Google Shopping and the Quest for a Legal Test for Self-preferencing Under Article 102 TFEU

Printer-friendly version

Abstract: Does a dominant firm abuse its market power in violation of EU competition law and, more specifically, art. 102 TFEU if it accords more favourable treatment to its own products or services than to those of its rivals? In answering this question in the affirmative and holding that self-preferencing constitutes a novel type of abuse of dominance, the European Commission's Google Shopping decision (case AT.39740) expanded the scope of art. 102 TFEU into unchartered territory. The Commission found that Google had abused its dominant position by guaranteeing its own comparison shopping website a more prominent placement on the result page of its general internet search engine than rival comparison shopping services. In November 2021, this decision was upheld on appeal by the General Court of the European Union (case T-612/17 Google and Alphabet v Commission (Google Shopping) ECLI:EU:T:2021:763). This Insight critically reflects on this watershed ruling. The General Court’s Google Shopping judgment poses once more the unsettled question of the exact boundaries of the concept of abuse of dominance under art. 102 TFEU. This Insight analyses different attempts by the General Court to redraw the scope of art. 102 TFEU by pinning down the constitutive elements of abusive self-preferencing. Although the Court considered various pathways to determine the legality of self-preferencing, it failed to articulate a clear legal test that establishes limiting principles as to when self-preferencing by a dominant firm violates EU competition law. In light of this finding, this Insight sketches an alternative pathway that would have enabled the General Court to ground the novel theory of harm of self-preferencing within a legally and economically sound framework.

Keywords: self-preferencing – case T-612/17 Google Shopping – Article 102 TFEU – Case C-7/97 Bronner v Mediaprint – unequal treatment – refusal to deal.

I. Introduction

On 10 November 2021, the General Court of the European Union handed down its long-awaited judgment in the Google Shopping case.[1] The ruling upheld one of the most controversial and consequential competition law decisions by the European Commission of the last decade. After seven years of investigation and three failed commitment packages, the Commission concluded in 2017 that Google had abused its dominant position in violation of art. 102 TFEU. At the heart of the Commission's finding of an abuse of dominance lay the novel theory of harm that Google had designed the result page of its well-known general internet search engine “Google Search” in a way that favoured its own Comparison Shopping Service (CSS) Google Shopping, while placing rival CSS websites at a competitive disadvantage. According to the Commission, Google's abusive self-preferencing consisted of two elements: Google was found to have i) consistently afforded its own CSS greater visibility on the result pages of its general search engine by displaying it amongst the highest ranked and most visible search results, and ii) simultaneously actively demoted competing CSS on its general search result pages to lower-ranked links and pages. The Commission not only sanctioned Google's conduct by imposing a then-record fine of 2.42 billion euro, but it also ordered Google to put an end to its self-preferencing conduct and ensure equal access to all third-party providers on its general search website.

In holding that Google's practice of self-preferencing amounted to a standalone abuse of dominance, the Commission pushed the boundaries of art. 102 TFEU outward. The Google Shopping decision sent out the bold message that art. 102 TFEU could be used to interfere in the way in which dominant firms design their products and services and impose a far-reaching obligation of equal treatment on these design choices. There was little to no precedent on which the Commission could rely in support of this conclusion.[2] Most notably, the Commission argued that it was not required to demonstrate that Google's self-preferencing amounted to a refusal to deal as defined in the Bronner case.[3] Instead, the Commission took the view that it was not bound to show that Google's general search result page constituted, consistent with the demanding Bronner test for refusal to deal, an essential facility or indispensable input for competing CSS to be able to operate effectively.[4] At the same time, however, the Commission failed to articulate any alternative legal test in support of its finding that self-preferencing by a dominant firm may breach art. 102 TFEU. The Google Shopping decision soon fuelled a controversial debate about the appropriate legal test and limiting principles that would indicate when self-preferencing constituted an abuse of dominance and delineate the prohibitive scope of art. 102 TFEU.[5]

Since the Commission's decision in Google Shopping, the novel theory of harm of self-preferencing took on a life of its own. Self-preferencing is singled out by a number of expert reports as emblematic example of the new types of anticompetitive conduct that dominant platforms, most notably Google, Apple, Amazon and Facebook, have used to leverage their market power and entrench their grip over digital markets.[6] Self-preferencing also lies at the core of a number of high-profile antitrust investigations that the Commission and other competition authorities have recently opened against Amazon[7] and Facebook.[8] Moreover, the Commission's Google Shopping decision importantly informs recent initiatives for new platform regulations in Germany, the EU, and the UK. Indeed, self-preferencing figures prominently as one of the blacklisted practices that are (set to be) outlawed by the new rules for powerful digital platforms in all three jurisdictions.[9]

In confirming the Commission's finding that self-preferencing by a dominant firm constitutes a standalone abuse of dominance under art. 102 TFEU, the General Court's Google Shopping judgment provides a major clarification. Most notably, it endorses the Commission's move to expand the prohibitive scope of that article. It also emboldens recent initiatives across Europe to impose ex ante rules on powerful digital platforms that establish a per se prohibition of self-preferencing. The judgment also makes a number of important points on the economic and legal analysis under art. 102 TFEU, notably with respect to the relevant effects-based analysis,[10] the as-efficient competitor test,[11] and the requisite standard of proof.[12] But does it also address the fundamental question of how far art. 102 TFEU can interfere with the design choices of dominant firms and prohibit them from according favourable treatment to their own products or services?

