Market Definition in the Google Android Case

http://lib.ulg.ac.be http://matheo.ulg.ac.be

Market Definition in the Google Android Case

Auteur : Lamesch, Joé Promoteur(s) : Gautier, Axel

Faculté : HEC-Ecole de gestion de l'ULg

Diplôme : Master en sciences économiques, orientation générale, à finalité spécialisée en Economic Analysis and Public Governance

Année académique : 2015-2016

URI/URL : http://hdl.handle.net/2268.2/1812

Avertissement à l'attention des usagers :

Tous les documents placés en accès ouvert sur le site le site MatheO sont protégés par le droit d'auteur. Conformément aux principes énoncés par la "Budapest Open Access Initiative"(BOAI, 2002), l'utilisateur du site peut lire, télécharger, copier, transmettre, imprimer, chercher ou faire un lien vers le texte intégral de ces documents, les disséquer pour les indexer, s'en servir de données pour un logiciel, ou s'en servir à toute autre fin légale (ou prévue par la réglementation relative au droit d'auteur). Toute utilisation du document à des fins commerciales est strictement interdite.

Par ailleurs, l'utilisateur s'engage à respecter les droits moraux de l'auteur, principalement le droit à l'intégrité de l'oeuvre et le droit de paternité et ce dans toute utilisation que l'utilisateur entreprend. Ainsi, à titre d'exemple, lorsqu'il reproduira un document par extrait ou dans son intégralité, l'utilisateur citera de manière complète les sources telles que

mentionnées ci-dessus. Toute utilisation non explicitement autorisée ci-avant (telle que par exemple, la modification du document ou son résumé) nécessite l'autorisation préalable et expresse des auteurs ou de leurs ayants droit.

Market Definition in the Google Android Case

Jury

Promoteur : Axel Gautier Lecteurs : Nicolas Petit Sébastien Broos

Mémoire présenté par Joé Lamesch

En vue de l’obtention du diplôme de Master en sciences économiques

Année académique 2015-2016

A BSTRACT

The definition of a relevant market is crucial in many antitrust cases. The assessment whether a firm is in a dominant position or not depends highly on the market that has been defined as comprising all competitive constraints to its behaviour. The ongoing case of the European Commission against Google’s mobile operating system Android is an example of such a case, which additionally includes a very specific type of product: multi-sided platforms. The particular nature and functioning of such platforms leaves most common methods of market definition useless and inappropriate. This thesis will use the growing literature on multi-sided platforms to assess the functioning of these products. Further, a prominent precedent case and a close consideration of the general business strategy of Google will help to reveal and to understand how Android works. What role it and other adjacent software play in this wider business strategy of Google. And what the competitive environment is in which these products are operated. The main insights will be that it makes little sense to consider these products without taking into account their interdependent relationship and the competition they face from firms operating with a different business model. It are in fact these various business models that are competing with each other on a higher level, rather than the single products with each other.

Comparing these findings with the markets the European Commission has defined as being the relevant markets in this case, hints that the Commission might possibly be missing important aspects of the competition that is taking place between these products and their parent firms.

La définition du marché est un élément crucial pour la plupart des procès touchant les lois de la concurrence. L’évaluation d’une éventuelle position dominante dépend fortement du marché pertinent sur lequel se fond l’analyse suivante.

Le procès actuel de la Commission européenne contre le système d’exploitation Android de Google constitue un exemple d’un tel cas. En plus, des produits d’une nature très spécifiques sont impliqués : des plateformes bifaces. La nature et le fonctionnement particulier de ces derniers rend la plupart des méthodes du marché traditionnels inutiles et inaptes pour la définition du marché.

Le présent mémoire va utiliser la littérature croissante des marchés bifaces pour évaluer le fonctionnement de ce type de produit. De plus, un fameux cas précédent et une considération précise de la stratégie d’entreprise générale de Google vont aider à révéler et à comprendre comment Android fonctionne, quel rôle le système et d’autres systèmes liés jouent dans le contexte de cette stratégie d’entreprise et à établir l’environnement concurrentielle dans lequel ces produits sont exploités.

Les conclusions principales seront que, une considération isolée ignorant les interdépendances de ces produits entre eux est inadéquate afin d’évaluer la concurrence à laquelle ils font face car cette concurrence peut provenir de firmes suivant un modèle d’entreprise tout à fait différent.

En fait, la concurrence se déroule entre ces différentes stratégies d’entreprise à un niveau plus élevé qu’entre ces produits en question.

En comparant ces résultats avec la définition actuelle du marché effectuée par la Commission européenne, on constate qu’il y manque plusieurs aspects importants à l’égard de la concurrence entre ces produits et leurs sociétés mères.

1

C ONTENT

I. Introduction ... 2

II. Multi-Sided Platforms ... 4

1. Definition and Aspects of Multi-Sided Platforms ... 5

2. Antitrust Implications ... 9

III. The Microsoft Precedent ... 13

1. Legal Proceedings and Antitrust Accusations ... 13

2. Microsoft’s Windows Software ... 14

3. The Market Definition ... 15

IV. Google’s Strategy ... 19

1. The Google Universe and Business Model ... 19

2. The Android Software ... 20

3. Antitrust Concerns and European Commission’s Investigation ... 22

V. Market Definition in the Android case ... 26

1. Android as an Multi-Sided Platform ... 27

2. The Whole Picture ... 30

3. Markets for different Platforms ... 33

4. Competition between Systems... 37

5. Comparison to the Market Definition of the European Commission and Differences to the Microsoft Case ... 39

VI. Concluding Remarks ... 41

Rerences... 42

2

I. I NTRODUCTION

When talking about the economic reality we live in, the word ‘market’ figures surely among the most used words. It is used to describe sets of economic transactions of more or less one specific kind and in some geographical area. As a synonym for words like industry, branch, sector or field of activity, this economic term has entered our everyday language. In contrast, however, to this mostly vague use of the word, the term has a very specific and precise meaning in the sphere of competition policy. A market includes all products in a geographic area which exert competitive constraints on each other. In merger cases or alleged abuses of dominant position, the definition of the market relevant for the case is often of considerable importance. Whether a firm can be found as being dominant depends highly on the market which it is found to operate in. A metaphorical illustration would be the question whether a football club is dominant.

