can use a bonus payment based on sales to incentivize the choice of effort by employees, and when the MSP can charge variable fees (or subsidies) to professionals based on sales in order to induce them to internalize spillovers.
arise because the more services/professionals are made available, the more clients will become aware of the firm/MSP (e.g., through word of mouth, reputation effects, or other sources of information and review) or the more likely a given client that is informed about the services available through the firm/MSP will be to become a repeat customer. In other words, we allow for positive agglomeration effects, contained in m(n). The variable ai captures the level of some costly, non-contractible, transferable action that raises the number of clients for the service offered by professional i. An example is the marketing and advertising of the professional services being offered. By “ transferable” we mean that control over this action can be transferred between the firm and the professional. Specifically, it is undertaken by the firm under the VI mode and by professional i under the MSP mode. A cost 12 a2i is incurred to provide this activity by the firm in VI mode and by the professional i in MSP mode.4 While the analysis becomes considerably more complicated, the results and intuition obtained are similar. The term θi measures the impact of ai on demand. We assume that θi is an i.i.d. random variable drawn from some distribution function with ! " support θ; θ and whose realization is private information known only to the professional i at the time ai is chosen. Assume θ N0. This private information can for example capture that each professional has a better idea of how much advertising in its services expands its client pool. The expected value of θ is denoted E(θ) and the variance is Var(θ). OLThe term x is ai spillover parameter, and captures the strength (and direction) of spillovers across professionals from their choice of activity ai. The term ā−i denotes the average level of activities across all services other than service i. Thus, the spillover affects the number of EV forETWORK clients service i whenCON the level of the activity ai differs from the average. We allow spillovers to be either positive or negative, but restrict attention to x b θ so that Di is always increasing in ai. As an example, consider hair salons: the activity could be the extent to which professionals' profiles and services are advertised. Salons that employ hairdressers (VI mode) make those advertising decisions themselves. Salons that rent out space/chairs allow each of their independent
The direct interaction requirement separates two-sided platforms from re-sellers and vertically integrated firms as in both setups each side only interacts with the intermediary. Affiliation sets the input supplier model apart because only the side buying the input actually interacts 3.1. The model with the supplier. There are N N 1 independent professionals. Each professional i = This makes ita continuum easy to classify Google Search and Shopping. Indeed, using 1,..., N definition supplies a unique service. There is of potential clients for each service. Each potential client incurs an “ affiliation” cost c traditional definitions, it is not clear that Google is a two-sided market. For example, Manne to be able to purchase from the firm or MSP. This can capture travel costs or other fixed costs involved in dealing with the firm or MSP. Assume each potential client of professional i's service obtains surplus above c, and so wishes to affiliate and purchase the service. The corresponding revenue generated per client is assumed to be exogenous and equal to π N 0.3 These assumptions allow us to focus on the key trade-offs between the two business models without introducing any pricing distortions on the buyer side. The number of clients for the service from professional i is given by
29
See Rochet & Tirole, supra note 22. See Rysman, supra note 19. 31 See Luchetta, supra note 17. Di ¼ mðnÞ þ θi ai þ xða−i −ai Þ þ γei ; 32 Simon P. Anderson & Jean J. Gabszewicz, The media and advertising: A tale of two-sided markets, in which implies that the revenue generated professional i is πDi. HANDBOOK OF THE ECONOMICS OF by ART AND CULTURE ; V . 1, 567–614 (Victor A. Ginsburgh & David Throsby, In the expression of Di above, m(n) is the baseline number of clients eds., 2006). See Auer & Petit, supra note 25, 20, for a comparison of many industries under the light of different for each professional and is allowed to depend on the number n of prodefinitions. fessionals who decides to join the firm/MSP. We assume m(n) ≥ 0 and 33 allow m to increase (weakly) in n to indicate that clients are more likely Andrei Hagiu, Merchant or two-sided platform?, 6 R . N E 1 (2007); Andrei Hagiu & Julian to come to the firm/MSP when more professionals are present (or when Wright,more Marketplace or reseller? 61 for MaANAGE . Snetwork CI. 184 (2015a); and Andrei Hagiu & Julian Wright, Multi-sided services are available). Thus, we allow cross-group effect to arise between professionals and client demand. This could platforms, INT. J. IND. ORGAN. Forthcoming, (2015b). 34 “MSP” stands for “multi-sided platform” and “V.I.” for “vertically integrated”. 30
3 These assumptions hold if professional i faces a downward-sloping demand function d(pi) for how much of its service is consumed by each client. Assuming professionals and the firm cannot commit to pi, so that the optimal pi under both modes is ∞ pm = argmaxp{pd(p)}, then provided ∫pm dðpÞdp Nc, all potential clients will consume and the revenue made on each individual client is π = p m d(p m ). Alternatively, p m could be exogenously determined, perhaps reflecting some industry norm or regulation, in which case π = pm if professionals face unit demand.
