As readers of this blog will know, on Tuesday 10 November 2021 the General Court (“GC”) delivered its judgment in case T-612/17, Google and Alphabet v Commission, largely upholding the Commission’s 2017 decision in Google Search (Shopping) (AT.39741). This is a landmark (and very aggressive) judgment and a major vindication for the Commission.
By way of reminder, in the contested decision the Commission had found that the more favourable positioning and display by Google, in its general search engine results pages (“SERPs”), of its own comparison-shopping service (“CSS”) compared to competing comparison-shopping services infringes Article 102 TFEU. The Commission imposed on Google a fine of approximately EUR 2.4 billion. Shortly thereafter Google filed an action for annulment, arguing among others that its conduct was not likely to have anticompetitive effects, amounted to competition on the merits, and in any event was objectively justified.
Even before the contested decision was adopted in 2017, the Commission’s investigation had sparked a massive debate in the literature over “self-preferencing” and whether it counts as a distinct form of abuse (and if yes, under what conditions it should be considered anticompetitive) or whether it should be subject to the strict Bronner conditions for a refusal to deal (see e.g., Vesterdorf’s paper from 2015 and Petit’s reply, or Ibáñez Colomo’s more recent paper on self-preferencing). Last Tuesday, the GC confirmed that self-preferencing (which it referred to as “favouring” and “internal discrimination”) can constitute a distinct form of abuse under Article 102 TFEU, there being no need to satisfy the strict conditions laid down in Bronner. But this is only one part of the Court’s ruling. In its (very) long judgment, the GC makes a series of points of principle that could have wide-ranging implications for EU competition law and Article 102 TFEU in particular, and which are not necessarily limited to tech cases. In this post, I would like to further explore some of these points and share my thoughts. As always, comments are most welcome (I bet that people will have different interpretations of certain parts of the judgment, and I am looking forward to the debates). Apologies for the long post but this is a massive judgment.
On leveraging and “competition on the merits”
The first threshold issue of law the GC examines in detail is under what conditions leveraging is anticompetitive. This part of the judgment (paras. 139-198) includes an interesting discussion of “competition on the merits,” a concept which has always been a bit nebulous under EU law.
By way of background, Google had claimed that its practices were quality improvements that fell within the scope of “competition on the merits” and could not be treated as abusive.
In response, the GC started by recalling well-known EU case law on the “special responsibility” of dominant undertakings (para. 150), the notion of abuse (para. 151; adopting the formulation of the CJEU in Post Danmark I that refers to consumer harm), the fact that the list of abusive practices in Article 102 TFEU is not exhaustive (para. 154), and the fact that exclusionary effects are not necessarily detrimental to competition, as competition on the merits may lead to the marginalisation and departure of less efficient competitors (para. 157; citing the CJEU’s seminal judgment in Intel). Interestingly, the GC states that the abuse may take the form of an “unjustified difference in treatment” (para. 155; citing GT-Link, Aeroports de Paris, and Irish Sugar, while reference is also made to the general principle of equal treatment under EU law).
The GC then observes that the mere extension of a dominant position to an adjacent market (often referred to as leveraging) does not suffice to characterise such conduct as anticompetitive, even if it leads to disappearance or marginalisation of rivals (para. 162, citing Intel and Post Danmark I). Leveraging is a generic term that may capture various practices, such as margin squeeze, loyalty rebates, or tied sales (para. 163). Leveraging as such is not prohibited (para. 164), but may infringe Article 102 TFEU if certain additional conditions are met. What are these conditions?
- First, the conduct in question has anticompetitive effects (even if only potential, as discussed below).
- Second, the conduct in questions departs from normal competition (competition on the merits) (see paras. 174-175 and 195).