The short answer to this question is: no. Those who expected the General Court to delineate the redrawn boundaries of art. 102 TFEU after Google Shopping or had hoped for a limiting principle that determines how far a dominant firm’s obligation of equal treatment reaches have certainly been disappointed by the judgment. The General Court missed the opportunity in Google Shopping to set out a clear legal test to establish which elements must be fulfilled for self-preferencing to qualify as an abuse of dominance. Even after the General Court's Google Shopping ruling, we remain in the dark as to when exactly self-preferencing amounts to an abuse of dominance and when it does not.

The remainder of this Insight develops this argument as follows. Section II discusses the General Court's holding that self-preferencing constitutes a free-standing abuse of dominance under art. 102 TFEU as the main takeaway of the case. Section III critically reflects on different unsuccessful attempts by the General Court to pin down a legal test that spells out the constitutive elements that self-preferencing has to fulfil to qualify as an abuse of dominance. Section IV proposes an alternative route that the General Court could have taken to devise a clear analytical framework and test for the analysis of self-preferencing in future cases. Section V concludes.

II. Self-preferencing as a free-standing abuse of dominance

The most important takeaway and clarification brought about by the Google Shopping judgment is that self-preferencing, or what the General Court calls “favouring”, on the part of a dominant firm is liable to constitute on its own an independent form of abuse of dominance. The General Court thus sided with the Commission's position that, as a matter of principle, unilateral conduct whereby a vertically integrated dominant platform gives greater visibility to its own product/service relative to competing products/services and downgrades the “findability” of competing products/services may amount to a standalone breach of art. 102 TFEU.[13]

The Google Shopping ruling also upheld the most important elements of the Commission's reasoning that supported its finding that Google's self-preferencing violated art. 102 TFEU. The General Court recalled that art. 102 TFEU provides a non-exhaustive list of examples of abusive practices. The mere facts that self-preferencing constitutes a novel form of abuse and that art. 102 TFEU does not explicitly refer to self-preferencing cannot preclude the Commission from characterising self-preferencing as a standalone abuse of art. 102 TFEU.[14] Moreover, the General Court rejected the claim advanced by Google and learned commentators[15] that the Commission had to establish on the basis of the Bronner criteria that self-preferencing is tantamount to a refusal to give access to an indispensable input for it to be prohibited under Art. 102 TFEU.[16] The Google Shopping ruling thus makes it plain that self-preferencing by a dominant firm may fall foul of art. 102 even if access to the platform (or infrastructure) on which it occurs is not indispensable for rivals active on an adjacent upstream or downstream market and the self-preferencing does not eliminate all competition on that adjacent market.[17]

In upholding the Commission's finding that self-preferencing may constitute a standalone abuse under art. 102 TFEU, the General Court endorsed the Commission's expansive reading of the special responsibility[18] of dominant firms under art. 102. This bold reading implies that art. 102 TFEU may, in certain circumstances, impose limits on the dominant firm's discretion to alter or even improve the design of its products and services.[19] Google Shopping also reaffirmed the open-textured and versatile nature of the prohibition of art. 102 TFEU. For dominant firm conduct to be in breach of art. 102 TFEU, it does not have to fall within well-established legal categories or economic theories of harm. Instead, art. 102 TFEU empowers the Commission to take action against any form of unilateral conduct by dominant firms that is inconsistent with the principle of competition on the merits because it excludes competing operators from the market, not on the basis of better quality or performance but by reason of the dominant firm's superior market power.[20]

This reaffirmation of the “openness”[21] of art. 102 TFEU in Google Shopping is a welcome development. It serves as an important reminder of the “conceptual elasticity”[22] of art. 102 TFEU, which enables the EU competition law system to respond to novel types of abuses and harms to competition. This flexibility and responsiveness undoubtedly constitute a strength of the system, as they allow EU competition law to adapt to changing circumstances and secure the preservation of competition even in fast-moving environments, such as digital markets.

III. A legal test unfound

This openness of art. 102 TFEU is, however, not without limits. Openness only strengthens the system as long as it does not come at the expense of the predictability and clarity of competition law and thereby undermine its integrity.[23] The interpretation of art. 102 TFEU thus calls for limiting principles that strike a balance between the openness and the integrity of the provision by ensuring legal certainty as to when specific conduct accords with lawful competition on the merits, and when it does not.

Yet the General Court's ruling is much less cogent when it comes to laying down limiting principles that provide guidance as to when self-preferencing clashes with competition on the merits. In its infringement decision, the Commission relied on two somewhat distinct theories of harm to support its finding that Google's conduct amounts to an abuse of dominance. On the one hand, it advanced a discrimination theory of harm. It found that Google had subjected competing CSS to unequal, and hence unfair, treatment by giving its own CSS greater visibility on its general search result pages while downgrading the visibility of competing CSS, thereby depriving them of a substantial amount of end-user traffic.[24] On the other hand, the Commission also argued that Google's self-preferencing was abusive because it allowed the company to leverage its dominant position on the market for general search into the adjacent market for comparison shopping services.[25] The Commission, however, failed to put forth a clear legal test for either of these theories of harm.