Defining the sphere of competition as being of national extent (i.e. the Belgian Jupiler League), clubs like Anderlecht or Club Bruges could be seen as dominant. Extending the competition, however, to a European level (i.e. the Champions League or Europa League), the picture changes quite tremendously.

The topic of the following thesis is the market definition in the ongoing case of the European Commission against Google’s mobile operating system Android. This analysis aims at assessing the type of competition Android takes part. Thereby, it attempts to identify the relevant sources of competition and compare it to the definitions the European Commission has defined as relevant in the current case.

To assess the competitive constraints software like Android faces is not that obvious as it might seem at the first moment. The world of IT or online firms is characterised by various products given away for free to users. Intuitively though, a price is a vital characteristic of a market. It is therefore that traditional methods of market definition reveal themselves as useless in such cases. In most cases, this free distribution is caused by the nature of these products which operate as platforms.

The growing literature on multi-sided platforms will be used to understand how such platforms work and what consequences derive for antitrust purposes. Furthermore this theory will provide guidance on how markets can be defined in the absence of prices.

3

When talking about the current Google case, one has to inevitably mention the antitrust accusations Microsoft faced about 15 years ago. Microsoft was alleged to be in a dominant position for desktop operating systems and faced similar charges as Google’s Android software today. In this sense, the Microsoft cases represent an important precedent in terms of substance, form and scope. At that time, however, the economic literature of multi-sided platforms was still in its early stage of development. Moreover, the overall industry of IT firms was structured and functioning in quite a different way.

In order to understand not only the differences to the Microsoft cases, but also to catch the whole picture of the economic environment in which Android is operating, a closer look at the overall business strategy of Google imposes itself. Since Android is just but one element of this wider course of action.

By considering these different aspects, the platform literature, the precedent Microsoft case and the overall business strategy of Google, the following analysis will be provide some insights on the sort and sources of competition Android is facing. The findings suggest that the Commission might be missing important competitive aspects in its market definition. Its view seems to be focused on a quite narrow perception of markets ignoring the competition taking place on a higher level. In other words, the competition for the link or path between users and the contents on the internet. By neglecting this form of competition between various firms, antitrust authorities’ perception of relevant markets might be flawed. More precisely, two aspects seem to be left out by European officials. First, the similar service browsers and app stores provide in accessing the internet. Secondly, products of closed systems, like the ones of Apple, are excluded from the markets.

The thesis will be structured as follows. Section II will summarise the main insights from the literature on multi-sided platforms. Section III will look back on the proceedings in the Microsoft cases. A more wide perception of Google’s overall business model will be provided in section IV before section V attempts to assess the competitive constraints and the markets for mobile operating systems and adjacent software. Concluding remarks can be found in section VI.

4

II. M ULTI -S IDED P LATFORMS

The case at hand is an example of economic activities with quite remarkable characteristics. Remarkable in the sense that they contrast with classical economic theory. For instance, as in the present Google Android case, there seem to be no prices, or zero prices. A rather counterintuitive feature when we think about economic markets.

Nevertheless, this phenomenon appears to be not that rare and may even be perceived as becoming more and more frequent in the various business models of the ‘new economy’ of the internet. They often come without direct monetary prices for consumers. Examples of such models would be: search engines, comparison portals, software, social networks, etc.

All these businesses are thought of as being platforms, where different type of consumers, or users, can interact.

Although the prominence of this platform perception has increased with the emergence of the internet and its new economic activities, the concept of platforms can be found in various industries and at all times in human history. A traditional market place would be one example, here buyers and sellers can meet and trade. Dating clubs are often described as being platforms for men and women to meet. Another classical example would be payment card systems like Diners Club, VISA or MasterCard, which allow clients and merchants to conclude transactions. Further examples would be gaming consoles, software, money or even languages.

From these illustrations above, a defining aspect of platforms becomes already clear, the presence of different types of consumers willing to interact with one another. Hence the denomination ‘multi-sided platforms’.¹

Despite the fact that academic research had already vaguely tackled the issue of MSP, the interest in the topic really took off with a 2002 paper by Richet and Tirole.² This was the beginning of an amplified treatment by scholars. Albeit this increased interest led to a

1 These platforms are also often referred to as „multi-sided markets“. However, since a platform not necessarily constitutes a market on its own, this term might be misleading for the purpose of market definition. Furthermore, the platforms used as examples are normally linking two different sides.

Therefore, it has become standard to use the terms „two-sided“ and „multi-sided“ interchangeably. In the following thesis, the term „multi-sided“ will nonetheless be preferred and the term multi-sided platform will be abbreviated by „MSP“

2 (Rochet & Tirole, 2003)

5

deeper research, no overall accepted definition or general theory has established itself so far.³

The following section will outline the main aspects of the theory of multi-sided platforms, how it compares to other branches of literature and highlight the implications for competition policy.

1. D

A

M

-S

P

The most important and defining feature of MSPs is the specific interdependence between different groups among its customers or users.⁵

For a firm to qualify as MSP, at least one group of consumers has to value the presence of another group. This circumstance distinguishes such businesses essentially from common single-sided undertakings, where such a connection is not existing. To give an example, any one sided shop, a restaurant for instance, serves various kinds of consumers who may differ considerably in terms of socio-economic or demographic characteristics, such as sex for example. However, none of these different consumers is particularly concerned about the attributes of the other guests or their presence in the restaurant. In contrast, for a dating bar or club (which are considered to function as MSP), the picture is quite different. For them it is vital to attract men and women at the same time. Men will only enter when they know they can meet women inside and vice- versa. Hence, each group values the presence of the other.

This specific characteristic of interdependent consumer groups, makes MSPs a quite particular type of business with noteworthy aspects and features.