The question of course is how to determine whether conduct falls outside the scope of competition on the merits. In the present case, the GC was satisfied that the Commission had shown that Google’s conduct departed from competition on the merits, on account of three specific factors (“relevant circumstances”), namely (i) the importance of Google search traffic for CSSs; (ii) the behaviour of users when searching online (users tend to concentrate on the first search results); and (iii) the nature of Google search traffic as being not effectively replaceable (paras. 169-175).
The GC did not stop there, however, even though in para. 175 it seems to consider that these three circumstances sufficed to characterise Google’s conduct as falling outside the scope of competition on the merits. The GC made three further points to support the conclusion that Google’s conduct deviated from normal competition:
- Incentives of a search engine: Google’s self-preferencing conduct seemed to involve a certain form of “abnormality,” in that a search engine is in principle “open” to results from external sources (in fact this is how it generates value), which distinguishes it from other infrastructures consisting of tangible or intangible assets. For a search engine, limiting the scope of its results to its own “is not necessarily rational, save in a situation, as in the present case, where the dominance and barriers to entry are such that no market entry within a sufficiently short period of time is possible in response to that limitation of internet users’ choice” (para. 178). The GC seems to consider that, in the absence of market power, a search engine would not have the incentive to promote its own specialised results over those of rivals.
- Legislation: The GC referred to EU legislation imposing on internet access providers a general obligation of equal treatment, noting that such an obligation on the upstream market for internet access providers “cannot be disregarded” when analysing the conduct of an operator like Google on the downstream market, given Google’s “undisputed ultra-dominant position” and its associated “special responsibility” not to impair competition (para. 180). The GC notes that Google has a higher “special responsibility” in light of its “superdominant position” in general search and being a “gateway to the internet” (paras. 182-183).
- Change of conduct: Finally, the GC noted that a departure from competition on the merits of the conduct at issue was all the more obvious as “it follows a change of conduct” on the part of Google, in that the way Google displayed search results changed when it developed its own CSS (para. 181).
As for Google’s arguments that its conduct amounted to product improvement, the GC relegated these claims to the stage of efficiency considerations (paras. 188 and 266). The GC also dismissed as irrelevant the fact that Google may have intended to compete on the merits, in line with the settled case law of the CJEU (paras. 257 and 263, citing Tomra).
All in all, the GC appears willing to consider as anticompetitive leveraging practices that do not necessarily fit into a traditionally recognized box under Article 102 TFEU (such as tying or margin squeeze). In this case, leveraging took the form of favouring. At the same time the GC also placed some limits; it did not accept that any kind of leveraging is prohibited under Article 102 TFEU, and it noted that the Commission cannot rely solely on the effects of the conduct either. It must in addition be established that the leveraging practice falls outside the scope of competition on the merits. I have to say this part remains a bit puzzling. While in that case the GC considered that this last condition was satisfied, it did not provide general guidance that could be easily transposed to other cases. After all, the “relevant circumstances” which the Commission took into account (the importance of Google search traffic, its nature as not being replaceable, and the behaviour of users), seem to be very closely related to the effects of Google’s conduct. Now, the GC took into account a number of additional points (incentives of a search engine, legislation, change of conduct), hence it seems that the Court relied on a web of evidence. One could say this is in line with Intel where the CJEU stressed the need to evaluate the firm’s conduct in light of “all the relevant circumstances.”
On Bronner and essential facilities
The next important legal point is whether “self-preferencing” is prohibited only if it meets the strict conditions of Bronner. Google had argued that the Commission wrongly refused to apply the Bronner conditions (and in particular, the requirement to show indispensability), when in reality the case was about a refusal to supply as the Commission in practice imposed on Google a duty to supply rival CSSs. The GC disagreed with Google, holding that Bronner did not apply. In my view, this was expected following the CJEU’s judgment earlier this year in Slovak Telekom, as explained here.