8 4
In an earlier version of this article we considered the case that ai represented the price of service i, in which case revenue is aiDi and we need to assume θi b 0.
Please cite this article as: Hagiu, A., Wright, J., Multi-sided platforms, Int. J. Ind. Organ. (2015), http://dx.doi.org/10.1016/j.ijindorg.2015.03.003
and Wright35 and Luchetta36 argue that it is not (or at least that it is unclear) because network externalities are only running in one direction namely, advertisers care about the number of consumers but consumers do not care about the number of advertisers. Nevertheless, under Hagiu and Wright’s37 definition, network externalities are not an essential characteristic. There is thus no ambiguity: Google Search and Google Shopping are two-sided platforms. There is clearly an interaction in the form of a click that is traceable to a consumer and to the ad placed on Google’s webpage. It is the seller and not Google that sets the price of the good. In turn, each side is also affiliated to Google. Both consumers and advertisers choose to use Google Search or Google Shopping instead of another search engine and therefore incur an opportunity cost. As stated above, Amazon is both a two-sided platform and a reseller. Its Marketplace is two-sided for the same reasons as Google Search and Google Shopping. There is an interaction in the form of a sale that is easily traceable, each side is affiliated to Amazon and it is the seller that sets the most important strategic variable: price. But Amazon is also a onesided reseller when it sells goods itself. Indeed, in that case there is no direct interaction between the producer of the good and the consumer. Moreover, Amazon as a reseller controls all the strategic variables such as price setting and advertising activities. 3.2 One or two markets for two-sided intermediaries? 3.2.1 Decisional Practice The decisional practice of the Commission has seen an involution towards defining two different markets where two sided intermediaries are involved. In the case concerning the acquisition of DoubleClick by Google, the Commission ventured into defining one relevant market for “online advertising intermediation activities”38 (Emphasis added). In the wellknown VISA case, the Commission acknowledged that the two-sided nature of “card payment systems” deserved the definition of one relevant market. The Commission could thus consider existing competitive constraints between different payment card schemes. The Commission did not however concede VISA’s claim that card payment systems competed in a larger “payment systems” market.39 The Commission has nevertheless shifted its approach and started defining different relevant markets where two-sided platforms are involved. Its position was clearly stated in MasterCard where it defined two relevant markets and strongly opposed the consideration that two-sidedness meant defining one relevant market: “Two sided demand does not imply the existence of one single joint product.”40 (Emphasis added). It is interesting to note that MasterCard appealed the decision and did not challenge the market definition before the General Court. However, it appears that while MasterCard overlooked in first instance the importance of such differentiation, it sought to remedy this before the Court of Justice. The ground was rejected: “the appellants have not directly challenged the General Court’s
35
Geoffrey A. Manne & Joshua D. Wright, Google and the limits of antitrust: The case against the case against Google, 34 HARV. JL. & PUB. POL’Y. 171 (2011). 36 See Luchetta, supra note Erreur ! Signet non défini.. 37 See Hagiu & Wright supra note 33 (2015a and 2015b). 38 COMP/M.4731 – Google/ DoubleClick, at §86. 39 For a discussion on this, see Section III.D. 40 Cases COMP/34.579 MasterCard, COMP/36.518 EuroCommerce and COMP/38.580 Commercial Cards at §257.