First, the interdependence between consumers results in an externality. For each consumer, the value of joining a platform depends on the number of consumers on the other side. For the platforms, this means that they face two distinct demand functions which are interdependent. Therefore, the demand elasticity of each function is higher than it would be without these externalities. Consider a platform that serves two groups of consumers, group A and B. A price increase on one side will lead to a direct reduction

3 (Auer & Petit, 2015) (Rebecchini, 2014)

4 Thi spart is based on (Evans, 2003) (Rochet & Tirole, 2003) (Rysman, 2009) (Evans & Schmalensee, 2005) (Evans & Schmalensee, 2012) (Hagiu & Wright, 2015)

5 Hereafter, the terms „users“ and „consumers“ will normally refer to the end-users of devices. Unless indicated otherwise.

6

of consumers on that side. Since the drop of consumers on side A makes the platform less attractive to side B, the number of consumers of type B will drop as well. This will again lead to a further reduction of type A consumers. Thus, this phenomenon is commonly referred to as an indirect network effect.

Second, since the nature of a MSP is to enable different consumers to interact in one way or another, the efforts of a platform are most of all devoted to achieve this pivotal goal of attracting all sides necessary for its functioning. As a consequence, the costs of a platform are mainly borne in order to facilitate the interaction between consumers. It is very difficult or even impossible to attribute these costs to a side which would have caused the costs. Therefore, the computation of marginal costs on a certain side makes little or no sense in the case of MSPs.

Thirdly, the need to ‘get all sides on board’ causes pricing strategies of MSPs to deviate considerably from that of single-sided companies. As the consumers of platforms differ in some basic aspects, they do not value the platform in the same way. Hence, to appeal to all and to induce clients from all sides to join the platform, the MSPs have to price them differently. It is quite common for such platforms to charge one side substantially less than the other side. It might in fact be that one side is not charged at all, or even subsidised to join the platform. As an example, viewers of free-to-the-air television do not pay anything to see the TV programs. On the other side, advertisers are charged huge prices for advertisement slots that the TV station sells to them. In multi-sided businesses it is therefore not only the price level but the price structure that matters in order to maximise profits. In more formal words, the price structure of MSPs has to be non-neutral, meaning that a price increase on one side accompanied by a corresponding drop in the price on the other side, changes the overall number (i.e. is non-neutral) of consumers.

A first important implication is that for one side, there might be no prices, or in other words, prices are zero. Another important consequence is that on one side prices could be below marginal costs. Although these measures cannot be computed for either side, it is nevertheless reasonable to assume that an additional consumer on any side entails some non-negative costs to the platform. In the case of zero prices, these would then be below the marginal costs.

7

All these features might sound familiar and in fact, the treatment of MSPs touches various other topics of economic literature.

For instance, the description of platforms may remind one of firms selling multiple products. Software perfectly illustrates this point. As in the case of Microsoft which tied its internet browser to its software system (cf. Section III), the parallels seem striking.

“But firms that make multiple products for several one-sided markets [...] or several complementary products for a distinct set of consumers [...] do not secure profit opportunities from internalizing indirect network effects.”⁶

Furthermore, the fact that platforms charge different consumers different prices prompt the conclusion that the strategy of a platform can be classified as well-known price discrimination. After all, the goal of price discrimination is as well to increase revenues by increasing the number of consumers, notably by attracting different kinds of consumers. However, for a MSP, it is not only a strategy to increase its revenue, but first of all a necessity to ‘get all sides on board’ and to generate demand on any of its sides.

Most of all, the theory of MSPs seems to relate to the literature of network goods. These contributions discussed the implications of such externalities, the fact that the value of a good depends on the number of consumers using it. It might seem that the treatment of MSPs is an additional contribution to this literature. This perception, however, would not account for the distinguishing fact that MSPs serve more distinct consumer groups.

Although related, this clearly differentiates the two branches of theory.

Finally, the discussion of MSPs and the internalisation of externalities obviously evokes the Coase theorem. This explains how well defined property rights can lay the ground for the internalisation of externalities. It therefore becomes clear that MSPs arise in situations in which this theorem does not hold and individuals are unable to internalise the externalities between them on their own.

Despite the lack of an all-inclusive, generally accepted definition of MSPs, several contributions have tried in finding one.⁷ This thesis will refrain from discussing them all.

Since, they are quite complementary, the common main aspects of these definitions will be outlined instead.

According to these contributions, a firm is multi-sided when

6 (Evans, 2003) p. 338

7 (Evans, 2003) (Evans & Schmalensee, 2005) (Rochet & Tirole, 2003) (Hagiu & Wright, 2015)

8

• there are two or more different groups of consumers

• between whom arise externalities

• which these groups are unable to internalise on their own

• the MSP serves then as intermediary,

• enabling direct interactions between these groups

• the MSP applies a non-neutral price structure

• each side is affiliated in some way to the MSP (through membership fees, opportunity costs, etc.)

At the end, it has to be noted that unlike the above thoughts (and the term ‘multi-sided market’) might suggest, multi-sidedness is not an inherent characteristic of some markets or industries. The decision to operate as a MSP is a genuine choice for a certain business model. MSPs and single-sided businesses can end up competing for the same consumers in the same markets (at least on one side).⁸

This decision is a quite complex one as it involves choices about which sides a platform wants to serve, how open it wants to be and through which pricing strategy and which precise prices it wants to achieve this.⁹

Furthermore, the theory of MSPs unifies various aspects of other well-known economic activities (price discrimination, multiple products, network effects) which are quite, if not all at the same time, common in the real economy. Multi-sidedness might be considered more as a question of degree rather than a clear-cut business model in which a firm fits in or not.

Obviously, all of this makes the treatment of MSPs a rather complicated issue. This complex nature has therefore several implications on antitrust policy and the handling and approach of cases in this field.

8 (Hagiu & Wright, 2015) (Rysman, 2009)

9 (Rysman, 2009)

9

2. A

I

As MSPs have their own distinctive properties and characteristics, they alter the features and nature of antitrust cases involving multi-sided firms.