After recalling the case law on essential facilities under EU law and the rationale for the indispensability criterion laid down in Bronner, the General Court notes that what is at issue in the present case are “the conditions of the supply by Google of its general search service by means of access to general results pages for competing comparison-shopping services” (para. 219). The contested decision does seek to provide rival CSSs with access to Google’s SERP and ensure equal treatment with Google’s CSS (para. 222). Next, the GC observes that Google’s SERP has characteristics akin to those of an “essential facility” (para. 224), and in practice the Commission found Google’s traffic to be indispensable for competing CSSs (paras. 225-227).
Most importantly, the GC considers that the practices at issue should be distinguished from the refusal to supply in Bronner (para. 229). Not every issue of access means that the Bronner conditions have to apply (para. 230). The GC gave the following reasons why the Bronner test does not apply in the present case:
- A refusal to supply implies that (i) there is an express refusal to an access request (to be contrasted to an implicit refusal); and (ii) the impugned conduct triggering the exclusionary effects lies “principally in the refusal as such, and not in an extrinsic practice such as, in particular, another form of leveraging abuse” (para. 232).
- There are a number of cases that raised issues of access to a service but where it was not necessary to show indispensability per the refusal to supply case law, namely cases of margin squeeze (citing TeliaSonera) and tying (citing Microsoft) (para. 235). Reference is made to the infamous dictum from TeliaSonera (para. 55) according to which it cannot be inferred from Bronner that the conditions to establish a refusal to supply must also apply when assessing the abusive nature of conduct which consists in “supplying services or selling goods on conditions which are disadvantageous or on which there might be no purchaser” (para. 236). While in the past it had been argued that this dictum was limited to cases of margin squeeze (the type of abuse at issue in TeliaSonera), both the CJEU in Slovak Telekom, and now the GC in this case, have expanded this dictum to other practices (in this case, self-preferencing).
- The conduct at issue consisted in internal discrimination (para. 237). The case is thus not concerned merely with a unilateral refusal of Google to supply a service to rivals, but with a difference in treatment (para. 238). The Advocates General of the CJEU have consistently distinguished between cases of difference in treatment from cases of refusal to supply (para. 239). Moreover, the practice at issue consisted in “active” behaviour in the form of positive discrimination (as opposed to “passive” refusal to deal; citing the CJEU’s 2021 judgment in Slovak Telekom) (para. 240).
As for Google’s argument that the contested decision required it to supply its rivals, the GC stressed that there can be no automatic link between the criteria used for classifying an abuse and the corrective measures taken to remedy it (para. 244). The applicability of the Bronner criteria does not depend on the remedies ordered by the Commission (para. 246).
The GC has essentially limited Bronner to cases of explicit refusals to supply a service or provide access to infrastructure. This is in line with the approach of the CJEU in Slovak Telekom. This has created a paradoxical situation, where it is better for a dominant firm to refuse rivals access outright than provide access to its service or infrastructure but on unfavourable terms (in that the standard for intervention is lower in the latter case).
On anticompetitive effects
Google had argued that the Commission engaged in pure speculation about the anticompetitive effects of the practices at issue. The GC largely upheld the effects analysis of the Commission. By way of background, the Commission’s effects analysis comprised two parts: first, a very detailed examination of the consequences of Google’s conduct on the traffic flowing to rival CSS; and, second, an examination of the anticompetitive effects flowing from such a reduction of traffic. Only the latter are stricto sensu anticompetitive effect; however, the GC noted that finding consequences on the traffic was a prerequisite for establishing anticompetitive effects. In other words, the GC drew a distinction between effects on traffic and anticompetitive effects. While showing the former was necessary, it was not sufficient to establish an infringement; the Commission also had to show anticompetitive effects, understood as effects on the ability/incentive of rivals to compete with Google (as inferred from para. 451).
First part – effects on traffic and causal link
Google argued that the Commission had failed to establish a causal link between the conduct in question and the decrease of traffic to rival CSSs. According to Google, the decrease in traffic (which Google did not contest) was solely attributable to the operation of Google Search adjustment algorithms.