9
assessment in respect of that definition, namely the acquiring market”.41 In light of the above, it is safe to consider that the Commission has left the door open to defining one relevant market in cases where two-sided intermediaries are involved. It seems it will be for the parties to prove the pertinence of defining only one market. This pertinence is what we seek to prove hereunder. 3.2.2 An economic approach to the Google case As hinted above, the proper definition of a relevant market in cases involving two-sided intermediaries is a complex stumbling block for economists. The pertinent question to ask is whether there are two relevant markets or just one relating of the two sides.42 Some commentators have already discussed this riddle and both arguments in favour and against the pertinence of defining one or more relevant markets in the Google case have been made. Höppner considers that defining a single relevant market for a variety of multisided intermediaries cannot work because each side requires an independent substitutability analysis.43 On the contrary, Ratliff and Rubinfeld argue that a market for online search is not a meaningful market definition, as it does not capture the interdependence of both sides of the platform.44 The discussion is not without purpose. Considering Google or Amazon within independent markets as delineated by the boundaries of each side of their platforms could lead to misleading conclusions on market power and competitive constraints. In this line, and as others have before us,45 we understand that the most important factor to defining one relevant market is that an interaction between both sides takes place and is observable, be it on Google Shopping or Google Search. Because Google is a transaction intermediary, only one market should be defined. This implies that it does not make sense to talk about the “search market” or the “advertising market”. What Google sells is a link between the buyers and sellers. The observability of the interaction allows Google to charge advertisers a transaction fee each time their ad is clicked or their product is bought. This is in fact the core of Google’s business. The service that Google offers and is paid for is thus the intermediation that it provides. Consumers do not pay to search and sellers do not pay to advertise per se. Indeed, prices are not per impression but per click. There is a single price and this is paid when an interaction, a click, between the two sides occurs. This is an extremely important point because it helps clarify the difference between this type of advertising and traditional offline advertising.46 Let us take for instance the yellow 41
“In the present appeal, the appellants have not directly challenged the General Court’s assessment in respect of that definition, namely the acquiring market”, see C-382/12P, MasterCard v Commission, not yet published, at §159. 42 See Filistrucchi et al. supra note 27. See also Lapo Filistrucchi, Damien Geradin, Eric van Damme, Simone Keunen, John Wileu, Jan Boone, Tobias Klein & Thomas Michielsen, “Mergers in Two-Sided Markets - A Report to the NMa”, NETHERLANDS COMPETITION AUTHORITY, pp. 1-183, Available at https://www.acm.nl/download/documenten/nma/NMa_Two-Sided_Markets_-_Report_-_16_July_2010.pdf 43 Representing complainants against Google, see Thomas Hoppner, Defining Markets for Multi-Sided Platforms: The Case of Search Engines”, 38 WORLD COMP. 349, 352-356 (2015). 44 Under financing of Google, see James D. Ratliff & Daniel L. Rubinfeld, Is There a Market for Organic Search Engine Results and Can Their Manipulation Give Rise to Antitrust Liability, 10 J. COMPETITION L. & ECON. 517 (2014). Along the same lines see Lars Wiethaus, Google’s Favouring of Own Services: Comments from an Economic Perspective, 6 J. EUR. COMP. L. & PRAC. 506 (2015). 45 See Filistrucchi et al. supra note 27. 46 We do not propose that newspapers and other more traditional media do not compete with Google. More simply we do not know, but this point helps to show that our definition of the market is not boundless and does not imply that Google competes with anything and everything.
10
pages, which have a structure that is similar to Google. It is a list where free and paid items coexist. A major difference however is that on Google, sellers pay for an interaction while in the yellow pages, they pay whether consumers visit their store and buy later or not. There is usually no way to relate the yellow pages’ ads and the sale. This also has an impact on consumers who cannot buy immediately, in contrast to the online world. Further, unlike what the Commission has stated, the market in which Google Search operates is therefore not “general online search”. Such a definition would imply that a market can exist without price, transaction and profit.47 In the case of search intermediaries, defining two separate relevant markets would lead to meaningless definitions. In this sense, the Commission has itself been reluctant to identifying such thing as “a viewer market” in cases relating to free-to-air