Self-Preferencing: Between a Rock and a Hard Place

Following Google Shopping, competition authorities in Europe – supported by a flurry of reports on digital competition – have wasted little time establishing self-preferencing in competition law’s lexicon. But while self-preferencing looks here to stay, there is little clarity over the legal test which applies and how self-preferencing fits within the established canon of “leveraging” abuses. This article considers where self-preferencing should fit in light of “prior beliefs” as to its likely efficiencies and potential for competitive harm. It further considers how arbitrage of different legal standards by regulators (reverse regulatory arbitrage) can be avoided. These are critical questions as regulators and courts weigh up how to address the alleged harm from self-preferencing without undermining the dynamic efficiency of digital markets.

By Christian Ahlborn, Will Leslie, & Eoin OReilly¹

 

I. INTRODUCTION

Since the first Statement of Objections on Google Shopping was announced, competition practitioners, commentators and academics have grappled with the concept of “self-preferencing.”² What harm is it intended to address? What is the correct legal test? And how does it fit into the established canon of abuses under Article 102 TFEU?

Much ink has been split on whether “self-preferencing” is even an abuse or, as some have suggested, merely an inventive way of circumventing the (high) legal standard for “refusal to supply.”³ Others have pointed out that competitors have always had access to Google search results, and that the indispensability requirement should not be “extended” to cover situations where the dominant entity “actively changed its entire business to drive out competitors.”⁴

Yet, this debate risks being overtaken by events. Even while Google Shopping is under appeal before the General Court, the Commission has wasted little time establishing self-preferencing in the lexicon of competition law. Various official reports have endorsed self-preferencing as necessary to address the modern platform economy.⁵ Furthermore, the Commission and the Member States have taken steps to entrench self-preferencing in legislation. The Digital Services Act is expected to prohibit major online platforms from “certain forms of self-preferencing“ with similar initiatives underway in Germany and France.⁶

But while self-preferencing looks here to stay, there is little clarity over the legal test which applies and why. The General Court, which has recently heard the arguments in the Google Shopping appeal, now has an opportunity to provide some much needed clarity.⁷ This article outlines the key issues facing the General Court and how it may want to approach them.

 

II. WHAT IS SELF-PREFERENCING?

Self-preferencing occurs where a dominant firm uses an asset to give preference to its own complementary services over those of third parties. The favoring may concern different types of inputs including preferential access to “entry points” for customers and preferential access to data. In Google Shopping, the favoring concerned the positioning of Google’s comparison shopping service on the ‘first general results page in a highly visible place’ whereas in Amazon Marketplace the Commission has expressed concerns that Amazon is favoring its retail and private label business through preferential positioning in the “buy box” and access to seller data.⁸ Ultimately, self-preferencing is thus a special form of discrimination.

Competition authorities have, as a starting point, been wary of restricting firms from discriminating between customers. The Commission has only typically brought discrimination cases where discrimination fits as an element into an established pattern of exclusionary conduct.⁹ This is because price discrimination is both ubiquitous and frequently pro-competitive. Indeed, the growing use of algorithmic pricing means that pricing for services such as air travel, hotels and car rental fluctuates in real time. Furthermore, firms are in principle able to increase output – and thus welfare – by charging customers according to their willingness to pay.¹⁰ The ubiquitous and pro-competitive nature of discrimination is also seen in multisided markets where different pricing structures on either side are necessary to solve the chicken-and-egg problem of attracting all sides.

The vertical aspects of discrimination in favor of firms’ own upstream or downstream products add further layers of complexity: on the one (pro-competitive) hand, discrimination is often a natural consequence of vertical integration with firms seeking to integrate their operations and deliver efficiencies.¹¹ In this context, discrimination in favor of one’s own services is common to almost all markets: enterprises are, at their heart, a combination of tangible and intangible inputs which firms combine to produce a product or service.¹² Self-preferencing is thus inherent to the basic building blocks of undertakings.¹³ On the other (potentially anti-competitive) hand, firms may seek to leverage a dominant position into a related market by favoring their related product (resulting in anti-competitive foreclosure of rival third parties dependent on the dominant firm’s input).

Finally, neither vertical integration nor discrimination are binary questions. There are different types of vertical integration and discrimination. At one end of the spectrum, total refusal to supply is arguably the most extreme form with a firm opting not to supply the relevant input to any third parties (“closed systems”). A firm may also operate a partially closed system by supplying an input to some third parties but refusing to supply others. At the other end of the spectrum, vertical integration is open by its nature for many platform businesses with the commercial objective being to attract users to both (or more) sides of the platform (“open systems”). Equally, the intensity of vertical integration varies between firms and industries. Discrimination motivated by vertical integration is likely in many cases to be justified by efficiencies. In that regard, a firm may, for example, eliminate the “double margin” charged in the upstream and downstream markets or deliver significant efficiencies from integration of the business divisions themselves.

