Comments on Android (I): some questions for economists on market definition
Like many others, I guess, I jumped on the Android decision as soon as it became available (see here). If you have not done so, you should definitely read it. It provides abundant food for thought. It is not an exaggeration to say that a whole competition law course could be taught around it. Pretty much all the (unresolved) issues are there, in a fascinating factual setting with plenty of intriguing ingredients.
It would be great if the decision were widely discussed. The general interest is not served when only those with an interest in a case talk and write about it. When there is so much at stake, the atmosphere tends to get ugly and aggressive (footballised, if you want). The youngest among you should not feel discouraged: you can change the tone of debates by asking questions and/or sharing your views.
I thought I would kick off (no pun intended) with the issue of market definition and dominance, which has attracted a great deal of interest. One of the most intriguing aspects of the original press release was the suggestion that the Apple and Android ecosystems do not constrain each other. The detailed rationale behind this conclusion is now there for all to see.
Since the definition of the market and the assessment of dominance are essentially economic exercises, the right way to go about it is to raise some questions to specialists. It would be wonderful if you could share your thoughts on the comments section.
I will not discuss all aspects of these two questions, just the two that I find to be particularly interesting. Some people may react by saying that these two questions are not even decisive. To which I reply: perhaps, but I am not, and have never been, interested in the outcome of individual cases. Plus, there are more posts for discussion coming up.
One of them relates to the app store. According to the decision, there is a separate market for Android app stores. The most salient implication is that Apple’s app store would be on a separate market. In the same vein, the constraint placed by Apple’s app store would not be sufficient to rule out a finding of dominance.
The second concerns the market for operating systems. The Commission concludes in the decision that there is a separate market for the licensing of smart mobile OS. Again, non-licensable smart mobile OS (read: Apple iOS) are found not to be on the same market. What is more, the constraint coming from Apple iOS is not deemed strong enough to exclude a finding of dominance.
App stores as two-sided markets? What are the implications?
The first set of questions is for economists in general and for Lapo Filistrucchi in particular. As most of you know, Lapo – together with his co-authors – has written influential work on two-sided markets. In part, his research seeks to tame people’s enthusiastic tendencies to see two-sided markets everywhere. So whenever I notice a setting that might be a two-sided market, I ask myself what he would say.
My (cautious) impression is that the app store is a two-sided market. Would that be correct? Using Lapo’s helpful categorisation, I am inclined to conclude that it is a transaction market (in an app store, there is a transaction between the two sides of the market). Hopefully full points so far.
If I understand the scholarship on two-sided platforms correctly, it would follow that there is a single market encompassing both sides. Even more interesting is the analysis of the competitive constraints. My question here I guess is: how are the constraints evaluated in practice?
In this regard, the decision provides a fascinating case study. If you read the decision (paras 652-673), you will see that the Commission focuses, by and large, on Android end-users’ inability and/or unwillingness to switch OS (and thus phones). I guess my question here is: is this factor decisive?
According to the Horizontal Merger Guidelines, the absence of switching opportunities does not rule out a horizontal overlap (that is, a competitive constraint). I would say that, in a two-sided market, the fact that one side of the market cannot (or would not) switch is even less decisive. I would welcome thoughts and clarifications in this sense.
How about the other side of the market? Less importance is given to app developers in paras 652-673. But it is worth reading para 668, where the Commission states (uncontroversially, I would say) that app developers would not switch away from Android because they could not afford to do so.
My question here is: does it follow from that – uncontroversial – finding that Android’s and Apple’s app store do not constrain each other? As I read para 668, I thought of an Australian-like country with two very large supermarket chains (I visited Australia in 2013 and what I paid for groceries still haunts me). Many suppliers would not be able to give up either supermarket chain in such a scenario. Would it follow that the two chains do not constrain each other?
Android as a franchise and the analysis of indirect constraints
When I first wrote about Android, I suggested that the underlying business model is best understood as a form of franchising. I am still of this opinion after reading the decision. Just like McDonald’s, Google licenses its formula so that third-party OEMs can sell their products alongside vertically-integrated manufacturers (euphemism for Apple).
Seen from this perspective, the Android decision claims that the licensing of the formula by the franchisor is a separate product market. Even more interesting is the finding that the franchisor is not constrained by Apple. As I see it, it would be tantamount to suggesting that McDonald’s would not be constrained by vertically-integrated fast-food hamburger restaurants.
It may be a counterintuitive claim, but it is extensively supported by two sets of arguments. First (paras 483-496), I understand the decision as suggesting that the OS is one of many features found in a smartphone, which would allow McDonald’s (I mean, Google) to decrease the quality of its operating system without suffering the consequences. In other words, the Commission conducted an SSNDQ analysis. I find it great that we will be get some guidance about the robustness of this exercise (question: has the GC ever reviewed an SSNDQ assessment?).
Second (paras 497-559), the same arguments discussed above (including the fact that app developers would not be able to afford abandoning Android and Android users’ inability and/or unwillingness to switch OS and devices) are also advanced to rule out that vertically-integrated systems constrain non-integrated ones. So the same questions I raised above would be relevant here too.