The GC disagreed, noting that the Commission took issue with the combined effect of two practices of Google, namely (1) Google displaying its CSS in a prominent and eye-catching manner in its SERP without that CSS being subject to adjustment algorithms used for general searches, and (2) at the same time displaying rival CSSs on its SERP in the form of blue links prone to demotion due to adjustment algorithms (para. 369).
As a result, the only counterfactual scenario Google could have put forward would have been one in which none of the above practices was implemented (para. 376). Google did not do so.
Interestingly, the GC notes that in a case like this, identifying a credible counterfactual scenario may be an arbitrary or even impossible exercise (para. 377). The Commission cannot be required either spontaneously or in response to a counterfactual analysis put forward by the defendant, systematically to establish a counterfactual scenario (para. 378). This would also oblige the Commission to demonstrate that the conduct at issue had actual effects, whereas it suffices to establish potential effects (para. 378).
This is followed by a very interesting discussion on when a causal link is established. In para. 382, the GC seems to lay down a two-stage test. First, it is for the Commission to establish such a causal link, by showing that (i) there is a correlation between the conduct and the actual evolution of the market; and (ii) there is additional information – e.g., assessments of market participants, their suppliers, their customers, professional or consumer associations – that is capable of demonstrating the causal link. Second, it “may” in some circumstances be appropriate for the dominant undertaking to bear the burden of putting forward information casting doubt on that causal link.
Applying this test to the present case, the GC considered that the Commission had established such a causal link, in that first, the Commission (i) had shown there was a decrease in overall traffic to rival CSSs; and (ii) there was information derived from rival CSSs attributing such decrease to Google’s adjustment algorithms, and second, Google failed to provide evidence to the contrary (para. 394).
Does this amount to a reversal of the burden of proof? The answer is probably no, since the Commission still has to complete the first step and show at least a correlation between the practice and the evolution of the market, supported by additional information. While some may argue that this is far from amounting to causality, we have to be pragmatic. Proving causality in a scientific manner can be impossible. The GC seems to acknowledge that and sets a lower standard based on correlation, while allowing for the dominant undertaking to then provide information casting doubt on the causal link.
Second part – anticompetitive effects
The GC starts by recalling well-known case law on the concept of effects under Articles 101 and 102 TFEU. It is interesting to note that the GC puts forward a rather broad interpretation of Intel, extending it to cases beyond exclusivity payments: if the dominant undertaking disputes, with documented evidence (e.g., information on the actual development of the market), that its conduct was capable of restricting competition, then the competition authority must analyse “all the relevant circumstances” (paras. 439-440).
The GC notes that for the purposes of establishing an infringement of Article 102 TFEU, it suffices for the Commission to show there is a potential anticompetitive effect (para. 438, citing Deutsche Telekom and TeliaSonera). There is no need for the Commissionto identify actual exclusionary effects (para. 442, citing TeliaSonera and Telefonica; implicitly rejecting Google’s argument based on the GC’s Article 101 TFEU judgment in Servier, according to which when an agreement has been in place for years, the Commission needs to establish actual effects). Nor, a fortiori, is the Commission required to show the possible negative consequences from a reduction of competition, e.g., price increases or less innovation, since the weakening of competition is highly likely to have such consequences (para. 443; in effect, the GC states that a reduction of competition is presumed to have such negative effects).
In the present case, the GC held that the Commission identified several potential anticompetitive effects in the markets for CSSs, in that according to the contested decision the practices at issue were capable of: leading rival CSSs to cease their activities; reducing their incentive to innovate; reducing Google’s incentive to innovate; and thus, affecting the competitive structure of the market (para. 451).
On the other hand, the Commission’s reasoning was speculative as regards potential anticompetitive effects in the national markets for general search services, hence the GC annulled the decision on this point (paras. 456-459). This did not affect other parts of the decision, nor the amount of the fine since the Commission did not take the value of sales on the general search market into consideration in order to determine the basic amount of the fine (para. 660).