III.1. The leveraging theory of harm

The General Court's ruling in Google Shopping does not fare any better. With respect to the “leveraging” theory of harm, the General Court observed that leveraging is not a specific form of abuse in itself. Rather, it constitutes a generic term that describes practices that occur in one market and have appreciable results on competition in related adjacent markets.[26] Leveraging is, therefore, best understood as an umbrella term to designate various practices such as tying, refusal to deal, margin squeeze, or rebates.[27] While leveraging by a dominant firm may amount to an abuse of dominance if it takes the form of one of these practices, the Court asserted that the mere fact that a dominant undertaking leverages its market power from one market to another is insufficient to qualify, in itself, as an abuse. Accordingly, something more than mere leveraging must be established for there to be an abuse of dominance in breach of art. 102 TFEU.[28]

How much more, however, remains unclear. Google Shopping clearly resists a restrictive reading that would only outlaw leveraging that falls within the established category of a refusal to deal as defined in Bronner.[29] To make this point, the General Court walked a thin line. The Court went to great lengths to distinguish self-preferencing from refusals to deal covered by Bronner. It held that the Bronner test only applies to express[30] refusals to deal, while self-preferencing, as well as other forms of constructive refusals to deal, such as margin squeeze, are not governed by the strict Bronner criteria.[31] At the same time, however, the General Court undermined its efforts to narrow the scope of Bronner to “express” refusals to deal by asserting that it emerges from the Commission's decision that Google's general search result page had “characteristics akin to those of an essential facility”.[32] According to the General Court, the Commission had – at least implicitly or unwittingly – established that the traffic generated by Google's general search result pages was “indispensable”[33] for rival CSS to operate because it could not be realistically replicated by other sources[34] and that Google's practice led “to the potential elimination of all competition” on the CSS market.[35]

The General Court’s discussion of Bronner appears to have it both ways: the Court contends that the Commission was under no obligation to demonstrate that self-preferencing amounted to a refusal to deal as defined in Bronner, only to tell us in the same section of the judgment that the Commission had, in any case, implicitly established that Google's conduct met the Bronner criteria. A charitable reading may interpret this fundamental tension in the General Court's reasoning as a sign that the judges were deeply divided over the requisite legal standard to be met in order for self-preferencing to qualify as an abuse. A less forgiving reading would suggest that it defies logic. The Court’s orbiter dictum on the quasi-essentiality of Google’s search result page also raises broader questions about the scope of the General Court's judicial review because it de facto substitutes, at least in part, its own factual assessment for that of the Commission.[36]

While the Court made it clear that, as a matter of principle, the Commission was under no obligation to demonstrate that self-preferencing fulfilled the Bronner criteria, it omitted to clearly spell out any alternative test or limiting principle to determine when self-preferencing ought to be condemned as unlawful leveraging. Such a test is, if at all, only implicitly articulated in the judgment. The Court highlighted that the Commission did not content itself with establishing that self-preferencing enabled Google to leverage its market power, but that it also adduced additional circumstances that indicated how this leveraging entailed anticompetitive effects.[37] According to the Court, the Commission had relied on five relevant elements to characterise self-preferencing as an abuse. Google's conduct i) gave a more prominent display to its own CSS, while ii) downgrading the visibility and ranking of competing CSS on its general search results page.[38] This affected iii) consumers’ behaviour and iv) led to a reduction in traffic coming from Google's general search results page to rival CSS, and v) Google’s search results could not be replicated in any effective or economically viable manner through other sources.[39] Though these five elements may constitute the rudimentary scaffolding for a legal test to distinguish lawful from unlawful self-preferencing, the General Court by no means signalled that these elements were binding on the Commission when establishing unlawful self-preferencing in future cases.

III.2. Unequal treatment and the “no economic sense test”

With regard to the discrimination theory of harm, the Google Shopping ruling does not provide much clarity either. Instead, the General Court again in broad brushes followed the reasoning of the Commission. It held that Google's self-preferencing was objectionable because it treated rival CSS less favourably than its own CSS and thereby placed its competitors at an unfair disadvantage.[40] The General Court highlighted that the principle of equal treatment not only constitutes a general principle of EU law[41] but also forms the very basis of equality of opportunity between competitors which is a prerequisite of undistorted competition.[42]

The General Court then introduced some form of a novel “no economic sense test” [43] to describe when and why self-preferencing constitutes an undue form of unequal treatment or discrimination. It contended that Google's initial business model for its general search engine consisted of providing consumers with neutral, objective, and unbiased search results which are most relevant to their search queries.[44] The Court took the view that Google's self-preferencing of its own CSS relative to competing CSS sites regardless of their relevance constituted an “abnormality”[45] as it runs counter to Google’s own business model.[46] Not only, so the Court, was the self-preferencing only possible by virtue of Google's market power, as consumers would otherwise have switched to other more objective alternatives; it also implied that the self-preferencing could not be explained by any economic rationale other than Google's goal of excluding competing CSS providers.[47]

This iteration of the “no economic sense test” is fraught with difficulties. Above all, the General Court's reasoning implies that a dominant undertaking's special responsibility may prevent it from altering its business model if such a change turns out to be disadvantageous for competing products/services relative to the dominant firm’s own product/service.[48] By implication, this imposes an obligation on Google to run its general search service like a public utility and to guarantee equal access for all interested third-party providers. Indeed, Google Shopping goes as far as suggesting that even changes in the product design which contribute to an improvement of the dominant firm's business model may be incompatible with the special responsibility of dominant firms not to impair competition any further.[49]

Such a sweeping interpretation of the special responsibility of dominant firms is not a problem in itself, as long as its underpinning rationale and limiting principles are clearly articulated and its implications for innovation duly considered. However, no such explanation is provided apart from the fact that Google held a “superdominant position” and operated as “a gateway to the internet” within a market characterised by high entry barriers.[50] The implications of the General Court's “no economic sense test” also deviate from the recent CK Telecoms ruling. In this judgment, the General Court held that the mere fact that a merged entity adopts changes to its post-merger business strategy cannot result in a significant impediment to effective competition even though it may considerably harm competitors.[51]