TV, where the “commercial relationship only [exists] between the program supplier and the advertising industry”.48 A position that has nevertheless not been consistently endorsed: “It can be left open whether the viewer market constitutes a distinct relevant product market in the strict economic sense of the notion”49.50 Moreover, a “search” definition would completely set aside the feedback effects that are typical of many two-sided intermediaries. More than typical, in the case of Google feedback effects are a necessary condition to its viability. Defining the relevant market as (search) advertising would not make sense either for the same reasons.51 Even if a billion consumers look at an ad, no price is paid. The same holds true for Google Shopping. It now remains to be seen whether the unique markets that should defined for Google Shopping and Google Search are identical. 3.3 Search and Shopping operate in the same relevant market The Notice on Market Definition states that “a relevant product market comprises all those products and/or services which are regarded as interchangeable or substitutable by the consumer by reason of the products' characteristics, their prices and their intended use”. Thus, in a streamlined fashion we should consider whether Search and Shopping impose competitive constraints on each other. In other words, whether customers buying these goods, be they consumers or advertisers, see them as effective substitutes. These are the questions that we try to answer in this subsection following the price, characteristic and intended use framework. But first, let us dismiss what could be an intuitive tool to assess these issues: the SSNIP test. Indeed, besides traditional issues with the SSNIP test in two-sided markets52, two other issues make it difficult to use in this particular setting. First, one side, consumers, does not pay anything and firms have to compete on other attributes which are less quantifiable, such as quality. There is no way for the SSNIP test to capture that aspect. Second, there are literally billions of prices, one for each keyword paid by advertisers, and they are constantly changing. Billions of prices mean billions of different situations and it is probably not feasible to use a SSNIP test on such a large scale and obtain meaningful conclusions. Moreover, because each 47
See also Miguel Sousa Ferrero, Ceci n’est pas une marché’: gratuity and competition law, REV. CONCURRENCES 1, 10-13 (2014). 48 Case COMP/M.2876 - NEWSCORP/ TELEPIU’, at §41. 49 Case COMP/M.878 - RTL 7, at §7. 50 For an exhaustive discussion of the decisional practice of the Commission regarding “free markets”, see Sousa Ferrero supra note 47. 51 See Manne & Wright supra note 35, calling out the appropriateness of using such a concept, at p. 24. 52 See Filistrucchi et al. supra note 27; David S. Evans & Michael Noel, The analysis of mergers that involve multi-sided platform businesses, 4 J. COMPETITION L. & ECON. 663 (2009) and Lappo Filistrucchi, A SSNIP test for two-sided markets: the case of media, 34 NET Institute Working Papers, (2008) available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1287442.
11
keyword is sold through an auction, using the SSNIP might lead to the extreme conclusion that each keyword auction is a market in itself.53 We turn now into the precise features of each side and the characteristics of its interaction with Google and the other side. 3.3.1 Consumers Do they see search ads and Shopping ads as substitutes? As we mentioned above, Google Shopping and Google Search are not always in the same relevant market. Indeed, in many instances, while “sponsored links” are displayed, “Shopping links” are not. It must therefore be that in these cases Google considers that Shopping links are not useful to consumers. Either they would not click and precious space would be wasted on the page, or they would and results would be irrelevant, leading to displeased consumers. The reason why they do not appear does not matter, the important point is that if Shopping ads are not displayed, there is, by definition, no scope to degrade anyone. It is therefore only worth considering those instances in which both “sponsored links” and “Shopping links” appear together. A first argument in favour of substitution is that Google itself knows that the intended use of consumers will lead them to click on both types of ads. Otherwise, it would be economically unsound to display both kinds and lose advertising space that Google could sell to other advertisers. As a matter of fact, both types of ads are also answering the same intrinsic need of consumers: the search for a website selling a particular good. A consumer clicking on a “sponsored add” or a “Shopping add” expects the same result: a good that corresponds to his search. Moreover, the links appear side by side on the page. Google has often argued that “competition is a click-away” because consumers can reach competing search engines. That argument may seem unconvincing because a minor switching cost remains, after all, a switching