The General Court confirmed on 14 September 2022 [1] the EC finding in 2018 [2] that Google had abused its dominance by tying its Android operating system with its app store suite, the Google Play Store (“GPS”), thus leveraging power from GPS (a service that was “must have” to OEMs equipping their devices) to cement the monopoly of Google Search (and the Chrome browser) on mobile devices. The case started in Europe informally in 2014, formally in 2015, the infringement decision was issued by the EC in July 2018, and it has taken another 4 years to get through this first round of appeals. Mercifully less than the Google Shopping case, which took over 10 years from its uncertain beginnings (and early settlement attempts) to last November’s judgment upholding the Commission’s 2017 decision – but still a chunky amount of time.
It was pretty existential for the EC to win this endorsement from the Court, at a time when the Court had been patchy in its support (see Intel, and Qualcomm, and O2/Three judgments among others [3]) in ways that had inevitably affected the institution’s appetite for taking risks. Relief at DG COMP has been palpable, and understandably so. That said, the substance of the case was strong, and it was very hard to imagine that the Court could have pushed back on substance: it is a clear case of exclusionary abuse based around tying, that does not create angst to the traditionalists because it is squarely within the confines of the Microsoft precedent; in addition, it is grounded in robust economic theories of exclusionary tying, updated to a world of two-sided platforms and free services on one side. The Court mostly endorsed the Commission’s analysis and use of evidence, pushing back on Google’s often gravity-defying efficiency justifications. While the Court did reject one of the grounds of infringement (exclusivity), this was the least important part of the case (and the real significance of this rejection is that it confirms the Courts are not deferential to the Commission’s application of price-cost tests and will engage with them in full – a sign of progress to some, to me however also a dubious legacy of the “more economic approach” that has been oversold as “science” but in fact has weakened enforcement).
In addition to a few observations on the case and the judgment itself, there are at least three broader reflections arising for me in the wake of the Court’s endorsement. First, while the judgment is a necessary boost to the posture and appetite of the agency, the case is also more than most emblematic of the inherent challenges and failures of ex post enforcement. The reality is that Google’s practices had succeeded in moving the Search monopoly from desktop to mobile pretty much entirely by the time the case was formally started by the EC (and before it, the Russian FAS). What was underway already by 2008-10 was a significant “pivot” into a new technology (desktop to mobile) which could have provided in principle a major opportunity for challenging the incumbent in the previous technology. Yet monopolisation of mobile search was largely accomplished by the time the investigations got going, and the investigations did not change or slow down the conduct. This reminder of the slow grinding of the antitrust enforcement wheels is especially poignant when – as now – there are major technological pivots underway, for instance the emergence of potential new paradigms such as immersive reality and “horizon worlds”. [4] What can ex-post enforcement hope to do here, and should the agencies and the Courts not be much more focused on tools like merger control to pre-empt deals which will facilitate “monopoly pass-on” and early platform envelopment?
Second, the case is also emblematic because, notwithstanding a great theory of harm, fines, and a “remedy”, it has led to absolutely no change on the ground – the “remedy” could not realistically undo the harm in a digital market which has tipped. And the Court judgment will of course not change that.
Third, the slow grinding of the antitrust wheels in this and other cases is itself the reason that digital enforcement has strongly shifted – at least in Europe – to a huge (and at the moment highly uncertain) experiment in digital regulation. Is this more likely to succeed?