The most important shortcoming of the General Court's version of the “no economic sense test” is that it leaves open the matter of exactly when unequal treatment in the form of self-preferencing amounts to an “abnormality”. It thus fails to put forth any limiting principle that delineates the scope of the special responsibility of dominant firms under art. 102 TFEU. Nor does it give guidance to dominant firms with regard to how far they can go in designing their products in a way that grants preferential treatment to their own business units. Indeed, the “no economic sense test” tells us nothing about the constitutive conduct elements of unlawful self-preferencing, let alone the exclusionary effects that must be present for it to breach art. 102 TFEU. The General Court put major emphasis on the fact that Google's self-preferencing consisted of two elements, namely the upgrading of its own CSS and the demoting of competing CSS.[52] In no instance, however, did it state that self-preferencing must contain both conduct elements – that is, involve the upgrading of the dominant firm's own service plus the active and consistent downgrading of its competing services. It remains hence unclear whether, in future cases, only one of these two elements would be sufficient to establish the abusive character of self-preferencing conduct.

The General Court's failure to clarify this point is a major omission. The question of whether the two elements of upgrading of the dominant firm's own business and the downgrading of competitors have to be cumulatively present for self-preferencing to qualify as abuse is determinative of the evidentiary burden for competition authorities and private plaintiffs under art. 102 TFEU. It is also of paramount importance for dominant undertakings to know how far they can go in designing their products and services in a way that gives greater visibility to their own offers. Moreover, it affects the evidentiary burden for the defendant undertaking when pleading an objective justification or putting forward a counterfactual as rebuttal evidence. That becomes evident in the General Court's discussion of Google's objective justification and counterfactual analysis. The Court observed that Google could not justify its otherwise unlawful self-preferencing on the grounds that it gave its own CSS greater visibility to improve its service quality. While this explanation may serve as justification for the upgrading of Google's own CSS, the Court held that it failed to provide a legitimate explanation for the downgrading of competing CSS.[53] Along similar lines, the General Court also rejected the relevance of two counterfactual studies Google produced during the proceedings. These studies showed by means of a “difference-in-difference analysis” and an “ablation experiment” that the greater visibility of Google’s own CSS on its general search result page had only a marginal impact on traffic.[54] The General Court objected that both studies focused exclusively on the upgrading element of Google's self-preferencing conduct in isolation, without considering the downgrading of rival CSS. As a result, the counterfactual discounted the fact that Google's self-preferencing was a combination of two conduct elements that produced combined anticompetitive effects through two channels.[55]

One is left wondering whether this means that unlawful self-preferencing always consists of upgrading and downgrading, for both of which dominant firms have to advance an objective justification or other forms of rebuttal evidence. Or does it imply that even self-preferencing consisting of only one out of the two conduct elements criticised in Google Shopping may run afoul of art. 102 TFEU, but that, in this case, dominant firms face a lower evidentiary burden because they only have to adduce an objective justification or counterfactual for this one element?

Greater clarity on the constitutive elements of unlawful self-preferencing under art. 102 TFEU may also inform the current policy discussion on the new platform regulations. Above all, it may provide guidance on the scope of the ex-ante prohibitions of self-preferencing enshrined in the Digital Markets Act (DMA), which draws up a new regulatory framework for powerful “gatekeeper” platforms.[56] In its current state, art. 6(1)(d) of the DMA proposal is framed in a way that requires gatekeeper platforms to refrain from any favourable ranking of their own products or services, irrespective of whether competing products are actively downgraded. If Google Shopping is to be read as requiring upgrading and downgrading as two cumulative constitutive elements of self-preferencing, it would be up to the EU legislator to take the necessary steps to finetune art. 6(1)(d) lest it become a major source of inconsistency.

IV. The route not taken

If the General Court in Google Shopping was in search of the legal test or limiting principle for self-preferencing, its quest might have taken a wrong turn somewhere between Place Madou and Kirchberg. The arguably most surprising feature of the Google Shopping ruling is that the General Court discusses Google's self-preferencing against the backdrop of long-standing case law[57] interpreting the prohibition of discriminatory behaviour on the part of dominant firms under art. 102(2)(c) TFEU with no mention of the recent MEO ruling. The MEO judgment added to this case law by further clarifying when a dominant firm's unequal treatment of upstream or downstream customers violates art. 102(2)(c) TFEU. In a nutshell, that is the case if a dominant firm i) applies dissimilar conditions to equivalent transactions and thereby ii) places other trading parties at a competitive disadvantage.[58]

The major contribution of MEO was that it clarified that the Commission could no longer content itself with establishing that the dominant firm had subjected trading partners to unequal treatment. Instead, it would also have to assess “all the relevant circumstances” to determine whether the discriminatory conduct had a foreclosure effect on upstream or downstream customers, and it would have to ascertain the seriousness of this effect.[59] Amongst the relevant circumstances for this effects-based analysis of the dominant firm's unequal treatment are, for instance, the negotiating power of upstream or downstream customers, the conditions and duration of the discriminatory arrangements, the existence of an exclusionary strategy, and, notably, the impact of the discriminatory practice on the relevant customers' cost.[60]