 

III. THE LEGAL TEST FOR SELF-PREFERENCING

Turning to the legal concept, self-preferencing fits under the umbrella of “leveraging abuses” which comprises several distinct forms of abuse where a dominant firm seeks to extend its market power into an adjacent market, each of which has established legal tests. Ignoring the established heads of “leveraging abuses,” the Commission veered off the beaten path in Google Shopping and adopted “self-preferencing” as a new head of abuse. The new abuse has three conditions:

• The dominant firm must engage in self-preferencing conduct which discriminates against competitors with a significant impact on an important parameter of competition;
• the self-preferencing has a significant impact capable of an anticompetitive effect; and
• there is no objective justification for the difference in treatment.¹⁴

Defending the decision not to apply an established test, the Commission argued that the conditions to establish one form of abusive conduct do not necessarily apply when assessing another form of conduct, unsurprisingly citing the paragraph in Teliasonera where the CJEU found that indispensability is not a requirement for a margin squeeze abuse.¹⁵

 

IV. WHERE DOES SELF-PREFERENCING FIT?

The Commission’s adoption of a “new” head of leveraging abuse is not merely an academic question. There are already a range of specific heads of leveraging abuses, namely: refusal to supply, tying and bundling, margin squeezes and abusive discrimination. The different leveraging abuses – and now self-preferencing – have different legal standards (i.e. the legal threshold to which the Commission must demonstrate that the conduct is abusive). Understanding how self-preferencing “fits” is thus critical to the substantive and administrative coherence of Article 102 TFEU.

A. Potential Overlap with Refusal to Supply and Abusive Discrimination

There are, in the first instance, clear parallels between self-preferencing, and on the one hand, refusal to supply and, on the other, abusive discrimination.¹⁶

Refusal to supply can be characterized as an extreme form of self-preferencing. But while all refusal to supply cases could be characterized as self-preferencing the same is not true in reverse. Self-preferencing also deals with situations where the dominant company continues to provide access to the relevant input, but on less favorable terms.

Conversely, while all self-preferencing cases can be classified as discrimination, not all discrimination is self-preferencing. Pure “second-line” discrimination, although rarely enforced, concerns situations where a dominant undertaking discriminates between two non-affiliated undertakings.

B. Differing Legal Standards for Self-preferencing and Abusive Discrimination

The fact that self-preferencing straddles refusal to supply and abusive discrimination is all the more salient because the three heads of abuse apply different legal standards:

Refusal to supply applies the “indispensability” standard as established in the seminal cases Commercial Solvents (1974) and Bronner (1994). To establish an abusive refusal to supply, the Commission or a private party must demonstrate that: (a) there is an actual or constructive refusal to supply an input; (b) the input is indispensable to compete effectively on the downstream market (i.e., there is no alternative input, even if less advantageous); (c) refusal to supply the input is likely to lead to the elimination of effective competition on the downstream market; (d) the refusal is likely to lead to consumer harm.¹⁷

Abusive discrimination, in contrast, applies the lower “effects” standard – as set out in MEO (2018)¹⁸ – which builds on well-established line of cases, notably Deutsche Bahn (1997),¹⁹ Tetra Pak (1994),²⁰ Clearstream (2009).²¹ To establish abusive discrimination, the dominant firm must have: (a) entered into equivalent transactions with other trading parties; (b) applied dissimilar conditions to those transactions (i.e. discrimination); and (c) the discrimination must place other trading parties at a competitive disadvantage (such that there is an anti-competitive effect).²²

Self-preferencing applies an effects standard manifestly closest to the effects standard elucidated in MEO in relation to abusive discrimination. In particular, the test eschews the indispensability and elimination of effective competition elements which, in practice, limit abuse conduct to scenario where firms held a monopoly on input which is critical for downstream competition. Indeed, the reference by the Commission in Google v. Commission to “capability” to restrict competition echoes the even lower “by object” standard which applies to tying abuses.

C. Does It All Add Up?

The similarities between self-preferencing and abusive discrimination and refusal to supply as well as the differing legal standards applicable to each give rise to the following important questions. First, given that similarities between the standards for self-preferencing and abusive discrimination, why treat self-preferencing as a distinct category? Second, can the different standards between self-preferencing and refusal to supply be justified? Third, how can the arbitrage of legal standards be avoided?