1. Great substance: a tying case for multisided digital platforms
I will not take time retelling the theory of the case, as there are multiple commentaries out there that do just that. In a nutshell, though, the EC found that Google’s offer of providing its app store suite (Google Play Store, or GPS) for free to OEMs that agreed to place Google Search as a pre-installed default on the device, along with related restrictions, amounted to tying. This is because the availability of Google’s “must have” GPS to Android OEMs was made contingent on them adopting Google Search as the pre-installed default at every “point of entry” on the device (this is what the Mobile Application Distribution Agreement, or “MADA”, involved). Other search engines were foreclosed as they could not realistically replace Google Search (even though there was no explicit veto to installing other engines: default position and pre-installation were enough). Additional anti-fragmentation agreements (“AFA”) prevented OEMs from developing alternative versions of the operating system with different rules on the position of search engines. Google’s technological advantage in search and its dominance of online advertising could be threatened only by a competing search engine with enough scale to improve its search algorithms, and attract enough search queries to expand revenues in online ads. Google’s conduct made this impossible and the Court essentially agreed, rejecting Google’s contrived efficiency defences.
Perhaps a small nugget of antitrust curio is that when the case started the focus was on dominance in Search as the obvious locus of Google’s power from which leveraging would be expected to occur. Regulators were scratching their heads about how to formulate the theory of harm: leveraging from where to where? Search was the natural starting point, but rivals were foreclosed in Search. So what power was being leveraged where? The penny finally dropped and it became apparent that the right economic case was the other way round: dominance was being leveraged from the GPS store, which was the real “must have” input for OEMs, into Search by inducing OEMs to pre-install Google Search and place it as the default as a condition for getting GPS (and Android) for free. This formulation of the theory was first articulated in the case brought by the Russian FAS following Yandex’s complaint and was then replicated in the EC case. Thus, while the case was about tying, in the Microsoft tradition, the direction of the tie was initially counterintuitive.
Substantively, the case matters also because it attracted significant interest by economists on the anticompetitive mechanism at play, prompting work that extended traditional theories of harm based around tying, network effects and naked exclusion to incorporate multisidedness and zero-price constraints in ad-funded business models. [5] This matters, as enforcement in digital markets requires substantive work developing credible mechanisms that agencies can rely on and will be harder to refute and discredit by the defendant’s consultants and in the mainstream economic discourse. The decision does not dwell much on it but there were discussions with the Chief Economist Team (led at the time by Tommaso Valletti), and economic papers have formalised the mechanism in ways that increased the credibility of the case.
Of course, we have long known from the traditional antitrust literature that tying a primary product supplied by a monopolist (here Google, as the dominant provider of the GPS suite) with a secondary product (here its search engine) can be an aggressive strategy to foreclose entry in the secondary market. Given the high fixed costs of developing a search engine and reaching a viable scale in search, tying that reduced the profits of an entrant could deter entry and allow Google to pass on its monopoly to mobile search. This traditional mechanism was extended and adapted to the specific facts of the case, with analyses showing formally that by tying the GPS suite to Google Search, and committing to distributing Android with the GPS suite for free, Google prevented even a more efficient search engine from getting traction. Providing the GPS suite to all OEMs for free in addition to the Android operating system had additional value to consumers because it included unique applications such as Google Play and YouTube that were otherwise unavailable. Google monetized the suite on the advertising side, where it extracted rents through in-app advertising and revenue sharing agreements with app developers. Even if a rival search engine offered equivalent (or even better) quality to Google Search it just could not outbid Google, be pre-installed on Android devices, and challenge its dominance in search.
Note the tie was essential to the anticompetitive mechanism: without the tie, i.e. if Google had provided the GPS suite as a standalone product and there was competitive bidding for exclusive pre-installation of search engines, an entrant with a superior search engine could in principle have outbid Google Search for pre-installation. But by making GPS available for free if and only if Google Search was pre-installed as the default search engine, Google could ensure OEMs used its bundle (and compensated them also through revenue sharing). This strategy deterred entry, was profitable for Google, reduced the price-quality ratio of Android devices, lowered welfare (by deterring the use of a better search engine) and reduced consumer surplus. [6]
In short, the essence of the case – which the Court accepted – was the pre-installation and default positioning tied with the supply of the Android system for free. The case showed that existing robust foreclosure mechanisms can be extended to the specificities of digital markets, multisidedness and zero prices to consumers. There is no need to appeal to exotic theories, the specific features of the market reinforced an economic mechanism that was mainstream.