IV.1. Applying the MEO test to self-preferencing

While the MEO case involved discriminatory pricing conduct, nothing precludes the application of this test to determine the legality of self-preferencing. Indeed, the wording of art. 102(2)(c) TFEU, which prohibits the application of “dissimilar conditions to equivalent transactions”, is sufficiently flexible to accommodate both discriminatory pricing and non-price conduct.[61] Nor does the fact that MEO was not vertically integrated and only discriminated amongst third-party downstream customers suggest that it constitutes an inapposite framework for the analysis of self-preferencing that relates to discrimination by a vertically integrated firm between its own upstream/downstream division and competing third-party services. On the contrary, the Advocate General[62] and the Court[63] agreed in MEO that the fact that a dominant firm is vertically integrated and uses discriminatory conduct to favour its own upstream or downstream unit would make it a more plausible foreclosure strategy. This observation is of particular relevance for the analysis of unequal treatment in digital markets involving platforms with a hybrid business model which, like Google, operate both a marketplace and sales function.[64] As their upstream or downstream sales function is exposed to a cannibalisation effect by competition at the relevant level of trade, vertically integrated hybrid platforms may have strong incentives to design their marketplace in a way that places upstream or downstream rivals at a disadvantage.

The MEO test would certainly provide a conceptually sounder foundation for the analysis of self-preferencing than the General Court's “no economic sense test”. Elegant though it may appear, the “no economic sense test” often puts competition authorities and courts in the uncomfortable position of second-guessing the business model of the dominant firm. Such an exercise inevitably involves strong assumptions which can be easily challenged and which, being subjective in nature, sit uneasily with the fact that the notion of “abuse of dominance” is an objective concept.[65] This becomes apparent from the General Court's bold claim that Google's self-preferencing runs counter to its business model and therefore does not make any economic sense. This proposition appears out of sync with commercial reality. Strategic product placement of private labels or products by “category captains” are commonplace in the grocery sector[66] and has been recognised to be a source of important efficiencies.[67] This is the case even though the business model of grocery stores also hinges on providing consumers with access to a broad choice of products.[68] The difference between the self-preferencing employed by Google and that employed by grocery stores or their category captains does not necessarily lie in their respective business models. It revolves instead around their respective market power and the ensuing potential to generate substantial foreclosure effects.[69] It is on these effects, rather than the economic sense of Google's conduct, where the focus of the art. 102 TFEU analysis should rest.

IV.2. The merits of the MEO test

An assessment of self-preferencing under art. 102(2)(c) TFEU based on the MEO test would have the merit of putting this theory of harm back on a conceptually sound foundation and of providing a clear framework that ensures legal certainty to dominant firms and their competitors as to when self-preferencing amounts to unlawful abuse. Considering self-preferencing within the framework of art. 102(2)(c) TFEU, as interpreted in MEO, would have enabled the General Court to “retrofit” the Commission's analysis into well-established conceptual and legal categories.

Applying the MEO test to self-preferencing would also do greater justice to the economic analysis that the Commission deployed in Google Shopping. The beauty of MEO is that it realigns the assessment of secondary-line discrimination with the “raising rivals' costs” paradigm that provides an economically sound and unified framework for the analysis of arguably any type of exclusionary conduct under art. 102 TFEU. MEO suggests that unequal treatment of upstream or downstream competitors by a dominant firm is harmful to competition if it raises their costs to the extent that they are no longer capable of exerting meaningful competitive pressure and, as a result, competition is materially impaired.[70] Such a strategy of raising rivals' costs can be implemented through the pricing of an important input controlled by the dominant firm that competitors (upstream or downstream) need in order to be able to compete effectively. But it can also take the form of non-price foreclosure of important distribution channels that prevents competitors from achieving minimum efficient scale or forces them to distribute through less effective and more costly sales channels.[71]

Arguably, the Commission's analysis endorsed by the General Court in Google Shopping closely followed the blueprint of the raising rivals' cost analysis that informs MEO as well as other leading “effects” cases under art. 101[72] and art. 102 TFEU.[73] The Commission first assessed the foreclosure rate of Google's exclusionary practice by gauging the proportion of CSS traffic that was affected by Google's self-preferencing.[74] It concluded that Google's self-preferencing affected a “large proportion of the overall traffic of competing comparison shopping services”.[75] Second, it found that Google's self-preferencing resulted in a material reduction of the traffic that rival CSS received from Google's general search result page.[76] Third, it considered the extent to which competing CSS could have recourse to effective counterstrategies that would allow them to compensate for the loss in traffic from Google's general search result page by reaching end-users through alternative channels. It found that competing CSS could not offset the loss in traffic from Google's general search results through alternative distribution channels, such as text ads on Google, mobile apps, direct traffic, or other sources, which proved either ineffective or economically unviable substitutes for the traffic coming from Google's general search result page.[77] The traffic from Google's search result notably outperformed other traffic in generating clicks for CSS. Moreover, some of the alternative sources of traffic, such as text ads, were also considerably more expensive than search engine optimisation that ensures traffic from Google's general search result page.[78] In a nutshell, the Commission attempted to show that Google's unequal treatment of competing CSS raised rivals' costs by foreclosing a substantial fraction of their distribution channel and forcing them to reach customers through more expensive and less effective distribution channels, thereby preventing them from building up a critical mass of network effects.[79]