An appropriate treatment of a case involving a firm, which is or could be seen as a MSP, one should first assess whether network effects are present, how strong and how important they are. Not all businesses with similarities to MSPs will impose a fundamentally different form of the investigation than normal one-sided ones, simply because the externalities might not be that relevant for the case.

Nevertheless, some substantial differences and challenges can arise for competition policy.

The traditional use of market share as a first proxy for market power is even more problematic than it already is in single-sided environments. Since two or more, possibly very diverse, sides are served by a MSP, a figure of ‘market’ share or user share on one side gives an only incomplete picture of the economic power of a platform, or its position compared to competitors.

Moreover, the very nature of MSPs gives rise to various sources of barriers to entry. First of all, it could be very costly to start a platform in terms of capital requirements and fixed costs (development costs for software, construction of a shopping mall, costs of printing presses and distribution for a newspaper). Additionally, in order to get a platform started, the firm would have to reach, at least on one side, a critical mass of consumers. A further issue in this respect is the possible coordination problem that arises. Consumers on one side might be reluctant to join the platform when there is uncertainty about whether consumers on the other side are likely to join. This ‘chicken and egg’ problem has to be solved before a platform can be successful. However, if at least one side engages in ‘multi-homing’¹⁰coordination poses no real problem. As we can see, the handling of a platform is quite demanding in terms of technical as well as economic skill and knowledge.

All of these aspects could impose conceivable barriers to entry for newcomers. These theoretical considerations should, however, not lead to premature allegations of

10 ‚Multi-homing‘ refers to the fact that a user group on one side is affiliating with several simliar platforms to reach users on another side. ‘Single-homing’ in contrast means that a group of users only uses one side to reach users on another side.

10

anticompetitive barriers. Rather, they represent possible ways through which entry could be impeded. A more closer investigation should be undertaken in order to check whether these barriers actually materialise and constitute a real anti-competitive behaviour.

Market Definition

The first important step in merger or abuse cases is the definition of the relevant market which is quite different to the single-sided cases. The nature of MSPs cause the traditional instruments used for this purpose to be of no or only limited use.

For instance, the indirect network effects between the different sides lead to a higher demand elasticity. Given a two-sided business, a price increase on the first side, directly reduces the number of type 1 consumers. But through the indirect network effect, it decreases the number on the other side as well, which again reduces the attraction to consumers of type 1. Therefore, the profitability of a price increase is far more limited in a multi-sided setting than in a single-sided one. An imprudent use of conventional methods of market definition, notably the SSNIP test, would possibly lead to too narrowly defined relevant markets.

Moreover, the fact that the platform serves different consumers on different sides makes it clear that competitive constraints for actions can arise from any of the different sides of the platform. It would therefore be inappropriate to ignore these constraints from other sides of the platform.

Ultimately, the special pricing forms used by MSPs could lead to hasty and flawed perceptions of relevant markets. As seen above, one side of a platform is charged less than the other, often even charged at zero prices. This absence of prices could lead to the view that there is no market at all to investigate on that side. Examples are the difficulties national competition authorities have with defining appropriate markets for free-to-air TV markets.¹¹

The first contributions to the MSP literature mostly limited themselves on highlighting these problems for traditional methods of market definition, when confronted with multi-sided firms, without attempting to solve them. Only after a while did scholars tackle the challenge of defining markets in multi-sided environments.

11 (Rebecchini, 2014) (Filistrucchi et al., 2013)

11

Most contributions were thereby concerned about the problems traditional methods such as the SSNIP test face with MSPs. Either on a formal and general basis¹² or using a precise industry as field of application.¹³

These insights are only useful in an environment where consumers pay monetary prices, this is, however, not the case for the Google Android software. [cf section IV and V] This absence of prices can easily lead to the premature conclusion that there is no market to be defined and investigated. But, as has been shown above, a price of zero is not necessarily an odd thing for a MSP. In fact, for such businesses charging one side a price of zero might be a profit maximising strategy. Hence, the price of zero is no sign of an absence of markets but “it is exactly because the market is two sided that one side does not pay”.¹⁴

Thus, even if there are altered methods of market definition suited for the case of MSPs, they may not be applicable. This puzzle was tackled by Filistrucchi in a first paper in 2008¹⁵ for the case of media products. And further treated in 2013¹⁶. The course of action proposed for such settings will play an important role in the following analysis.

In his works, Filistrucchi argues that, as already mentioned, in cases with MSPs, it is important to take into account all sides of the platform and the possible competitive constraints that can arise from all these sides. But with MSPs, the question would arise whether one should define only one relevant market, comprising all sides of the platform or whether two or more markets need to be defined. According to these papers, the answer to this question lies in the assessment of the type of a MSP. Platforms which offer a way of interaction between different groups could be classified as a ‘two-sided transaction market’¹⁷ when the interaction between the groups is an observable transaction. In these cases platforms are able to charge an access fee as well as a usage fee (although they do not necessarily have to).

On the other hand, a platform can be characterised as a ‘two-sided non-transaction market’ when the interaction between the groups is no traditional transaction and the interaction is not observable.

12 (Evans & Noel, 2005) among others

13 (Argentesi & Ivaldi, 2005)(Filistrucchi, 2008)

14 (Filistrucchi et al., 2013) p. 25

15 (Filistrucchi, 2008)

16 (Filistrucchi et al., 2013)

17 This terminology can again be misleading when applied tot he purpose market definition. The terms

„transaction plaform“ or „non-transaction platform“ will therefore be used instead.

12

The distinguishing feature between both types of MSPs is that in a transaction market a firm is necessarily on both (all) sides of the market, whereas in non-transaction markets, it is possible that a firm is just on one side of the market. In other words, this distinction could answer the question, whether a service a MSP provides on one side should be seen as being in the same market as the service provided on the other side.

As an example, in the media scenario, a free-to-the-air TV station can be seen as being in the same market for advertising as newspapers. Advertisers may regard both as a substitute. But the TV station would not be in the same market on the user/reader/viewer side since consumers would not consider it to be a real substitute for newspapers. This view would be justified since the interaction between advertisement and readers/viewers (people seeing advertisements) would not be a transaction and could not be observed.