2. AECT, again? The Court’s appetite for price-cost tests
The one part of the EC case that the Court struck down is the claim that Google’s Revenue Sharing Agreements (“RSA”s) with OEMs gave rise to exclusionary effects. That Google signed agreements with certain selected OEMs to share revenues in exchange for exclusivity (i.e. if they agreed not to pre-install competing search engines anywhere on the device), was not in fact strictly an essential part of the theory of harm. The core of the theory of harm was the power of pre-installation, default and tying. Exclusivity remunerated through “revenue sharing” was more of a “reinforcement mechanism” that facilitated the foreclosure of others, with effects similar to exclusive dealing. But it was not the main engine for the exclusion.
The EC opted to throw this claim in the mix at the time, including an estimate in the decision of the “demand” accounted for by OEMs that entered into RSAs with Google. It concluded that competing search engines could not have offered OEMs enough revenues to compensate for the loss of Google’s portfolio-based revenue share payments across their Android devices. The Court however struck down the notion that the RSAs had exclusionary effects, essentially because the “as efficient competitor” test (AECT) was not robust, and in particular the estimated “coverage” of the demand accounted for by OEMs with RSAs was not enough to conclude rival engines were excluded from a large enough portion of the market.
The real significance of this is more in the interest and appetite the Courts appear to have developed for engaging with price-cost tests like “as efficient competitor”. The Intel infringement decision has been famously overturned on appeal by the ECJ last January, 13 years after the EC decision, on grounds that the Commission had been sloppy in applying its AECT. This is notable because on the one hand, it signals that AEC tests and price-cost tests in general will be taken seriously if they are relied upon at all – as the Courts apply themselves these days with some gusto to these calculations and do not hesitate to conclude the Commission had been imprecise. This is fair enough, and some will applaud this as a victory for rationality and “economic analysis”, forcing the Commission to “take the defendant’s economic analysis seriously” (comments to this effect are out there as the main “takeaway” of economic consultants and defence lawyers who know very well what economists are useful for).
On the other hand, to me it is also a concerning prospect. These price-cost tests are typically assumption-driven, and most often fragile: they rely finely on estimates of “contestable share”, margins, demand. The impression that there is reliable “science” behind them, and that we economists can estimate with reliable precision what the portion is of “market” demand which is taken off the table by certain practices (discounts, exclusivity) and whether that is enough to allow others to survive is one of the great successes of the “more economic approach” as pushed by economic consultants over the past 20 years. “We know how to do it, it’s scientific”, we said. Except it is not, and much of this is at best indicative but certainly not dispositive. Of course, economists have peddled these tests because they were working for defendants and the tests are deeply pro-defendant: who can be “as efficient” as a large incumbent? Ah, but still, the test says if you are not “as efficient” then you don’t deserve to live. This is a very deeply ideological stance which has been camouflaged under the guise of “science”, and the fact the Courts are putting significant weight on these analyses in these cases is problematic in my view. Because it will serve as a deterrent to agencies bringing cases with a price-cost aspect. The Android case did not depend on this, but other cases very well do. This will likely severely weaken the appetite for enforcement further in any cases involving exclusivity of sort.
3. Enforcement at times of technological pivots
The deeper reflections prompted by this case, however, are to me again about the challenges of ex-post enforcement in this space. This is not new, of course: we have lamented the slowness of enforcement for a long time – the product of complexity no doubt, but also of the incredibly elongated cadence of these cases and the grotesque exploitation of opportunities to delay and defer by defendants. The Shopping case saw an entire industry (comparison shopping) long dead by the time one got to the EC decision, never mind the Court judgment 4 years later.
