On the 19th of September the Google decision was finally made public after a long waiting period, which in hindsight may seem less surprising given its length (327 pages!) and extensive scope of analysis. As in the case of Microsoft, the decision provides us with lots of material for discussion. The decision covers the market definition as well as the analysis of abuses in the context of two (and multi) sided markets and potentially the matter of extraterritorial application of EU competition law. This post looks into some of the issues concerning the market definition. The remainder of the issues will be addressed soon in a follow-up post.
The market definition in the Android decision is undoubtedly one of the most disputed aspects of the case, which is not surprising as any abuse of dominance case depends on the outcome of the market definition for the purpose of establishing dominance. However, while market definitions are often disputed in practice, the market definition in the Android case constitutes somewhat of a novelty, as it involves (at least two) multi-sided platforms: namely the Android OS and the Google Play Store. Accordingly, when defining the relevant market under such circumstances it is not only about determining the scope of the relevant market but also about determining how many markets need to be defined. As discussed on multiple occasions, this is due to the fact that the multi-sided platforms facilitate the interaction between two or more separate customer groups that display an (inter)dependency of demand with regard to each other and the platform. Such interdependency is a result of the indirect network effects that are often in play in the case of platforms. The greater the interdependency and the more intense the bi- or- multilateral indirect network effects are, the more likely it is that the relevant market will cover more than one side of the platform (for more on this see previous post). Despite the fact that this challenge has been extensively debated in previous cases dealing with platforms, particularly in the context of Amex, this matter was not explicitly addressed in the case of Google. However that is not to say that no decision on this matter was taken. Quite the contrary, the manner in which the market were defined in this case can be instructive for future practice.
The Android decision covers the definition of four relevant markets that were considered to be affected by Google’s practices, namely (I) the market for licensing of smart mobile OS, (II) the market for Android app stores, (III) the market for the provision of general search services, and (IV) the market for non-OS specific mobile web browsers. In the context of two and multi sided markets it is primarily the first two relevant markets addressed in the decision that are interesting.
The definition of the market for the licensing of smart mobile OS is undoubtedly the one which was the focus of most debates on the Android decision. This is primarily because the definition of this market was often equated with the question of – are Apple and Google competitors? – which of course they are. Consequently, it is unsurprising that the finding of the Commission that the scope of the relevant market for smart mobile OS does not include Apple raised a few eyebrows. So was the Commission wrong in this matter? No. So how can Apple compete with Google but not be part of the relevant market? The answer is quite simple – asking whether Google and Apple are competitors in abstract has no value for the market definition process if the question is not accompanied by a specific perspective and aim. After all, the market definition process is not an aim in itself and cannot be done in abstract. The case against Google deals primarily with a (multi product) tying abuse. Accordingly, the aim of the market definition process in this case was to help assess whether such tying practices were capable of producing anti-competitive effects. Given that the tying practices were applied with respect to the OEMs it is required that the relevant market and the assessment of market power is performed from that perspective. In other words, the question of substitution is one that needs to be posed with respect to OEMs and not in general terms of Google vs. Apple. This is because in tying cases the relevant market must be defined at the same level of distribution as the tie. Once this approach (which was confirmed in the Microsoft case) is adopted, the conclusion that Apple is not part of the relevant market of Google with regard to app store and smart mobile OS is quite reasonable. After all, from the perspective of OEMs no matter what type of conditions Google introduces in its licensing contracts they could never switch to IOS and/or the App Store. That is not to say that Apple does not pose any competitive pressure on Google as it certainly does, however, that pressure is not with regard to the adoption of Android OS and Play Store by OEMs.