In contrast, a payment card system, such as VISA, should be seen as being in the same market on both sides, comprising the provision of payment cards for consumers and the provision of payment terminals for cards for merchants. The reason would be that the interaction between both groups would constitute a real transaction which would be observable. From these thoughts derives the main suggestion:

“In two-sided non-transaction markets, two (interrelated) markets need to be defined.

In two-sided transaction markets, only one market should be defined.”¹⁸

Other challenges for antitrust policy arise in the sphere of actual enforcement. Despite the fact that considerable scientific attention has been devoted to the study of multi- sided businesses, no general approved, ‘unified’ theory has emerged so far. While this is surely due to the complex issue at hand, it makes it difficult to use these theoretical insights in a systematic manner in legal cases. As a consequence, the lack of clear and unambiguous definitions and terms leave the door open for arbitrary use of the insights of this branch of theory in antitrust cases. This shortage of theoretical consistency and the rather common aspects of multi-sided businesses, notably the network externalities, could even lead to a kind of overuse of multi-sided theory and a kind of policy bubble.¹⁹ Finally, leaving these conceptual uncertainties aside, the already stressed complex nature of cases involving MSPs could likely lead to cumbersome and lengthy

18 (Filistrucchi et al., 2013) p. 10

19 (Auer & Petit, 2015)

13

investigations, increasing overall enforcement costs and decrease legal resilience of findings.

III. T HE M ICROSOFT P RECEDENT

The ongoing antitrust case against Google has had a prominent precedent. At the turn of the millennium, Microsoft faced serious antitrust accusations on both sides of the Atlantic. These were the first big cases in the ever-growing high-tech industry of information technology (IT) firms. Hence, they were and are thought to set out a first landmark on how such cases could or should be treated.

And indeed, there are striking similarities between the charges Microsoft faced back then and the current accusations against Google.

This section will outline the legal proceedings and economic issues of the Microsoft cases in the U.S. and in Europe. It will highlight the differences and parallels between them and take a look on how the relevant markets were defined.

1. L

P

A

A

Already in 1994, Microsoft was under investigation by both, the U.S. Department of Justice (DOJ) and the EC. Object were its agreements with PC manufacturers. Both cases were settled between Microsoft and a joint team of the DOJ and the European Commission (EC).²⁰

In the United States the case resumed in 1998, when several states and the DOJ filed suit against Microsoft. The accusation was that Microsoft was illegally tying its Internet Explorer to its OS through its agreements with PC manufacturers. They would have had to exclusively acquire Microsoft’s browser to the detriment of rivals.²¹

The case was settled in 2002 with a consent decree and behavioural remedies.²²

20 (Rubinfeld, 2003) (Press Release IP/00/141, European Commission)

21 (Rubinfeld, 2003)

22 http://www.seattletimes.com/business/microsoft/long-antitrust-saga-ends-for-microsoft/

14

In Europe, the EC was concerned that Microsoft would leverage its dominant position for PC OS into the market for server OS. In August 2000, Microsoft received the statement of objections from the EC. It was alleged to withhold interface and protocol information from competitors so that only Microsoft products could interoperate with its Windows software. This would have constituted a significant disadvantage for rival server OS given that Windows was the by far most used OS in the world. By this conduct Microsoft would foreclose the market for server operating systems and artificially construct barriers to entry for competitors.²³

In August 2001, this statement of objections was extended by the accusation of illegally tying its media player to its OS. Thereby, it would offer to its own media player a powerful channel of distribution, which would not be available for rival players.²⁴

Eventually, in March 2004, Microsoft was found guilty of the accusations and was forced to disclose its interfaces for its OS, to offer an alternative version of Windows 2000 without its media player and a fine of €497 million was imposed.²⁵

Subsequently, in 2007 a browser rival filed a complaint against Google with the EC, accusing Microsoft of illegally tying its Internet Explorer to its dominant PC OS. By this conduct, it would foreclose the market for competitors. In 2009, a statement of objections was sent. Microsoft offered a series of commitments in order to resolve the alleged antitrust issues. After a year of negotiations, the case was settled.²⁶

2. M

’

W

S

At the heart of the Microsoft cases was its popular operating system for desktop computers, Microsoft Windows. Operating systems allow other software (applications) to access the hardware of a device. Users on the other hand can run these applications on their devices for various purposes, but only if these apps are compatible with the OS on their device. Developers can sell their software only to users if their application is compatible to the OS on costumers’ device. It is only through this enabled interaction

23 (Press Release IP/00/141, European Commission)

24 (Press Release IP/01/1232; European Commission)

25 (Press Release IP/04/382, European Commission)

26 (Case COMP/C-3/39.530 - Microsoft (tying), European Commission)

15

between users and applications that operating systems unfold their true commercial value.²⁷

Hence, the product of operating systems is clearly characterised by network effects. On the one hand, users value it more, the more different applications are compatible with the software and increase the possibilities of their devices. On the other hand, it is more interesting for application developers to write software compatible with a certain OS, the more consumers are using it. The indirect network effects are quite obvious as each group of users cares about the number of users on the other side.

The pricing strategy of Microsoft seems to endorse the first impression of an MSP. While users were charged to get a version of the software (though indirectly trough PC manufacturers as intermediate agents), application developers were charged almost nothing. They were even subsidised in one way or another by Microsoft which was actively encouraging the writing of Windows compatible applications.²⁸ (Although some of the most popular applications were written by Microsoft itself). This course of action could fit the pattern of a non-neutral pricing strategy.

From these descriptions it seems quite clear that Windows satisfies all the characteristics of a MSP from section II. There are more distinct user groups, between which network effects are present. And for the internalisation of these effects, both groups need the operating system as an intermediary. Microsoft charges different sides different prices in order to create demand on any of the sides and maximise the overall number of users. And through prices and agreements, each side is clearly affiliated to the platform. Accordingly, software in general and operating systems in particular are an often cited example of MSPs.