In addition, the Commission emphasises the price differences between Apple and Android phones. Again, here my main question for economists would be: how decisive are these price differences? Similarly, I would also ask about the implications. How about high-end Android devices? Would it follow that further segmentation of the market is warranted?
As said above, I very much look forward to your reactions.
In case you were wondering (which I would understand): Alfonso and I generally comment on each other’s posts prior to publication. Not this time. I have not shared or discussed this post with him (nor will I share or discuss any of the subsequent posts in this series). So now you know the answer to the question ‘why are there even more typos than usual in Pablo’s post?’.
Chillin'Competition 
Pablo –
My two cents.
1. are app stores two-sided – absolutely. Are they transaction platforms – yes.
Does the apparent fact that developers are unwilling to switch necessarily mean that apple and android’s app stores don’t constrain one another? No. The two-sided nature means that this lack of substitutability wouldn’t matter if end-users would (consider) switching. But apparently they don’t. So separate markets and a lack of constraints across app stores makes sense to me if we accept the evidence on substitutability.
2. obviously an OEM cannot switch from android to using apple iOS. So apple imposes no direct constraint on the price (or terms of android). But if OS is important to end users (or developers) then it will impose an indirect constraint in the downstream – that is, android won’t be able to raise its price to OEMs (or lower its quality) because that will mean end-users switching from android phones to apple phones. But if end-users aren’t sensitive to changes in price (or quality) of the OS (apparently) then higher prices charged to OEMs won’t make much difference to how many android phones sold. in that case again apple won’t impose a constraint. So it’s more like the price of pepsi at macdonalds increasing, and consumers not caring enough to be willing to switch to burger king or a vertically integrated fast-food restaurant as a result, in the end perhaps it’s (in the eyes of end-users at least) a small piece of the price and the users reason for choosing macdonalds.
so for me the logic of the analysis certainly stands up, what matters then is the evidence on substitutability which I’m taking as given.
Chris
chris pike
4 October 2019 at 12:48 pm
[…] In der Zwischenzeit empfehlen wir den Digitalisti zwei Dokumente zur Lektüre: Zum einen hat die Kommission ihre Android-Entscheidung veröffentlicht, die viele „gold nuggets“ enthält, wie das ein ausländischer Kollege so schön formulierte. Pablo Ibanez Colomo hat mit der Auswertung schon mal begonnen. […]
SSNIPpets (34): In Vorfreude | D'Kart
4 October 2019 at 4:49 pm
Dear Chris,
Thank you so much for breaking the ice! And thanks very much for the extensive comments.
I am particularly interested in the first point. As I understand the decision, the Commission does not appear to claim that app developers are unwilling to switch. The point, as I understand it, is that app developers cannot afford to stop dealing with Android (paras 290 and 668).
And here my question would be: does it follow from the economic dependence vis-a-vis Android that Android’s app store is not constrained by Apple? This is the point with which I struggled in the decision.
I can understand that many suppliers may not be able to stop dealing with a large supermarket chain – maybe suppliers would not be able to forego the revenues coming from that large buyer. But I fail to see why it would follow that the large buyer in question would not be constrained by the other large buyer.
The conditions of competition may be terrible in such circumstances. But they could get even worse if the two large buyers decided to merge. Would the conditions of competition not be worsened if (assuming it is possible) Apple and Google decided to merge their app stores? If the answer is yes, would it not follow that the app stores already constrain each other?
On the second point, I guess my key question is how substitutability would be assessed in practice. I wonder, in other words, how the test would operate.
Thanks a lot again!
Pablo Ibanez Colomo
6 October 2019 at 6:32 pm
Dear Pablo
I would think, to use your supermarket analogy, that if a supplier cannot avoid dealing with a supermarket X then it also cannot switch from dealing with X to another supermarket Y. In which case supermarket X is under no pressure from the risk of suppliers switching away from it. As you suggest that doesn’t necessarily mean that supermarket X is not constrained by supermarket Y, but what it does mean is that if it is constrained it would have to be because users would consider switching. That means the key question is whether end-users consider switching or not. If they don’t then the supermarkets aren’t constraining one another and there’s no competitive constraint to be lost. So would an apple/google merger, or perhaps firm Z buying both app-stores reduce competition between app stores? On those facts I guess not (though that doesn’t mean apple/google wouldn’t be a problem for other reasons). The challenge i guess is that the app-store seems to be an aftermarket to the handset/device market, but that the app-store doesn’t figure much in users device purchasing decisions (apparently).
don’t know if that helps
look forward to reading more of your thoughts on the decision
chris
chris pike
7 October 2019 at 12:36 pm
Thank you for another excellent post.
My impression is that the error-cost framework that authorities want to introduce for digital abuse cases that would require incumbent companies to bear the burden of proving pro-competitive gains of the contested conduct (for the outrage of some commentators that this would actually constitute an inversion of the burden of proof), is making its way in a more subtle manner into the market definition, too.
Your reference to horizontal mergers is on point: “significant likelihood” that the the potential competitor would grow into an effective competitive force is sufficient to have significant anti-competitive effects. Interestingly, the Commission has used that reference to review recent innovation-heavy mergers (even though in relation to other aspects such as pipelines and other industries such as pharma and agro-chemicals).