Grounding the assessment of self-preferencing in the MEO test and the underlying raising rivals' cost theory of harm would also have allowed the General Court to put forward a more cogent explanation as to why the Commission was under no obligation to demonstrate that non-price abuses, such as self-preferencing, are capable of foreclosing equally efficient competitors.[80] Indeed, the General Court seems to erroneously assume that the as-efficient competitor test, which seeks to determine whether the impugned dominant firm conduct is capable of foreclosing equally efficient competitors, can only take the form of an (incremental) price-cost test and can therefore only be used for the assessment of pricing, but not non-price abuses.[81] The raising rivals’ cost literature suggests otherwise. One way of determining whether non-price conduct on the part of a dominant firm forecloses an equally efficient competitor from the market is to look at the foreclosure rate of the impugned conduct and determine whether it is sufficiently large to hinder competitors from achieving minimum efficient or viable scale (MES/MVS test).[82] If that is the case, it is safe to assume that even equally efficient competitors would not be able to withstand the dominant firm's conduct and that the impugned conduct is likely to lead to market power effects. Yet, even in its iteration as an MES/MVS test, the as-efficient competitor test may prove under-inclusive because it only catches conduct that is capable of fully foreclosing an equally efficient competitor from the market. The as-efficient competitor test – both in its operationalisation as (incremental) price-cost or MES/MVS test – thus fails to address anticompetitive conduct that falls short of full foreclosure but nonetheless significantly raises rivals' costs and reduces their competitive impact, without however entirely eliminating them from the market.[83]

Above all, the MEO test would have enabled the General Court to lay down a principled and effects-based legal test that clearly indicates when self-preferencing by a dominant firm will qualify as an abuse in future cases. Simultaneously, it sets out a limiting principle as to how far a dominant firm's equal treatment obligation reaches. Consistent with existing case law, self-preferencing would be unlawful if the dominant firm i) applies dissimilar conditions to equal transactions and ii) thereby places other trading parties at a competitive disadvantage by foreclosing a large enough fraction of their distribution channel to raise their costs.[84] Such would notably be the case if the self-preferencing conduct covers an important distribution channel and rivals do not have access to effective and economically viable alternative sales channels. This test would introduce an administrable and economically informed limiting principle to distinguish legitimate and illegitimate self-preferencing or unequal treatment on the part of a dominant firm: a dominant firm would thus remain free to design its business in a way that favours its own products or services over those sold by rivals, as long as the unequal treatment or self-preferencing does not have the potential to result in significant foreclosure effects.

V. Conclusion

(Legal and economic) Tests have been close to the heart of competition lawyers, and not just since the outbreak of the Covid-19 pandemic. The Commission's omission in its Google Shopping decision to articulate clear benchmarks that specify when self-preferencing violates art. 102 TFEU has again brought the question of the appropriate legal test for finding an abuse of dominance to the forefront of European competition law debates. While the General Court's ruling brings about some important clarifications as to the legal status of self-preferencing under art. 102 TFEU, it failed to remedy this basic flaw in the Commission's decision by forging clear limiting principles that delineate when self-preferencing infringes art. 102 TFEU. Neither the General Court's pronouncements on the scope of the Bronner test, nor its discussion of leveraging and the “no economic sense test”, provide clear and precise guidance as to when self-preferencing is incompatible with competition on the merits, and when it is not. Even after Google Shopping, dominant firms and their competitors are left wondering which constitutive elements have to be fulfilled for self-preferencing to breach art. 102 TFEU. The failure of the General Court to establish a clear limiting principle determining the (il)legality of self-preferencing is unfortunate. Dominant firms will have little legal certainty as to how far they can go in altering their business model and design of their products or services without risking antitrust liability. This may chill their incentives to undertake product improvements and to innovate. By the same token, competitors and competition authorities also have little certainty as to how far the protective scope of art. 102 TFEU reaches and when they can effectively challenge self-preferencing by dominant firms. This uncertainty may also take its toll on competitors' incentives to innovate.

This unsatisfactory outcome could easily have been avoided. The MEO test, which clarifies long-standing case law on when discriminatory conduct by dominant firms constitutes an abuse of dominance, could have provided a well-established legal framework that sets out the relevant constitutive elements for unlawful self-preferencing. It would also have allowed the General Court to link the analysis of self-preferencing with the “raising rivals’ costs” paradigm, which offers a unified economic framework for assessing exclusionary conduct under art. 102 TFEU. This road has not been taken by General Court. As a result, competition lawyers, dominant firms and their competitors will continue to be in search of the legal test for self-preferencing. A counter-intuitive takeaway of the Google Shopping saga is that obsessive attempts to shoehorn novel forms of exclusionary conduct into existing legal categories enhance neither analytical clarity nor legal certainty. The alternative route of grounding the assessment of exclusionary conduct in the "raising rivals‘ cost" paradigm would provide an economically well-established and sound, principled and unified framework for an effects-based analysis of any type of exclusionary conduct under Art. 102 TFEU, which ensures legal certainty and strikes an adequate balance between the openness and integrity of the system.

Some may object that the legal question of the applicable test for self-preferencing will soon be moot because the new DMA will, in any case, introduce a per se prohibition for self-preferencing by gatekeeper platforms in the near future. The reality may, however, turn out to be slightly more complex. The difficulty of pinning down a clear definition for self-preferencing may soon also creep into the enforcement of the DMA. The vague definition of self-preferencing in art. 6(1)(d) of the current DMA proposal does not bode well in this regard.[85] At the same time, it is also conceivable that the failure of the General Court to establish a clear test for self-preferencing under art. 102 may facilitate cross-fertilisation between the DMA and the assessment of art. 102 TFEU. For better or worse, the per se approach against self-preferencing under the new platform regulations may thus also creep into the enforcement of art. 102 TFEU against self-preferencing by dominant firms that are not designated as gatekeeper platforms. Google Shopping is thus also a missed opportunity for the General Court to contribute to greater consistency between art. 102 TFEU and the nascent DMA and, thereby, to prevent undue fragmentation of EU competition law.