But Android feels much worse. Because what was really happening in Android was that the practices in question (providing the GPS suite together with the Android OS to OEMs without charge) were adopted much earlier and were crucial to convincing OEMs to adopt Android when first launched as well as to persuade consumers to opt for Android devices. Google had realised as early as 2005-6 that the world was switching from the desktop to mobile devices, and it was this fundamental pivotal moment that could have been a threat to its search monopoly as the world shifted to a new form factor. With other proprietary OSs at the time such as Symbian and Windows Mobile (and their application stores) as the main options for OEMs, Google bought Android but had to convince OEMs to adopt it. By committing to an open-source OS with a free app store and free APIs, it got app developers to build apps for Android and created the basis for the quality advantage of its app store and of Android devices. It could do this and accelerate adoption because it could monetize its services on the advertisers’ side, by sharing revenues with app developers through in-app advertising powered by Google products, and through the promotion of the Google environment generating revenues from search advertising.
Google was thus able to “pass on” the Search monopoly from the desktop to the mobile and avert all threats to itself at the most perilous time. By 2014, when the EC started its informal investigation, it was all over. The search monopoly was established on mobile. In fact, as the EC had been busy on Shopping, this passed everyone by. The practices continued but it was all done and dusted by the time the case started. This is what has also happened in IoT and connected cars: while the regulator is busy looking narrowly at one case at a time and trying to train its limited resources on it, the monopoly is being passed on the next thing. Cars, appliances. Regulators are just starting over there, but it is already late.
What all of this points to is that (as we know) ex post enforcement will never stand a chance of meaningfully addressing anticompetitive conduct in these spaces, unless it were possible to hugely accelerate the process. But this lag is particularly dramatic, and serious, in situations where there is a major technological pivot to a new form factor underway, and while enforcement is looking the other way incumbents are able to swing their monopoly into the new space.
One implication is that merger control needs to be looked at as a major tool here. We cannot wait until entrenched monopolies are formed, moats built, and all chances of competition beaten back, to engage in ex post enforcement. [7]
Think most obviously of the current aspirations for immersive reality as a new computing form factor. Meta and others have been making multiple app acquisitions in this space, with the declared aim of growing a major position into this new reality. Meta’s history in particular involves serial acquisition of “user -generated and -driven” social networking platforms (Facebook, WhatsApp, Messenger, Instagram), and it has made no secret it is now focused on “horizon worlds” (spanning both augmented and virtual reality), for which it is making hardware too. Meta’s business model aggregates consumer engagement in all things social - chat, video, video phone, timelines, video, images etc., in order to sell advertising, and interactive experiences are sources of some of the most engaged ad inventory there is. So, while this is of primary focus to incumbents like Meta, and it is centrally on Meta’s roadmap, in this “pivot” there may be a window for competition.
But how is it going to be possible to ensure through ex post enforcement that at this time of key paradigm shift incumbents like Meta will not inevitably occupy the new space of immersive reality on horizon worlds? Acting on ex post complaints will be too late. If we allow Meta (and others) to acquire more social experiences on the new form factor, while they dominate key media already, we may well be foregoing new independent experiences gaining independent critical mass. So, if we think augmented and virtual reality are a likely new paradigm to engage consumers on social experiences, and we are concerned this will be occupied early, we should be taking a clear stance on deals which are intended to accumulate social media experience in the new paradigm. The riposte that this is “speculative” does not cut it. We should pick our rails: it is not good to say that ex post enforcement is always too late and then balk at tighter merger control because “it’s too speculative”. Unfortunately, this seems to be the approach in recent Court decisions in the US.