 

V. HOW TO FIT SELF-PREFERENCING WITHIN THE EXISTING LEVERAGING ABUSES

Addressing these three questions in turn:

A. Distinguishing Self-preferencing and Abusive Discrimination

The Commission’s adoption of self-preferencing may reflect both the need to differentiate a narrower subset of discriminatory conduct as potentially more problematic and avoid any ambiguity over the validity of applying abusive discrimination to self-preferencing.

Self-preferencing in the context of dominant online platforms is increasingly considered likely to harm competition given the inherent conflict of interest between the platform as both host and competitor, as well as perception on the part of users that platforms are neutral arbiters.²³ This is consistent with the Commission’s cases which have so far focused on dominant platform operators that allegedly control an “open” ecosystem which serves as a gateway into neighboring markets. As Vestager put it candidly, “Some of these platforms, they have the role both as player and referee, and how can that be fair?”²⁴ While these cases could fall within the scope of abusive discrimination, the Commission may want to signal that self-preferencing represents greater potential for competitive harm than previous discrimination cases by attaching a new label.

The Commission may have also wanted to dispel any questions over the legitimacy of applying abusive discrimination – as set out in Article 102(c) – to self-preferencing. Some commentators have previously questioned its applicability contending that the Article 102(c) is manifestly intended to address the (rare) occurrences of pure second-line discrimination rather than exclusionary conduct, which is more properly addressed under Article 102(b).²⁵ Equally, it is not clear that the conditions of Article 102(c) – which require the application of dissimilar conditions to ‘equivalent transactions’ would necessarily cover self-preferencing conduct where – at least in the case of Google Shopping – the self-preferencing did not concern transactions per se but rather the treatment of third parties under a search algorithm.

B. Distinguishing Self-Preferencing’s and Refusal to Supply’s Legal Standards

The Commission’s adoption of an effects standard for self-preferencing rather than an indispensability standard suggests that intervening to prohibit self-preferencing is regarded as less problematic (or more beneficial) than intervention to mandate refusal to supply. This has two elements.

The selection of different standards reflects, in the first place, a prior belief that self-preferencing carries a higher likelihood of competitive harm and a lower likelihood of counterbalancing efficiencies than refusal to supply. As Advocate General Jacobs outlined in Bronner, the rationale behind the high standard for obliging firms to provide “access” to their assets is the ‘careful balancing’ of short-term (allocative efficiency) considerations and long-term (dynamic efficiency) considerations.²⁶ Apart from certain “essential facilities,” refusals to supply are generally associated with high dynamic efficiency – and therefore a high error cost of regulatory intervention – which therefore justifies the higher legal standard. Conversely, the effects standard abusive discrimination reflects the ambiguous effects of price discrimination. As Advocate General Wahl outlined in MEO that ‘it is well established that a practice of discrimination, and a differential pricing practice in particular, is ambivalent in terms of its effects on competition.’²⁷ The two different standards thus reflect underlying degree to which the relevant conduct is likely to be efficient or have anti-competitive effects.

Similarly, the selection of an effects standard also suggests that it is more straightforward to successfully address the competitive harm from self-preferencing than from refusal to supply. The high standard for refusal to supply reflects in part the different nature of the mandatory remedies necessary to address the underlying concern. As Pablo Ibanez Colomo has observed, there is ‘a fundamental difference’ between prohibitory (e.g. cease and desist from discrimination) and mandatory remedies (supply the relevant third parties).²⁸ This is because mandatory remedies are inherently more difficult to implement as they require a competition authority or court to design them (e.g. to impose appropriate access terms and pricing). There is accordingly a greater error cost associated with refusal to supply remedies which justifies the higher standard. Conversely, abusive discrimination intuitively only entails a requirement that the dominant platform treat its affiliates according to these rules (i.e. a prohibitory cease and desist). This is in line with the Court’s statement in Van Den Bergh, which stated that indispensability was only applicable when the remedy required the dominant undertaking to ‘transfer an asset or enter into agreements with persons with whom it has not chosen to contract.’²⁹

C. Risk of “Reverse” Regulatory Arbitrage

Finally, “regulatory arbitrage” is often used to refer to a corporate strategy whereby firms use regulatory inconsistencies in order to circumvent unfavorable regulations. In this context, however, we are confronted with a different issue: the potential for regulators to exploit inconsistencies in law in order to avoid more burdensome legal standards.

The risk of reverse regulatory arbitrage primarily arises in relation to the overlap between the two heads of abuse were regulatory authorities to use self-preferencing rather than refusal to supply to take advantage of the former’s lower effects standard.