European Papers, Vol. 6, 2021, No 3, European Forum, Insight of 22 February 2022, pp. 1345-1361
ISSN 2499-8249 - doi: 10.15166/2499-8249/528

* Lecturer in Competition Law and IP, University of East Anglia Law School,

[1] Case T-612/17 Google and Alphabet v Commission (Google Shopping) ECLI:EU:T:2021:763.

[2] Indeed, the Commission struggled to ground its finding that self-preferencing breaches art. 102 TFEU in a solid line of precedent. See European Commission Decision of 27 June 2017 case COMP/AT.39740 Google Search (Shopping) paras 334-336.

[3] Case C-7/97 Bronner v Mediaprint ECLI:EU:C:1998:569 paras 41, 44-46.

[4] Google Search (Shopping) cit. para. 651.

[5] See e.g. J Temple Lang, ‘Comparing Microsoft and Google: The Concept of Exclusionary Abuse’ (2016) World Competition 5; P Ibáñez Colomo, ‘Indispensability and Abuse of Dominance: From Commercial Solvents to Slovak Telekom and Google Shopping’ (2019) Journal of European Competition Law & Practice 532.

[6] J Furman, D Coyle, A Fletcher, D McAuley and P Marsden, Unlocking digital competition: Report of the Digital Competition Expert Panel (2019) para. 2(36); J Crémer, YA de Montjoye and H Schweitzer, Competition Policy for the Digital Era (2019) 65–67.

[7] European Commission, Press release IP/20/2077 Antitrust: Commission sends Statement of Objections to Amazon for the use of non-public independent seller data and opens second investigation into its e-commerce business practices (2020); Autorità garante della Concorrenza e del Mercato, Press release A528 Antitrust Amazon fined over € 1,128 billion for abusing its dominant position (2021) See also, Autorità garante della Concorrenza e del Mercato, Provvedimento A528 – Sanzione di oltre 1 miliardo e 128 milioni di euro ad Amazon per abuso di posizione dominante (9 December 2021) para. 716

[8] Competition and Markets Authority (CMA), Press release CMA investigates Facebook’s use of ad data (2021)

[9] Deutscher Bundestag, ‘Gesetz zur Änderung des Gesetzes gegen Wettbewerbsbeschränkungen für ein fokussiertes, proaktives und digitales Wettbewerbsrecht 4.0 und andere Bestimmungen (GWB-Digitalisierungsgesetz)’ (18 January 2021) Bundesgesetzblatt Jahrgang art. 19(a)(2)(1); European Commission, Proposal for a Regulation of the European Parliament and of the Council COM(2020) 842 final of 15 December 2020 on contestable and fair markets in the digital sector (Digital Markets Act) art. 6(1)(d); Department for Digital, Culture, Media & Sport and Department for Business, Energy & Industrial Strategy, ‘A New Pro-competition Regime for Digital Markets’ (20 July 2021) figure 4.

[10] Google and Alphabet v Commission cit. paras 435-45, 169-175 and 519-533.

[11] Ibid. paras 538-541.

[12] Ibid. paras 377-379 and 438-443.

[13] Ibid. paras 150-197.

[14] Ibid. para.154. This conclusion follows long-established precedent. See case C-6/72 Europemballage Corporation and Continental Can Company v Commission ECLI:EU:C:1973:22 para. 26; case C-52/09 Telia-Sonera Sverige ECLI:EU:C:2011:83 para. 26.

[15] See e.g. J Temple Lang, ‘Comparing Microsoft and Google’ cit.

[16] Google and Alphabet v Commission cit. para. 240.

[17] Interestingly, the General Court consistently refers to the elimination of “all competition” on the neighbouring market as one of the constitutive elements of a refusal to deal. See ibid. paras 213–217 and 228. It thereby departs from the lower standard endorsed in Microsoft and the Commission’s Guidance paper, according to which the refusal need only eliminate “any effective competition” on the neighbouring market to qualify as an abuse. See case T-201/04 Microsoft Corp v Commission ECLI:EU:T:2007:289 para. 332; see Communication C 45/7 from the Commission - Guidance on the Commission's enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings, para. 81.

[18] Case C-322/81 Michelin v Commission ECLI:EU:C:1983:313 para. 57.

[19] Google and Alphabet v Commission cit. paras 150, 188 and 612.

[20] Ibid. para. 152. Case C-457/10 P AstraZeneca AB and AstraZeneca plc v European Commission ECLI:EU:C:2012:770 para. 75; case C-413/14 P Intel v Commission ECLI:EU:C:2017:632 paras 133–134 and 136.

[21] This point draws on the insightful discussion in S Makris, ‘Openness and Integrity in Antitrust’ (2020) Journal of Competition Law & Economics 1.

[22] Ibid. 3.

[23] See in particular ibid. 52-60.

[24] Google Search (Shopping) cit. para. 336.

[25] Ibid. para. 334.

[26] Google and Alphabet v Commission cit. para. 163.

[27] Ibid.

[28] Ibid. para. 164.

[29] Bronner v Mediaprint cit. paras 41-47.

[30] Google and Alphabet v Commission cit. paras 231-233.

[31] Ibid. paras 229-247.

[32] Ibid. para. 224.

[33] Ibid. para. 227.

[34] Ibid. paras 226-227.

[35] Ibid. para. 228.

[36] In so doing, the General Court oversteps the limits of its judicial review recognised in ibid. para. 135.

[37] Ibid. para. 175.

[38] Ibid. paras 167-168.

[39] Ibid. paras 169-174.

[40] Ibid. paras 155 and 179-180.