3. T

M

D

During the described Microsoft cases, the relevant markets were defined as being the

‘market for PC OS’ (in the E.U.) and the ‘market for OS for Intel-compatible desktop PCs’

(in the U.S.). Moreover, according markets for media players and web browser were defined. Obviously, both competition agencies came to practically the same conclusion

27 (Evans & Schmalensee, 2005)

28 (Evans, 2003) p. 352

16

on behalf of the relevant market in question. And their reasoning was very similar as well.

In the U.S., Apple’s OS was excluded from the relevant market since the DOJ saw no real substitutability between both software products, using the product characteristics as a method to define the market. Similarly, the EC concluded that PC OS could not be used as server OS, which would form a separate market on its own.²⁹

Furthermore, on the demand side, the DOJ found it unlikely that non-desktop devices would be part of the market for PC OS, since it considered it not credible that a 5% - 10%

price increase for Windows would lead a significant amount of users to switch the product. It should be noted that this argument actually treated the question whether consumers would switch devices not directly if they were to switch OS.

However, manufacturers (the direct clients of Windows) indicated that such a price increase would not induce them to change the OS they used on their devices.³⁰

Finally, on the supply side, both agencies came to the conclusion that it was unlikely that other firms could easily step in and offer a comparable product than Microsoft’s OS. The explanation was the alleged “applications barrier to entry”. To effectively challenge Windows OS, a potential rival product would have to offer a considerable amount of compatible applications, equivalent to the most popular applications for Windows, and this right from the start. Possible competitors could not engage in developing and supplying a PC OS as well as the required applications without bearing substantial risks and costs.

By some, this relevant market definition was viewed to be too narrow. The emphasis on the relations between Windows and PC manufacturers was criticised for ignoring the relationship between Microsoft and ISVs (independent software vendors).³¹

Likewise, Microsoft criticised the relevant market definition as to static and inappropriate for the fast evolving and dynamic nature of the industry. The view that there were no close substitutes for Windows would only hold if the number of applications available for software products would be fixed. This, however, would not be the case since numerous developers would write applications, not only Microsoft. Hence, it would be absolutely possible that new applications were written for rival software or

29 (Rubinfeld, 2003) (Gil-Molto, 2008)

30 (Rubinfeld, 2003), a similar reasoning was applied in Europe.

31 (Economides, 2001)

17

that existing applications would be adapted to them.³² Likewise, Economides found that

“The dynamic view of market definition invites an examination of the incentives that ISVs have to create such applications”³³

Microsoft claimed that the relevant market was in fact the market for software platforms as such, offering a platform to application developers and users. Operating systems were just one of these platforms but not the only one. Other middleware would serve the same purpose. Otherwise, Microsoft would not have needed to take actions to defend its OS against rivals such as Netscape Navigator.³⁴

Ultimately, Microsoft tried to show it would face pricing constraints. Using a short-term model it claimed it would have priced its OS tremendously higher, were it to have possessed monopoly power. Since it did not, there must have been other important competitors imposing this price constraint. Hence, the relevant market had to be wider than the one alleged by authorities.³⁵

Albeit the clear signs that Windows could be seen as a MSP, the definition of the relevant markets was based on rather traditional single-sided reasoning. The main focus were product characteristics and functionalities. Apple was excluded from the defined market (partly) on the basis of usage shares. The logic of a (one-sided) SSNIP test was applied by a kind of questioning of PC manufacturers whether a price increase would induce them to change the OS used on their devices.

Only by referring to the ‘applications barrier to entry’ authorities showed that the present network effects did not escape their assessment. Despite this consideration of the special nature of the product under investigation, both competition agencies defined a quite narrow relevant market, focusing on OS as an input for PC manufacturers, ignoring the side of application developers. Further, they defined similar narrowly the respective impacted applications, namely separate markets for web browsers and media players. Again, by recognising these markets as complementary, it appears clear that the competition authorities were to some extend aware of the interdependencies between these products, but did not perceived them a MSP.

32 (Rubinfeld, 2003) (Economides, 2001)

33 (Economides, 2001) p. 14

34 (Rubinfeld, 2003)

35 (Rubinfeld, 2003)

18

The view of a single market for operating systems was not really challenged by Microsoft at first. Only later in the process, as an attempt to defend its actions, the company argued that these markets were defined too narrowly. Microsoft stated that applications were written for a whole range of software products other than Windows.

Moreover, it claimed that all these products would serve as a platform between users and application developers, and that this platform functionality was the real economic service provided.

Although these arguments come already very close to the meanwhile adopted description of MSPs, both parties, the competition authorities and Microsoft, were lacking a pronounced perception and awareness of the nature of MSPs.

This incomplete view of the character of these software products may be due to the fact that the meanwhile prominent literature on MSPs had not yet emerged at the time of the investigations. Yet, the European browser case stared only in 2009, at a time when already a lot of academic attention was devoted to MSPs. Still, the markets were defined in the same way. This commitment to the previous market definition could be due to the desire of being consistent with the previous investigations and not to undermine the reasoning in these cases.

These defined markets could as well have been reflecting in an authentic way the economic reality. Whether these defined markets were appropriate or not it not of interest in this analysis.

19

IV. G OOGLE ’ S S TRATEGY

1. T

G

U

B

M

Google was created at the turn of the millennium. Its initial activity was performing an internet search engine. By now, the company has evolved into being one of the most influential technology and internet firms. Google Search is the most successful search engine in the world.³⁶ And Google expanded its field of activity in a considerable way, offering now a whole range of internet services including email accounts, map services, online videos, cloud computing, software, etc.

Its revenues stem primarily from advertisement which generates around 77% of total revenues.³⁷ These revenues include not just the mere offering of advertising space but also the revenues for key words firms buy in online search. This is a feature that distinguishes Google from some of its perceived competitors, for instance Apple, Amazon or Microsoft. Google does not primarily sell hardware (phones, PCs, tablets, etc.) or software to generate income. It’s business model instead is focussed on attracting internet traffic to its websites, applications and services. By owning highly frequented sites it is able to charge higher prices for advertising spaces and keywords.