cost. Nevertheless, here the switching cost is zero. Consumers do not even have to click to go from one the other, a simple glance is enough. Although a survey and economic analysis thereof are out the scope of this paper we believe it is difficult to argue that links placed side by side on the same webpage are not seen as substitutes by consumers when we consider that they are designed to answer the same need. This is further evidenced from the advertisers’ side. 3.3.2 Advertisers Do advertisers also see Shopping and Search as substitutes? As argued above, we do not pretend that Google Shopping and Search are always substitutes. Rather, Shopping is aimed at retailers54 while Search is aimed at all types of advertisers. It seems that for retailers the two types of ads are indeed substitutes. First, the reasons for which consumers see Shopping and Search as substitutes have their counterparts for advertisers. The group of consumers that advertisers can reach have the same intent in both cases. Indeed, Shopping and Search ads reach consumers who have been identified by Google as having the same intent: buying a good corresponding to their search criteria. A retailer willing to sell should not see a difference as they are targeting the same consumer base. 53
NIGEL PARR, ROGER FINBOW & MATTHEW HUGHES, UK MERGER CONTROL: LAW AND PRACTICE, (2005) at p.371. 54 See Google, About Shopping campaigns and Product Listing Ads, https://support.google.com/adwords/answer/2454022?hl=en
12
A caveat to this might be that “Shopping ads” are, in terms of characteristics, superior to “Search ads”. To put it in more economics terms, they are vertically differentiated. Indeed, “Shopping ads” display a picture, provide more information, are more expensive and have better click-through rates. Nevertheless, these characteristics do not necessarily put them in different relevant markets because they are targeting the same set of consumers.55 The Commission has not addressed this particular issue yet and has not sub-segmented the “online advertising” market.56 Although the Commission has left the question open, recent data seems to show that retailers are indeed substituting. According to a report by Adobe, sponsored search spending among retailers has declined by 6% between December 2013 and December 2014 while Shopping ads spending has increased by 47%57 during the same period.58 Simultaneously, total search engine marketing spending by retailers increased by 7%.59 3.4 A single market for Shopping, Search and Amazon As we underscored above,60 market definition is not a static given assumption and should thrive to incorporate dynamic arguments. The key question we ask is: do different business models necessarily imply different markets? The answer is no. Amazon’s main business model is that of a reseller: it buys goods and sells them to consumers. The business model of Google, be it Shopping or Search, is to sell a connection between consumers and retailers. We argue that these are simply two ways to provide the same service and therefore, that all these products compete in the same market. A fortiori, this is also true of Marketplace. Note once again that we do not propose that Amazon and Google are always competitors but only that, if Shopping ads appear, they are. 3.4.1 Consumers We start with consumers, the least contentious part. We use the same grid of analysis as for Google: characteristics, price and use. Whether they visit Amazon or Google, consumers search for a particular good and see an ordered list of items that link to retailers’ goods. These items usually display some descriptive information (price, rating, availability, etc.) and a picture. This is very similar to the structure of “Shopping ads” although the latter are usually smaller. From consumers’ perspectives, the two products thus seem to have the same characteristics. As on Google, the price of the service is zero for consumers. Finally, the conditions of use are very similar. Both Google and Amazon are used for the same purpose: finding an appropriate good. But a more important point is that the sequence of actions that consumers perform on Amazon and Google is exactly the same. A consumer searches for a good, sees a list, clicks, moves to another webpage, buys and leaves the 55
See Adrian Majumdar & Richard Murgatroyd, Looking beyond market shares: the theory, evidence and meaning of closeness of competition in the manufacture, wholesale and retail of fast-moving consumer goods in South Africa and Europe”, RBB Economics, 2009. 56 Cases COMP/M.7217 - Facebook/Whatsapp, at §79 and Case COMP/M.7023 - Publicis/Omnicom. 57 The big difference in the numbers is simply due to differences in absolute values of spendings. 58 See Global Digital Advertising Report, Adobe Digital Index Q4 2014, available at http://offers.adobe.com/en/na/marketing/landings/_64058_q414_digital_advertising_report.html 59 Idem. This data is for the US but we cannot however come up with a reason why advertisers would behave differently in the EU. 60 See Section III: Introductory remarks.