Regulatory arbitrage poses three concerns. First, it runs contrary to the principles of the rule of law. As Tom Bingham, the former Senior Law Lord in United Kingdom observed, ‘the law must be accessible and so far as possible intelligible, clear and predictable.’³⁰ In other words, dominant firms need to understand ex ante what rules are applicable to them.

Second, more practically, it incentivizes authorities at times to pursue the lower standard at the cost of inappropriate remedies. An infamous example is the failed remedies in the Commission’s Windows Media Player tying case. In reality, the Commission’s concerns in Microsoft WMP were Microsoft’s failure to “carry” alternative media players, i.e. a Bronner-type claim about access to the Windows “distribution network.” By framing its concerns as tying, the Commission was able to lower the required standard of harm (“by object/capability” standard rather than the indispensability standard). However, the price of this approach was that the Commission’s remedy to cease tying had no impact on competition in the market.³¹ Subsequently, when the Commission addressed a very similar set of facts with Internet Explorer, Microsoft agreed to implement a “browser choice screen” which resembled the type of “must-carry” remedy typically associated with refusal to supply cases.³² This example highlights a further reason to look closely at the remedies in question (in addition to the complexity issue outlined above): the appropriate remedy may help clarifying the type of abuse at issue: if a “must-carry” remedy is the right solution to the problem, then there is a fair chance that this problem is best characterized as a refusal to supply.

Third, and most importantly, regulatory arbitrage carries the risk there is a fundamental mismatch between the standard which the authorities apply and the correct ‘priors’ (in terms of likely harm and efficiencies) for the conduct in question.

 

VI. CONCLUSIONS

These three considerations shed some light on how self-preferencing fits, or should fit, between refusal to supply and abusive discrimination. The selection of an effects standard implies, on the one hand, that self-preferencing should not be applied to a “closed system” which would require a mandatory remedy (i.e. a dominant firm is required to provide access which it would not have otherwise given to any third party). Otherwise an indispensability standard would have been more appropriate given the potential implications for long-term investment incentives and the complexity of imposing a mandatory remedy.

It also roots self-preferencing within abusive discrimination, applying to “open” systems where a prohibitory remedy is possible (i.e. where the dominant firm is supplying its own upstream or downstream business as well as third parties).

The question mark remains, however, how self-preferencing will tackle the “hard” cases that straddle the divide. How, if at all, should self-preferencing apply to scenarios where the favoring relates to a “closed” feature of the dominant firm’s business? To what extent should self-preferencing apply to scenarios where regulators need to devise remedies for inputs that have not been made available to the market? Furthermore, if self-preferencing should cover these scenarios, is it still appropriate for an effects standard to apply? In particular, how will the Commission seek to weigh the trade-off between the short-term (allocative) efficiencies (from prohibiting self-preferencing) and the long-term (dynamic) efficiencies from potentially distorting investment incentives.

Similar questions arise in relation to regulatory arbitrage. Given that the self-preferencing is decisively rooted as an “effects” abuse, will the EU Courts delineate a clear dividing line between the scenarios to which refusal to supply applies (e.g. “closed” systems) and the scenarios to which self-preferencing applies? Similarly, will the EU Courts limit abusive discrimination to pure second line discrimination to make “space” for self-preferencing or simply allow the two abuses to implicitly overlap given the similar standards?

The answers to these questions, amongst others, will shape the boundaries of the new head of abuse.

---

1 Lawyers at Linklaters.

2 See Press Release of European Commission, Antitrust: Commission sends Statement of Objections to Google on comparison shopping service; opens separate formal investigation on Android (available here).

3 See, for example, Bo Vesterdorf, Theories of Self-Preferencing and Duty to Deal – Two Sides of the Same Coin?, Competition Law & Policy Debate, Volume 1, Issue 1, February 2015, pages 4-9; and the reply of Nicolas Petit, Theories of Self-Preferencing Under Article 102 TFEU: A Reply to Bo Vesterdorf., Competition Law & Policy Debate 1 CLPD, 2015. See also Pablo Ibanez Colomo, Indispensability and Abuse of Dominance: From Commercial Solvents to Slovak Telekom and Google Shopping, Journal of European Competition Law & Practice, Volume 10, Issue 9, November 2019, Pages 532–551.

4 See Thomas Hoppner, Duty to Treat Downstream Rivals Equally: (Merely) a Natural Remedy to Google’s Monopoly Leveraging Abuse, 1 European Competition and Regulatory Law Review (CoRe) Issue 3/2017, pages 208-221.

5 Jacques Cremer, Yves-Alexandre de Montjoye, and Heike Schweitzer, Competition Policy for the digital era, 2019 (available here); Federal Ministry for Economic Affairs and Energy (Germany), A new competition framework for the digital economy: Report by the Commission ‘Competition Law 4.0’, 2019 (available here); Digital Competition Expert Panel, Unlocking digital competition, 2019 (available here).