4. The remedies “fiction”
The next reflection is how utterly pointless the remedies theatre is. Like the Shopping case just before, the Android case ended with a “cease and desist” order, in effect “stop what you are doing, and don’t do it again in a way that leads to equivalent outcomes”. A “cease and desist” decision can only contain very limited indications as to what an appropriate remedy could look like. In the particular case, the EC mandated Google not to tie the Play Store with Search or engage in “equivalent conduct”. [8] But in practice, Google responded by engaging actually in a “de facto” tie: no longer formally tying the Play Store with Search (and Chrome) but offering the former to OEMs at a positive price and the latter at an equivalent discount. As a result, OEMs continued to be able to pre-install GPS as effectively at zero cost, on condition that they also pre-installed the Search app (or Chrome) as default. Following an outcry by complainants and observers, the “proposed solution” Google offered instead was an auction for “other” search engines to get a place in the range of choices the owner of a new phone in Europe will face on the homepage when getting the device out of the box. In practice, however, Google is always in the list while others (Bing, Duck Duck Go) had to pay for inclusion. And though a few names made it to the list, nothing is happening on the ground in terms of consumers actually choosing to pre-install alternatives to Google Search.
This is because players that are nascent or that have been weakened and marginalised as a result of years of the conduct will simply not gain visibility through auctions. Relying on consumers to then spontaneously select weak alternatives to the super dominant choice is not going to get us places given all we know about behavioural factors. Lack of familiarity with alternatives continues to drive consumers towards the dominant firm. A fortiori if the alternative is still in some measure disadvantaged by less attractive placement and other constraints.
The Court does not engage of course with the failure of the EC remedies. But this is an endemic failure, because it is simply the case that when markets have tipped, they cannot be “untipped” through these remedies. One more reason why the current prospects of ex-post enforcement leading to “more competition” in these spaces are bleak in my view. Multiple enforcement cases are currently lying in this “no-man’s land” where remedies are being slowly and very incrementally negotiated, with no real prospects of achieving more than perhaps a bit more “fairness” at the margin with platform dependants. But more true competition? I do not see it.
5. So regulation, right?
The pivot to digital regulation in the past two years, with the adoption of the DMA in record time has reflected deep frustration with all of the above. The Court judgment on Android confirms the case was good law, and it was also good economics, and yet we are no closer to creating competition in mobile search. Google remains a total monopoly.
Ex ante regulation is hoped to reach deeper and sooner. In reality, it is difficult to see clearly past the current early stages of implementation – “designation” of gatekeeping businesses is still underway, and then we will have the whole “compliance” effort unfolding (with predictable legal challenges at multiple steps along the way). With rules that are broad and not platform-specific, and cannot really anticipate how power will be swung where, is there real scope for pre-empting platform annexation, cascading market power and envelopment in new spaces? It continues to be the case in my view that there may be achievable targets in terms of greater “fairness” in dealing with platform dependants (sellers or developers) but the prospects for creating actual competition that is effective are limited. [9] Even more so when considering the massive investments that are required to create, for example, an alternative offering in immersive reality: who can realistically challenge Meta here, if not another giant? What “entry” are we contemplating?
We are better off with regulation than without, for sure. Although we do not want to see regulators meddling with product design for instance, and we can anticipate any effort to do so to be heavily resisted. I am sceptical the kind of pre-emptive conduct that Google so effectively implemented in Android to swing its monopoly to the new environment can be pre-empted through the current general ex ante rules – unless these are truly complemented with a platform-by-platform look at plans and strategies.
Conclusions
The Android judgment was widely expected to uphold the EC decision, which was good law and good economics. It is nonetheless a needed boost to the agency after recent court pushbacks. But it is also a poster child for how ex post enforcement finds it hard to prevent the “swinging of monopoly”, and this is an especially grave loss at times of pivotal technological change. It is also a poster child for how markets which have “tipped” cannot be “untipped” with remedies. The ability of ex ante regulation to catch the “swinging of monopoly” early and effectively remains to be seen. Merger control could be an effective tool, though moving away from the engrained presumption that “mergers are good” and in the main “procompetitive” will take some turning around as well. Overall, the legacy of Android is that enforcement against the conduct of certain digital giants continues to face major challenges, unless we succeed in drastically cutting down investigation time (yes, “rights of defence”, but much of it is also wasted time). We need to interrogate realistically objectives, and deliverables.