13
website. This is a particularly important point because it helps us distinguish the market of Amazon and Google from others’. For instance, “display ads” do not fit in this framework because there is a lack of willingness to buy on consumers’ part and there is no ordered list. Offline shopping does not fit in this framework either.61 The only major difference we could find between the two websites is that on Amazon, consumers pay via Amazon whereas on Google, they have to pay via the retailer’s system. There might thus be a trust issue with some consumers not willing to provide payment information to the retailers present on Google. The auction system however mitigates that concern. Indeed, to be displayed in Google, retailers have to win the keyword auction. This implies, that, except for exotic goods, most retailers are sizeable and well-known companies with whom consumers are used to transact and trust issues should remain marginal. From consumers’ perspectives, it therefore seems difficult to consider Amazon and Google as belonging to a different market. 3.4.2 Retailers This is the main stumbling block. Here, we must separate Marketplace from “reseller Amazon” because each has its peculiarities. We use again the characteristics, price, use framework. 3.4.2.1 Marketplace Marketplace is a two-sided platform and therefore more easily comparable to Google. From the point of view of characteristics, the two goods are similar. They are both a link in an ordered list queried by a potential consumer who can easily buy. As already highlighted before, the format of the link is very similar to that of Google Shopping, which itself is a sort of premium version of Google Search. The price is also similar although in a more intricate way. On Amazon, retailers pay a fixed fee and a variable fee each time a consumer buys. On Google, they pay each time their ad is clicked without certainty that this will turn into a sale. But if they are rational, they should be indifferent between different ways of selling. If a retailer believes that one in a hundred clicks leads to a sale, then it must be that the price of an ad is a hundred times smaller than the variable fee on Amazon. In other words, yes, observed prices are different but prices per sale are the same. Suppose for instance a 1/100 sale rate for an ad again and that an ad costs €0.01. Then, it must be that the variable fee of Amazon is €1. Prices are different (€0.01 v. €1) but prices per sale are the same (€1).62 What about the use? The use is to capture consumers looking for a good to buy online. It does not make a difference whether the consumer comes through advertising put on Google or a link on Amazon, as long a sale arises. The only distinction in use is, as already mentioned for consumers, in the payment system. On Marketplace, sellers have the benefit of being able to use Amazon’s payment system while if they use Google, they have to develop their own.63 We can therefore safely conclude that Marketplace, Shopping and Search are substitutes and therefore, belong to the same relevant market. 61
This does not mean that offline shopping is necessarily out of the relevant market. More simply, we do not know but usage does not point in that direction. 62 Note that including the variable fee does not change the result; it just makes the point less clear. The computation is only valid for risk-neutral firms. Otherwise, prices per sale will be different but “perceived prices” will be the same. This does not change the result. 63 They can also use a third-party system such as PayPal.
14
3.4.2.2 Reseller For Amazon reseller, the case rests upon differences in business models. In one situation, retailers pay for a click or a sale, while in another, they directly sell their good to Amazon. Does that suffice to imply different markets? According to the Commission, existing different business models are an indication of different relevant markets.64 In the broadcasting industry, merger control at the Commission and National level shows the importance that business models have in defining the relevant market. Broadly speaking TV broadcasters operate pay-TV and/or free-TV businesses. The main question is whether pay-TV and free-TV are considered as different relevant markets (for competition assessment purposes) or rather should be defined as an all-TV market. Put another way, the underlying and fundamental consideration is whether these business models (pay-TV and free-TV) are different ways of financing the same activity, i.e. broadcasting television services, or on the contrary are representative of different relevant markets. In BSky/KirchPayTV65, News Corp/Premiere66 and Antena3/La Sexta67 the Commission justified its view that pay-TV constituted a separate relevant product from the free-TV market on in the different financing mechanisms under which each business operated. “While, in the case of advertising-financed television, there is a trade relationship only between the programme supplier and the advertising industry, in the case of pay-TV there is a trade relationship between the programme supplier and the viewer as subscriber. In view of these trade relationships, the conditions of competition are accordingly different for the two types of television.” 68 (Emphasis added) Interestingly the Competition and Markets Authority of the UK (“CMA”) has taken a divergent approach. In BSky/ITV the CMA defined one relevant market –the market for allTV69– comprising different business models: pay-TV and free-TV; and, unlike the Commission, identified the two-sided nature of the business models.70 In turn, to understand why a FTA broadcaster could be a competitive constraint to a pay-TV broadcaster, it is necessary to shift the focus. That is, to analyse how “competition for viewers” takes place as this was the source of revenue for both pay-TV and free-TV, rather than on the particular sides of the trading relationships. This tension between the Commission and the CMA in selecting the role and extent that business models have when defining the relevant market proves of major importance in the Google case. On the other side of the Atlantic, the recently cleared merger acquisition of Orbitz by Expedia points in this direction. According to the DOJ, the fact that travel service providers have actual alternative ways to attract customers and obtain bookings is proven by the rapid evolution of ways of doing business.71 For instance, former “pure” information intermediaries like TripAdvisor are now entering into the online booking market. These dynamics brought 64