The question of how many markets need to be defined for a platform often boils down to the choice between the definition of one or more markets for the platform depending on the number of customer groups interconnected by it. Although the Android decision does not address this matter as explicitly as previously seen in the credit card cases in the EU and US, a decision on this issue was nonetheless taken implicitly in the assessment of substitutability. In both the market for the licensing of smart mobile OS and the market for Android app stores the number of relevant markets per platform was essentially one. In the case of market for android play store, substitutability was limited to android compatible app stores indicating that a single relevant market was defined for the app store platform including all its customer groups (developers and android users). Similarly, the assessment of substitutability for the Android OS was done, logically, only with respect to other smart mobile OS, meaning that the relevant market covered all the sides of the platform, namely OEMs, developers and consumers. Accordingly, the decision provides yet another example where a single relevant market can be defined for a multi-sided platform despite the skepticism expressed on multiple occasions with regard to the existence and/ or desirability of such an approach (see e.g. the dissenting opinion in the case of Amex).
In addition to this one market approach, the market definition of these two markets also shows how the number of sides of a platform can be a decisive characteristic for substitutability purposes. Due to the fact that Android OS is licensable, it can be seen as a three-sided platform connecting OEMs, developers and consumers. This was similarly the case with some android forks and Windows mobile. By contrast, Apple’s OS and Blackberry’s OS by not being licensable, can be seen as platforms missing the OEM side. This in turn of course limits substitutability among the various smart mobile OS from the perspective of OEMs. If, however, the assessment of substitutability and market power would have been done on the sides of the platforms that concern consumers or developers, the scope of the relevant market may have been different. Unlike OEMs, consumers and developers are likely to be less concerned with the question of licensing when considering switching from one platform to the other. Accordingly, when a single relevant market is defined for a two or multi sided platform the scope of such market will depend greatly on the perspective of the customer group chosen for the assessment and the presence of such customer group on other (potentially substitutable) platforms.
Finally, another interesting issue that can be identified in the context of the market definition in the Android decision is the use of the SSNRQ test as a replacement in of the SSNIP test. Like in the case of most platforms, the majority (if not all) of Google’s products are provided to some of its customer groups (mainly OEMs and consumers) free of charge. Accordingly, when defining the relevant market in such circumstances, the SSNIP test is not a very useful tool since there are no prices to which a theoretical increase can be applied and assessed. An increase of 5-10% of zero will still be equal to zero and theorizing an increase in price from zero to a positive number cannot be considered an adequate mathematical alternative. In light of this problem, it has been suggest to modify the SSNIP test into another tools suitable to deal with zero priced products or services, where the SSNRQ could serve as an alternative. In the case of the SSNRQ the matter of substitution concerns a hypothetical scenario in which the price of zero in maintained, however, the quality of the product or service provided by the dominant undertaking is reduced. Although the test is logically sound, and thus its adoption by the Commission is welcome, its application in practice requires answering some difficult questions which are not always answered in the Android decision.
First, what are the qualities that are selected for such a test and how does the selection process look like? In the Android decision this was done apparently by virtue of a survey made by Google for developers (see para. 286 and footnote 306). One may wonder whether this is the most suitable manner to make the selection in future cases rather than making a specific survey for the purpose of the investigation. Second, it is also important to consider what degree of quality decrease simulates the same effect as a price increase of 5-10%. This information is not included in the decision but would be very important for future practice as it is evidently directly connected to the assessment of substitutability. Third, it is unclear how the SSNRQ (or any test for that matter) should take into account the two or multi sided nature of platforms and the interdependency of demand among their various customer groups. In this matter too it is unclear from the decision how the test incorporated the multi-sided character of the platforms in this case in the process of application.
The Android decision delivers a valuable source of guidance and questions in the case of the market definition for two and multi sided platforms. It solidifies the possibility of defining a single relevant market for a two or multi sided platform that was first seen in Amex but without providing a clear motivation for such (logical) approach. Furthermore, it shows the importance of choosing a specific perspective and taking into account the multi-sided character of the concerned platform and its (possible) competitors for the purpose of substitutability assessments. Finally, it shows that the Commission is capable of adapting its tools to zero pricing strategies by replacing the SSNIP with the SSNRQ, however, it fall short in delivering the legal and economic justification for such a switch