However, despite the fact that Google has a vast offer of websites and services, its core product and cash cow remains Google Search. To protect this highly profitable business and prepare the ground for further growth, Google tries to lock in as many internet users as possible in its online ecosystem.³⁸ By linking and combining its offers to a more varied and all inclusive product, it aims to attract users, keep them using Google’s services and increase the monetary value of its additional services and websites.

However, some market participants and observers regard this course of action as possibly anti-competitive. A widespread claim is that Google tries to leverage its market power in the online search market into other spheres of online activities in order to promote own services and prevent competition from secondary markets. By tying and bundling its services together it would abuse the dominant position of Google Search

36 http://www.statista.com/statistics/216573/worldwide-market-share-of-search-engines/

37 https://abc.xyz/investor/news/earnings/2016/Q1_alphabet_earnings/

38 (Auer, 2016), http://arstechnica.com/gadgets/2013/10/googles-iron-grip-on-android-controlling- open-source-by-any-means-necessary/

20

which would be an infringement of competition law. Some examples of Google’s conduct in this regard are given hereafter.

Google is blamed to favour its own services in its online search engine. Hence, Google services would get beneficial placements in the list of results and would benefit from an improved format. Accordingly, Google Maps (a web mapping service) appears with an oversized preview of the location researched, Google Shopping (a specialised or vertical search engine for products) appears with depicted images and product information, YouTube results appear with integrated videos ready to be played, etc.³⁹ By this behaviour, Google would enable its services to become popular by leading a considerable amount of internet traffic to them on a free of effort basis. As a consequence, potentially more relevant rival services are less visible and users are less likely to click on them.

Another accusation consists in the use of other websites’ content for Google’s own vertical search engines (e.g. Google News: Articles from news publishers, or Google Places: a search engine for local businesses) In order to stop this use by Google, to which internet publishers may have never agreed on, the only way for them would be to opt out from all Google services, including Google Search. This could reveal itself as an unfeasible option for several internet websites.

The last two complaints relate to Google’s online advertisement business. Google is accused of forcing advertisers to exclusively lead online advertising campaigns on its AdWords⁴⁰ service if they want to advertise on Google Search. And furthermore, Google would prevent internet content publishers (basically any site) from using rivals to its AdSense service in order to sell advertisement on their sites.⁴¹

2. T

A

S

As seen above, Google proceeds in various ways to link its services and to promote them.

This thesis will underlay the specific case of Google’s use of its mobile operating system (mOS) Android. This conduct is currently under investigation by the European

39 (Edelmann, 2015)

40 Ad Words is a service provided by Google for firms wanting to purchase advertising possibilities on the internet. Ad Sense, on the other hand, is a service for owners of websites to sell their advertising space to advertisers.

41 (Edelmann, 2015) (MEMO/13/382, European Commission)

21

Commission (EC). Before considering the Commission’s concerns, the way Android works and its role in the Google universe will be examined.

Android is a company founded in 2003 with the aim of developing an operating system for mobile devices. In 2005, the company was bought by Google and in 2007 (the year the first iPhone came on the market), Google launched the Android Open Source Project (AOSP). Thereby, the source code of the mOS was made freely available. Simultaneously, Google founded the Open Handset Alliance (OHA), a consortium of hardware, software and telecommunication firms. The goal of the OHA is to oversee the project and promote further open source standards for the Android software and mobile devices. From this regard, the AOSP comprises more than the sole operating system. The AOSP is perceived as being composed of the mOS, middleware (particularly app stores) and applications to run on the mOS.⁴²

In 2008, the first smartphone running on Android was released. Since then, Android has become the most widely used mOS in the world. About 290 million smart phones sold in the first quarter of 2016 were Android devices, giving it a share of around 80%.⁴³ Furthermore, the number of available applications on Google Play has recently overtaken the number available for Apple’s iOS, the long time leader in this field.⁴⁴ These figures illustrate the immense success of Android in one of the most important segments of the internet economy. In 2014, the number of people using the internet on a mobile device surpassed for the first time that of users on a fixed one.⁴⁵

Consequently, a lot of application developers have engaged in developing applications compatible with the Android mOS (cf. Figures on Applications above). Google itself is not only developing the mOS but is contributing as well to the abundance of applications available for it. However, some of these applications are proprietary, i.e. the source code is not free (e.g. its app store Google Play, YouTube, Gmail, etc.).

Furthermore, Google is switching more and more from developing applications in the open source form (AOSP)to further develop them as proprietary applications. This means which were initially developed under open source licenses, are now being

42 (Auer, 2016)

43 http://www.statista.com/statistics/266136/global-market-share-held-by-smartphone-operating- systems/

44 http://www.zdnet.com/article/ios-versus-android-apple-app-store-versus-google-play-here-comes- the-next-battle-in-the-app-wars/

45 http://www.smartinsights.com/mobile-marketing/mobile-marketing-analytics/mobile-marketing- statistics/

22

developed as Google’s own proprietary applications. (e.g. Google Search application, Google Music, Google Calendar, etc.)⁴⁶

While these different parts of mOS, middleware and applications are definitely interrelated an inclusion into a single product definition seems nonetheless questionable. In the following, the term Android will always refer to the sole mOS.

3. A

C

E

C

’

I

General antitrust concerns

The precise antitrust concerns related to Android lead to three accusations towards Google. According to these, Google’s use of Android would prevent rival applications from gaining ground, prevent the growth of rival mOS and finally would prevent the creation of viable alternative versions of Androids (so-called ‘forks’).

The source of these antitrust concerns stems from the closed part of Android, namely Google’s proprietary applications. Since some of these applications are quite important to the functioning of the platform (e.g. Google Play offers by far the most applications for Android) or are highly popular (YouTube) and face practically no close substitutes, these applications could be considered as ‘must-have-applications’. A lot of hardware makers use Android due to its free accessibility and they want to offer their clients ready to use products. Hence, they will want to use these must-have-applications and possibly want them to be pre-installed on their device.