6 Commission’s Inception Impact Assessment for Digital Services Act package: ex ante regulatory instrument of very large online platforms acting as gatekeepers in the European Union’s internal market, 2020 (available here). See also, Federal Ministry of Economy and Energy (Germany), Referentenentwurf des Bundesministeriums für Wirtschaft und Energie Entwurf eines Zehnten Gesetzes zur Änderung des Gesetzes gegen Wettbewerbsbeschränkungen für ein fokussiertes, proaktives und digitales Wettbewerbsrecht 4.0 (GWB-Digitalisierungsgesetz), 2020.

Autorité de la concurrence, The Autorité de la concurrence’s contribution to the debate on competition policy and digital challenges, 2020.

7 Case T-612/17, Google and Alphabet v. Commission, [awaiting judgment].

8 AT. Google Search, para. 650; See Commission Press Release, Antitrust: Commission opens investigation into possible anti-competitive conduct of Amazon, 2019 (available here).

9 See for example Case C-209/1, Post Danmark A/S v. Konkurrenceradet, (CJEU, 2012), para. 30.

10 See Frank Ramsey, A Contribution to the Theory of Taxation, Economic Journal, 1927.

11 See Pablo Ibanez Colomo, supra note 3.

12 See Ronald Coase, The Nature of the Firm, Economica, New Series, Vol. 4, No. 16., 1937, pages 386-405.

13 See Alfonso Lamadrid, Google Shopping Decision – First Urgent Comments, Chillin’Competition, 2017.

14 Google and Alphabet v. Commission, supra note 7.

15 Case C-52/09, Teliasonera, para. 55.

16 We note that a number of commentators have argued that self-preferencing is, alternatively, best understood as a form of tying. See, for example, Benjamin Edelman, 2014, Leveraging Market Power Through Tying and Bundling: Does Google Behave Anti-Competitively? Harvard Business School Working Paper, No. 14-112, 2014; Edward Iacobucci & Francesco Ducci, 2019. The Google search case in Europe: tying and the single monopoly profit theorem in two-sided markets, European Journal of Law and Economics, Springer, vol. 47(1); and Fumagalli, Chiara, Massimo Motta, and Claudio Calcagno, Exclusionary Practices: The Economics of Monopolisation and Abuse of Dominance, Cambridge University Press, 2018.

17 See in particular Case C-7/97 Oscar Bronner GmbH (1998); Joined Cases 6 and 7/73 Commercial Solvents (1974); Joined Cases C-241/91 P and C-242/91 P, Magill (1995); Case T-201/04 Microsoft (2007). Some argue, and the Commission’s guidance supports, that there is, or should be, a lower standard for termination of an existing supply arrangement. However, this is not clearly reflected in the case-law nor is it clear whether the standard would significantly differ from the indispensability standard. [See Guidance on the Commission’s enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings (2009/C 45/02)].

18 Case C-525/16, MEO – Serviços de Comunicações e Multimédia SA v. Autoridade da Concorrência (CJEU, 2018).

19 Case T-229/9, Deutsche Bahn AG v. Commission (General Court, 1997).

20 Case T-83/91, Tetra Pak International SA v. Commission (General Court, 1994).

21 Case T-301/04, Clearstream Banking AG and Clearstream International SA v. Commission (General Court, 2009).

22 See Article 102(c) TFEU and MEO, supra note 18.

23 https://ec.europa.eu/competition/information/digitisation_2018/contributions/linklaters.pdf.

24 New York Times, Big Tech’s Toughest Opponent Says She’s Just Getting Started, 19 November 2019.

25 Damien Geradin and Nicolas Petit, ‘Price Discrimination under EC Competition Law: The Need for a case by case assessment’ GCLC Working Paper (2005).

26 Bronner, AG Op. Jacobs, para. 57.

27 MEO, AG Op. Wahl, para.62.

28 Pablo Ibanez Colomo, supra note 3.

29 Case C-552/03 P – Unilever Bestfoods (previously Van den Bergh Foods) v. Commission, para. 137.

30 Tom Bingham, The Rule of Law (Penguin: 2011).

31 See, for example the assessment of Nicholas Economides and Ioannis Lianos, A Critical Appraisal of Remedies in the EU Microsoft Cases. Columbia Business Law Review, Volume No. 2, 2010.

32 The Commission’s Google Android decision, which was also considered under the legal standard for tying, has led to Google agreeing to a similar “choice screen” for search and browsers on Android.