See Filistrucchi et al. supra note 27. They review some of these decisions but their focus is on the two-sided nature, rather than on competition between business models. 65 Case COMP/JV.37 – BSky/KirchPayTV. BSky was a broadcaster of pay-TV channels in the UK wanting to acquire joint control of KirchPay TV, a pay-TV operator in Germany and Austria. The Commission cleared the transaction subject to commitments. 66 Case COMP/M.5121 – News Corp/Premiere. 67 Case COMP/M.6547 – Antena 3/La Sexta, at §16. 68 Case COMP/JV.37 – BSky/KirchPayTV, at §24. 69 OFT in BSky/ITV, Dec. 14, 2007, at §§4.30-4.31. 70 Although we believe that the two-sided analysis of both models is incomplete, it is not our intention to discuss this here. 71 http://www.justice.gov/opa/pr/justice-department-will-not-challenge-expedias-acquisition-orbitz
15
about by the rapid evolution of business models should adjust the prism under which competition authorities define the relevant market and thus the complete picture of the intensity of competition. Economic theory also seems to disagree with the idea that different business models imply different relevant markets. In a recent paper, Calvano and Polo72 show that starting from the exact same conditions and characteristics, two firms can endogenously choose different business models, not to be in two different relevant markets, but rather to relax competition within the same market. Different business models can therefore be a proof of the existence of competitive constraints rather than a proof of its absence. These are the competitive constraints market definition seeks to unveil.73 Similarly, in a series of papers,74 Hagiu and Wright show that even a single firm selling to the same consumers can decide to use different business models or organization modes. For instance, they argue that a firm should privilege the two-sided mode when it does not have a lot of information about the goods sold, for example when there are many varieties. The idea that different business models can compete in the same market has also attracted the attention of business strategy scholars. Their inquiry is not one of market definition but of whether a firm with two business models can survive. In their view, dual business models within a firm illustrate the fear of self-cannibalisation and competition where one business model grows at the expense of the other. It is competition intra-firm, business model against business model, which has led strategy scholars to develop organizational solutions to manage the competitive constraints in their favour. The takeaway is that competition between different models is a market reality and solutions have been proposed strategists such as Burgelman75 or Markides76. In turn, Porter has even warned about using the term business models in the Internet for it misguides the approach to competition in online services by unfortunately “encouraging managers to view their Internet operations in isolation from the rest of the business, lead[ing] to simplistic approaches to competing using the Internet and increase the pressure for competitive imitation.”77 Besides these business models issues, the usual characteristics, price, use trichotomy could be used once again. We do not perform the tedious analysis because the arguments are very similar to those developed for Marketplace. Nevertheless, there is a seminal ability to be mentioned. The targeting capabilities of Google and Amazon are comparable and high performing. This ability to target the right consumers at the right time makes them interchangeable in the eye of sellers. This sets Google and Amazon apart from their offline counterparts and other forms of advertising.78 In this regard, the incorporation of business dynamics into competition analysis should not deter agencies. The qualitative nature of this analysis should however remain as scientific and objective as possible. It should be based on reliable evidence, and does not entail disregarding quantitative tools. In fact, as the General Court has already held in this respect “there is no need to establish any hierarchy between technical and non-technical evidence”.79 In the 72
Emilio Calvano & Michele Polo, Strategic Differentiation by Business Models: Free-to-Air and Pay-TV’s, (2014), available at http://www.cresse.info/uploadfiles/2015_pa2_p4.pdf 73 See Relevant Market Notice, “The purpose of market definition is thus to identify in a systematic way the competitive constraints that the undertaking involved faces” at §2. 74 See Hagiu & Wright supra note 33 (2015a) and Andrei Hagiu & Julian Wright, Marketplace or reseller? 61 MANAG. SCI. 184 (2014). 75 ROBERT A. BURGELMAN & LEONARD R. SAYLES, INSIDE CORPORATE INNOVATION, (1986). 76 Constantinos Markides & Constantinos D. Charitou, Competing with dual business models: A contingency approach, 18 ACAD. MANAG. EXEC. 22 (2004). 77 Michael E. Porter, Strategy and the Internet, 79 HARV. BUS. REV. 62 (2001). 78 Avi Goldfarb, What is different about online advertising? 44 REV. IND. ORGAN. 115 (2014). 79 Case T-342/07 Ryanair Holdings plc v Commission [2010] ECR II-03457, at §136.