To be able to pre-install these proprietary applications, device manufacturers enter into agreements with Google to license these applications. These agreements are known as Google’s Mobile Application Distribution Agreements (MADA). However, these agreements oblige the manufacturers to install a whole suite of Google applications.

Hence, a lot of Google’s applications are only available as a bundle.⁴⁷

Manufacturers could feel the need to pre-install this bundle of applications in order to have a commercially viable product. Not only are they required to pre-install certain Google applications, but some of them also have to be placed in prominent positions on

46 http://arstechnica.com/gadgets/2013/10/googles-iron-grip-on-android-controlling-open-source-by- any-means-necessary/

47 (Edelmann, 2015)(Auer, 2016)

23

the device’s user interface. Moreover, some Google applications have to be set as the default service on the device⁴⁸. Although other applications and services are allowed and can be downloaded by users, there is a strong suspicion that the privileged placement and default setting prevents users from doing so. Since some services are already available, users would see no need to have redundant services on their device and hence, would not engage in searching for possibly more apt rival offers.

An example for this accusation is the case of the Korean search engines Daum and Naver.

In 2011, they filed complaint against Google, claiming that the imposing Google Search as default search engine on the growing number of Android devices, would make it impossible to convince consumers to switch.⁴⁹

Another example is given by the fate of Skyhook, a company offering location based services (for instance, providing location data even if the user has no GPS function). In order to promote its own location service (Google’s Network Location Provider), it imposed manufacturers to use it as default. As a consequence, major manufacturers which were so far using the Skyhook service, switched to the Google service.⁵⁰

Moreover, Google is accused of preventing rival mOS to grow by tying its Android software to popular applications. In fact, it refuses to supply meta-data for its YouTube website so that rival mOS are unable to offer a competitive YouTube application. Since YouTube is one of the most visited websites and an accordingly popular mobile application this constitutes a serious competitive disadvantage for rivals. Microsoft filed complaint against this behaviour with the EC in March 2011.⁵¹

The last accusation is the prevention of viable alternative Android versions (forks).

Some see these forks as the fiercest potential competition for Google’s Android software.⁵² Again, its power over its proprietary applications is the mean by which Google achieves this goal. Google has brought together major device manufacturers and

48 (Edelmann, 2015)(Auer, 2016)

49 (Edelmann, 2015), http://www.nytimes.com/2011/04/16/technology/16google.html?_r=0

50 (Edelmann, 2015), http://arstechnica.com/tech-policy/2010/09/skyhook-google-made-oems-break- business-deals-ripped-patents/

51 http://www.foxnews.com/tech/2011/03/31/microsoft-files-antitrust-suit-google.html

52 http://time.com/money/3093863/google-android-biggst-threat-aosp/

http://arstechnica.com/gadgets/2013/10/googles-iron-grip-on-android-controlling-open-source-by-any- means-necessary/

24

software developers in the OHA. It eases their dealing with Android and its Google applications but binds them to refrain from developing alternative Android software.

While many observers and market participants share these views of Google’s anti- competitive behaviour, there are of course other voices downplaying these allegations.⁵³ Firstly, the claim that Google would foreclose the carrying of other applications on Android devices is criticised since Google would not force manufacturers to use its applications. For instance, Amazon’s Kindle runs on an Android based operating system, and comes without any Google application. Another example would be Yandex, a Russian search company which offers a whole application suite for Android devices, which is pre-installed by a Russian manufacturer. Furthermore, users are free to download and use any application they like on their devices, even if Google applications are pre-installed. The view that consumers would hesitate to do so in order to save on their device’s storage capacity is mostly seen as invalid, since an average application takes up only about 0.07% of an average device’s capacity.

Secondly, the allegation that devices without Google applications would be commercially unviable and would therefore, among other things, prevent other mOS from spreading, seems again less convincing with the examples of Yandex & Amazon in mind. An additional example would be the case of China where 70% of (mostly forked) Android devices are used without any Google applications.

Finally, the prevention of forks and the encouragement to pre-install a defined set of Google applications could also be seen as the standard way of competing in this particular industry. Other mobile devices offered by Apple, Black-Berry or Microsoft all come with a pre-installed set of applications, offering consumers a complete device and a so called “Apple experience” or “Microsoft experience”. In order to be able to compete with these firms, and to meet user’s expectations, Google has to ensure that Android devices come with the same basic functionalities and that the underlying mOS is compatible with a maximum of applications.

The European Commission’s Investigation

The European Commission’s (EC) investigation into antitrust allegations about Google started on the 30^th November 2010, following complaints by search service providers.⁵⁴

53 http://www.project-disco.org/competition/030314-observations-on-the-economics-of-mobile-app- suite-bundling/#.VxDowzB942x

25

Initially, the complaints were only concerned about an alleged abuse of dominance in online search.

By March 2013, the EC notified Google that it had isolated 4 concerns about Google’s behaviour in online search.⁵⁵ At the same time, an anti-Google lobbying group, Fairsearch, filed complaint about Google’s Android software.⁵⁶ And in April 2013, the EC acknowledged that it was examining whether these practices would also require a deeper investigation. On the 15^th April 2015, the EC launched a separate investigation apart from the investigation of the online search concerns [IP2015]. Finally, on the 20^th April 2016, the EC sent a statement of objections to Google where it expressed its views about Google’s alleged dominant position in the markets for general internet search services, licensable smart mobile operating systems and app stores for the Android mobile operating system.

More precisely, the EC accuses Google of breaching EU competition law by:

1) Requiring the pre-installation of Google applications and the default setting of some of these applications on Android devices as a condition to license Google proprietary applications.

2) Hindering the distribution of alternative Android systems.

3) Giving financial incentives to manufacturers to exclusively pre-install Google Search on their devices.⁵⁷

54 (Press Release IP/10/1624, European Commission)

55 (Press Release IP/13/371, European Commission)

56 http://fairsearch.org/fairsearch-announces-complaint-in-eu-on-googles-anti-competitive-mobile- strategy/

http://fairsearch.org/list-of-complaints-and-complainants/

57 (MEMO/16/1484, European Commission)