In the literature on killer acquisitions (“don’t call it that way”), the ambivalence is recognised more overtly. The start-up targeted by the incumbent may not be an actual competitive constraint in the core market. Yet. Expanding potential competition to include the Target in the competitive assessment would meet merger control concerns. But at the same time it could also underestimate the market position of the incumbent. The EU digital / Crémer report admits the trade-off and proposes checking a “cannibalisation effect” in a “technological or users’ space”, which is broader than the relevant markets.
Which practical solutions then? I have two modest suggestions. On the one hand, enforcers could rely more on internal documents and how companies see each other for the definition of the markets, rather than on functionalities or (solely) demand substitutability (for which sources are not always reliable or at least consistent).
On the other hand, counsel could respond to the wave of digital enforcement by searching throughout the toolbox seeing until where enforcers have pushed the envelope for a specific case or field and demanding that the same is taken into account for another (case-by-case analysis, OK; but give me some predictability). Independently from the outcome of the single cases, we are all on the same side, seeking clarity on theories of harm, which necessarily start from market definition (or, as recent digital reports suggest, include market definition assessment, which cannot be considered on a stand-alone basis in the digital industries).
Gianni De Stefano
7 October 2019 at 10:31 am
Thanks a million, Chris and Gianni, for your comments! This is exactly the sort of debate to which I hoped the post would lead.
Chris: I am really grateful for clarifying that, taken to its logical consequences, the Commission’s position in Android implies that merging Apple’s and Android’s app stores would be unproblematic. This is a most counterintuitive and fascinating conclusion for a non-economist like myself.
But Gianni’s comment suggests that agencies’ approaches when assessing mergers may be different. I understand Gianni’s comment as suggesting that the analysis of competitive constraints in the area of merger control has differed somewhat from the analysis found in Android. This is really a fascinating empirical question to evaluate systematically!
Pablo Ibanez Colomo
7 October 2019 at 4:45 pm
Dear Pablo, spot on as ever. Market definition and dominance are likely to be a bone of contention in future academic, practical and judicial discussions of the Android case. The EC reasoning comes very close to a finding that a supplier is dominant over its own brand of products/services. This is inconsistent with the well accepted doctrinal principle whereby low levels of intra brand competition can lawfully coexist with intense inter brand competition. Put differently, there is no reason not to view the reported ecosystems competition between AAPL and GOOG as inter brand competition, and in turn assess it within a properly delineated relevant market comprising licensable and non licensable smart mobile OS. In that market, rational consumers make periodic purchases of semi durable goods – recall planned obsolescence and rapid product iteration – which constrain the ability of each ecosystem to undertake a profitable SSNIP of SSNDQ on the after market for the app store. Unless you see this competition in the primary market, you end up with the weird conjecture – that your discussion with Chris rightly points out to – that an AAPL/GOOG merger would be prima facie unproblematic absent horizontal overlaps. Of course, the EC approach works all fine if premised on a vision of antitrust essentially focused on short term allocative efficiency. The point then is to protect users from “lock in” manifest in app store market power (recall SCOTUS decision in Kodak or GC reasoning in CEAHR). But this is not how the EC decision in Android was reasoned, in particular when it alleged harms to innovation. Moreover, this would remain an odd case bc GGOG is the less expensive and higher app choice offer amongst the two ecosystems. Apologies for typos, writing from iPhone 🙂
Nicolas Petit
9 October 2019 at 4:02 pm
Dear Pablo, thanks for sparking the discussion, very interesting read.
I was actually hoping for the Commission to use this opportunity to develop a more comprehensive theory on the two-sided markets that you mentioned. Regarding market definition, some commentators think that the CEAHR decision by the CFI applies to those markets but the Commission mentions the case only once, and without elaboration. This decision could serve as a blueprint for the definition of multi-sided markets (maybe you want to have a look, T-427/08, paras. 105-107).
Also with regard to dominance, the EFIM decision by the CJEU could apply but the Commission did not mention it which I find curious (C-56/12, paras. 37, 12). Perhaps we will get a clarification when the appeals decision is published.
Maximilian Volmar
11 October 2019 at 12:42 pm
Regarding the first question:
1) Think of a merger between supermarket Chain X and Y, where X owns all stores located in the western part of the country, and Y have all stores located in the eastern part of the country. For end users there is no substitutability between store X and Y – the chains exert zero competitive pressure on each other in the short run. There are no short run unilateral effects.
2) The merger can still be problematic due to loss of potential competition or harm related to buyer power.
3) I don’t think the Commission claims that there is absolutely no competitive pressure from the App store. With differentiated products, situations can arise where X has enough market power to be considered dominant, but a merger with its only competitor Y would make the situation even worse. Specifically, X could in some settings profitably make a large price increase after merging with Y, even if few customers would switch away from X in response to a small price increase. When assessing dominance, I believe the share of customers switching away from X in response to a price increase above some benchmark is key. When assessing unilateral merger effects, the share of customers that would divert between the merging firms in response to a price increase is key.
Andrew
11 October 2019 at 8:52 pm