16
absence of a public decision, the Expedia/Orbitz deal cleared by the DOJ, and the decisional practice of the CMA are on the right track. Fast changing business models are becoming an evident source of substitution that should be considered when defining the relevant market, rather than at the later stage of the competitive assessment. The impracticability of quantitative analysis in these scenarios should not deter competition agencies from resorting to qualitative evidence. 4. CONSUMER’S PATH TO SELLERS Some scholars80 have argued that Google uses its dominant position on the search market to prevent consumers from reaching competitors’ websites. We do not find this argument convincing. As we have shown in the preceding section, this market definition is flawed, and hence any conclusion on Google’s dominance in the “online search” market would be misleading at best. However, we believe it is still important to address whether Google is a gateway to the Internet or, to use an offline analogy, whether Google controls to road to competitors’ store. Indeed, one might be tempted to reach this conclusion from Google’s “market” shares in search (over 90% in many EU countries).81 However, search engines are just one way of accessing websites. According to data from Forrester Research, “in the third quarter [of 2014], 39% of U.S. online shoppers began researching their purchases on Amazon and only 11% started on search engines like Google.”82 Indeed, many consumers know Amazon’s URL by heart, use browsers’ autocomplete functions or have Amazon in their bookmarks. Even worse for Google, many consumers go directly to their competitors through branded apps and branded websites on their smartphones.83 This is corroborated by data from Alexa84, a web analytics tool. The share of visitors on Amazon.co.uk that originated from a search engine fluctuated between 11 and 16% in 2015.85 These figures are an upper bound for Google because they also take into account clicks from Bing, Yahoo and other search engines. Results for other Amazon stores are similar. Moreover, not all queries on search engines are equal. Some are what we call “brand terms”.86 For Amazon, these would be terms such as “Amazon”, “Amazon France”, “Amazon uk”, etc. According to data from Alexa, the five top key search terms leading to Amazon.co.uk are all brand terms and represent about 24% of search queries. For Amazon.fr they represent about 40%. We can safely assume that an overwhelming majority of the consumers who queried these terms would have found their way to Amazon even if Google had removed it entirely from its search engine. It therefore seems that Google, although it might indeed be degrading rivals, does not own the road to competitors’ stores. 80
See Frank A. Pasquale III, Dominant search engines: an essential cultural & political facility in THE NEXT DIGITAL DECADE, ESSAYS ON THE FUTURE OF THE INTERNET, 401-418, (Berin Szoka & Adam Marcus, eds. 2010); Benjamin Edelman, Bias in Search Results? Diagnosis and Response, 7 INDIAN J.L. & TECH. 16 (2011) and Benjamin Edelman, Does Google leverage market power through tying and bundling? 11 J. COMPETITION L. & ECON. 365 (2015). 81 See e.g. Case COMP/M.7023 - Publicis/Omnicom at §§468 and 503. 82 Rolfe Winkler & Alistair Barr, Google Shopping to Counter Amazon Testing ‘Buy Now’ Button, Other Enhancements to Build Online Commerce Site, Wall Street Journal, Dec. 15, 2014. 83 In 2013, more than half of consumers purchase directly through branded apps and branded websites. See Nielsen Report, Mobile Path to Purchase (November 2013), https://ssl.gstatic.com/think/docs/mobile-path-topurchase-5-key-findings_research-studies.pdf 84 Note that Alexa is owned by Amazon. However, the same result can easily be verified for other competitors. 85 See http://www.alexa.com/siteinfo/amazon.co.uk as data retrieved on Oct. 28, 2015. 86 Thomas Blake, Chris Nosko & Steven Tadelis, Consumer heterogeneity and paid search effectiveness: a large-scale field experiment, 83 ECONOMETRICA 155 (2015).
17
5. CONCLUSION This paper does not pretend to give a decisive nominal definition of a relevant market nor to bullet proof Google against a dominance accusation. Rather, it puts on the table a number of considerations that we believe are missing in the intersecting literature of antitrust and economics on the Google Shopping case. These seem unaddressed in the current proceedings in the light of the public information available. In this respect, business model substitutability has become an evident element in the competitive process and thus should not be overlooked at the stage of market definition. Recalling Schumpeter, effective competition “comes from the new commodity, the new technology, the new source of supply, the new type of organization—competition which commands a decisive cost or quality advantage and which strikes not at the margins of the profits and the output of existing firms, but at their foundations and their very lives.”87
87
JOSEPH A. SCHUMPETER, CAPITALISM, SOCIALISM AND DEMOCRACY (1942). 18