[41] Ibid. para. 155.

[42] Ibid. para. 180.

[43] GJ Werden, ‘The “No Economic Sense” Test for Exclusionary Conduct’ (2006) The Journal for Corporation Law 293.

[44] Google and Alphabet v Commission cit. paras 177-178.

[45] Ibid. paras 177 and 179.

[46] Ibid. para. 179.

[47] Ibid. para. 178.

[48] Ibid. cit. para. 183.

[49] Ibid. paras 150, 180 and 612.

[50] Ibid. para. 183.

[51] Case T-399/16 CK Telecoms UK Investments v Commission ECLI:EU:T:2020:217 paras 340-348 and 362-368.

[52] Google and Alphabet v Commission cit. paras 167-168, 187, 261 and 369-376.

[53] Ibid. para. 187.

[54] Google Search (Shopping) cit. paras 506-538.

[55] Google and Alphabet v Commission cit. paras 369-376.

[56] Proposal for a Regulation COM(2020) 842 final cit.

[57] The General Court notably cites case C-242/95 GT-Link v De Danske Statsbaner ECLI:EU:C:1997:376; case T-228/97 Irish Sugar v Commission ECLI:EU:T:1999:246; case C-82/01P Aéroports de Paris v Commission ECLI:EU:C:2002:617. See Google and Alphabet v Commission cit. para. 155.

[58] Case C-525/16 MEO Serviços de Comunicações e Multimédia ECLI:EU:C:2018:270 paras 24-25.

[59] Ibid. paras 25-28.

[60] Ibid. paras 29-37.

[61] This reading also finds support in the General Court’s analysis, which seems to assume that the display and ranking of Google’s own CSS and of relevant competing CSS on Google’s general search result page constituted equivalent transactions. Google and Alphabet v Commission cit. paras 278-299.

[62] Case C-525/16 MEO Serviços de Comunicações e Multimédia SA ECLI:EU:C:2017:1020, opinion of AG Wahl, paras 47 and 76-80.

[63] MEO Serviços de Comunicações e Multimédia cit. para. 36.

[64] SP Anderson and Ö Bedre-Defolie, ‘Hybrid Platform Model’ (CEPR Discussion Paper 2021).

[65] Case C-85/76 Hoffmann-La Roche v Commission ECLI:EU:C:1979:36 para. 91.

[66] RL Steiner, ‘Category Management – A Pervasive, New Vertical/Horizontal Format’ (2000-2001) Antitrust 77; PW Dobson and R Chakraborty, ‘Assessing Brand and Private Label Competition’ (2015) European Competition Law Review 76.

[67] In relation to category management agreements, see European Commission, ‘Guidelines on Vertical Restraints’ (19 May 2010) para. 213.

[68] Ibid. para. 210.

[69] This is clearly recognised by the Commission’s Vertical Guidelines, which consider category management agreements between parties with a market share of less than 30 per cent unproblematic, but which at the same time are alert to potential foreclosure effects if the parties possess considerable market power. See ibid. paras 209-210.

[70] Meo Serviços de Comunicações e Multimédia cit. paras 34 and 37.

[71] TG Krattenmaker and SC Salop, ‘Anticompetitive Exclusion: Raising Rivals' Costs to Achieve Power over Price’ (1986) YaleLJ 209 and 226.

[72] Case C-234/89 Delimitis v Henninger Bräu ECLI:EU:C:1991:91 paras 13-15 and 19-27; case C-345/14 Maxima Latvija ECLI:EU:C:2015:784 paras 27-31.

[73] Case T-65/98 Van den Bergh Foods Ltd v Commission ECLI:EU:T:2003:281 para. 160.

[74] Google and Alphabet v Commission cit. paras 169, 173, 446-448 and 519-522.

[75] Ibid. paras 448 and 520.

[76] Ibid. paras 448, 520 and 523.

[77] Ibid. paras 173-174 and 523-532.

[78] Ibid. para. 529.

[79] Ibid. paras 171 and 226. For the importance of network effects in dynamic “raising rivals‘ costs” theories of harm, see DW Carlton, ‘A General Analysis of Exclusionary Conduct and Refusal to Deal: Why Aspen and Kodak Are Misguided’ (2001) Antitrust Law Journal 659 and 671.

[80] Google and Alphabet v Commission cit. para. 539.

[81] Ibid.

[82] TG Krattenmaker and SC Salop, ‘Anticompetitive Exclusion’ cit. 234-238; DA Crane and G Miralles, ‘Toward a Unified Theory of Exclusionary Vertical Restraints’ (2010) SCalLRev 605, 607-609 and 639-641; JD Wright, ‘Moving Beyond Naïve Foreclosure Analysis’ (2012) George Mason Law Review 1163 and 1186–1187.

[83] SC Salop, ‘The Raising Rivals’ Cost Foreclosure Paradigm, Conditional Pricing Practices and the Flawed Incremental Price-Cost Test’ (2017) Antitrust Law Journal 371 and 385–387. This has been recognised by the General Court in relation to the incremental price-cost test. See case T-286/09 Intel v Commission ECLI:EU:T:2014:547 para. 150.

[84] Meo Serviços de Comunicações e Multimédia cit. paras 25, 34 and 37.

[85] This has not been remedied by the recent amendments of the DMA proposal by the European Parliament. See Draft Report on the proposal for a regulation COM(2020)0842 – C9‑0419/2020 – 2020/0374(COD) of the European Parliament and of the Council of 30 November 2021 on contestable and fair markets in the digital sector (Digital Markets Act), art. 6(1)(d).