On the as efficient competitor test
CCIA (intervener for Google) had argued that the Commission failed to demonstrate that rival CSSs were as efficient as Google, thus falling foul of Intel. The GC, in a rather aggressive part of the judgment, states that the use of an as efficient competitor (“AEC”) test is warranted in the case of pricing practices, but not in non-pricing practices (para. 538). The use of such a test, which involves comparing prices and costs, did not therefore make sense in the present case, since the competition issue identified was not one of pricing (para. 539).
While the above has been well known ever since Post Danmark II(recall the dictum that the AEC test should be regarded as “one tool amongst others”), the GC went even further to state that an AEC test could not in itself determine whether Google’s conduct was anticompetitive. The reason is that an AEC test may produce objective results only if competition is not in fact distorted by anticompetitive conduct (para. 540). It was thus “impossible to know” whether Google “was ‘more efficient’ that the other comparison-shopping services” when its practices were capable of distorting competition (para. 541).
On competition law and business models
Some scholars such as Ibáñez Colomo have argued that competition law has traditionally been careful not to interfere with the business model of firms and how they choose to monetize their business, save for exceptional circumstances (essentially when the Bronner conditions are met). I have not found this view particularly convincing, since antitrust intervention may well challenge a business model even in the absence of exceptional circumstances (see also the discussion here). The GC took a position on this issue in response to Google’s argument that the contested decision wrongly compared the treatment of Shopping Units (that is, product ads), with the treatment of free generic results, as these are not the same. According to Google, treating paid results differently from organic results is a natural consequence of operating a two-sided, ad-funded business model.
The GC held that the contested decision does not take issue with the two-sided business model of Google, in that the Shopping Units were part of Google’s CSS (para. 314). In any event, the GC unequivocally stated that “[i]f the funding model of an undertaking leads it, as in the present case, to take part in an abuse of a dominant position, there is nothing to preclude that funding model being caught by the prohibition under Article 102 TFEU. It is, moreover, a feature of many abuses of a dominant position that the aim is to improve an undertaking’s sources of funding” (para. 314).
This is a seminal judgment whose consequences will reverberate across the world. It will be interesting to see whether Google will appeal to the CJEU or stop here. Either way one thing is certain; the Commission has been emboldened in its Big Tech crackdown.
Photo by Pawel Czerwinski on Unsplash
3 thoughts on “General Court of the EU delivers landmark Google Shopping judgment (Google and Alphabet v Commission, T-612/17)”
Do you agree that the judgement signals a need for vertically integrated dominant companies ( such as google) to examine their pricing and QOS practices vis a vis third parties to ensure that they operate on a FRAND basis and don’t leverage their vertical integration?
Thank you for this excellent post – far more detailed than the other articles I have seen, which makes it an excellent complement to reading the judgment itself.
Re: “The GC has essentially limited Bronner to cases of explicit refusals to supply a service or provide access to infrastructure. This is in line with the approach of the CJEU in Slovak Telekom. This has created a paradoxical situation, where it is better for a dominant firm to refuse rivals access outright than provide access to its service or infrastructure but on unfavourable terms (in that the standard for intervention is lower in the latter case).”
This is a very interesting point and I agree it does seem potentially better for a dominant undertaking not to risking providing access at all. This may result in vertically integrated undertakings having more ‘closed’ levels of their supply chain. However, I am not sure it is “paradoxical”. My reading of the case law – but very open to other interpretations – is that if an undertaking has already *voluntarily* given access to a service/facility (even on unfair terms) it should not be able to subsequently benefit from the added protection of the Bronner line of case law, which concerns an *obligation* to deal where the undertaking would not otherwise do so (taking into account its right to property etc.).
Many thanks Charlie for your comment – apologies for not seeing this earlier! My reading of the case law is very close to yours. But at least intuitively I find it weird that it is better (from an antitrust perspective) for a firm to refuse to grant access to a service than granting access on unfair terms. Then again, one could say that this is also what margin squeeze is about.