Self-preferencing between all and nothing: in search for a definition under Brazilian competition law

Anna Binotto

PhD Candidate, University of São Paulo (Brazil), São Paulo, Brazil, ORCID https://orcid.org/0000-0001-7225-4104.

anna.massaro@usp.br

Patrícia Deluca

Master, University of São Paulo (Brazil), São Paulo, Brazil
https://orcid.org/0009-0005-0244-0223.

patricia.deluca@usp.br

Received: November 15th, 2022 | Accepted: May 14th, 2023

How to cite: Binotto Anna and Patricia Deluca. “Self-preferencing between all and nothing: in search for a definition under Brazilian competition law”. Latin American Law Review n.º 11 (2023): 73-92, doi https://doi.org/10.29263/lar11.2023.04

Abstract

The term self-preferencing has gained great prominence in the competition and regulation debate involving digital platform markets. However, it has increasingly been used to designate practices with distinct structure, economic rationale, and effects, including in contexts that fall outside of digital markets (even though those conducts already existed in preceding times and were widely scrutinized by competition authorities under the existing legal grounds). In Brazil, the situation is no different. This is due to the lack of a well-established legal test, including definitions about presumptions of illegality and burden of proof, and the attempt to concentrate, under the same “umbrella” diverse and distinct conducts. Following up on debates in international scholarship and, most relevantly, discussions within the European scenario, this article provides initial suggestions for the development of an adequate legal test for the practice of self-preferencing in Brazil. We make a preliminary attempt to differentiate between diverse categories of self-preferencing and suggest, among other aspects, that tests and theories of harm existing in CADE case law should be improved and adapted to drive antitrust analysis for those practices. This is the case, for example, for the classic (or “pure”) pattern of what the scholarship has equated to self-preferencing, being possible to turn to an adaptation of the CADE’s position on the essential facilities doctrine (EFD). We argue the essentiality test should not be completely abandoned, as it stands as a relevant divisor between distinct types of self-preferencing, with relevant enforcement implications, and provide an initial outline of an adapted legal test to assess essentiality.

Keywords

self-preferencing, digital markets, platforms, essential facilities, refusal to deal

Self-preferencing entre todo y nada: en busca de una definición según la ley brasilera de defensa de la competencia

Resumen

El término “self-preferencing” ha ganado gran prominencia en el debate sobre competencia y regulación en los mercados de plataformas digitales. Sin embargo, cada vez más se ha utilizado para designar prácticas con estructuras, fundamentos económicos y efectos distintos, incluso en contextos que no pertenecen a los mercados digitales (aunque estas conductas ya existían en tiempos anteriores y haya sido ampliamente examinadas por las autoridades de competencia bajo fundamentos legales existentes). En Brasil, la situación no es diferente, debido a la falta de un criterio legal bien establecido, incluyendo definiciones acerca de presunciones de ilegalidad y la carga de la prueba, y al intento de agrupar bajo la misma “sombrilla” conductas diversas y distintas. Siguiendo los debates en la bibliografía académica internacional y, especialmente, las discusiones en el escenario europeo, este artículo ofrece sugerencias iniciales para el desarrollo de un criterio jurídico adecuado para la práctica de self-preferencing en Brasil. Presentamos un intento preliminar por diferenciar entre diversas categorías de self-preferencing y se sugiere, entre otros aspectos, que los criterios y teorías del daño existentes en la jurisprudencia del CADE (Consejo Administrativo de Defensa Económica) deben ser mejorados y adaptados para analizar la competencia en relación con esas prácticas. Esto es válido, por ejemplo, para el tipo clásico (o “puro”) de lo que la bibliografía académica ha equiparado con la self-preferencing, siendo posible recurrir a una adaptación de la posición del CADE sobre la Doctrina de las Facilidades Esenciales (Essential Facilities Doctrine - EFD, por sus siglas en inglés). Argumentamos que la prueba de esencialidad no debe ser abandonada por completo, ya que representa una división relevante entre distintos tipos de self-preferencing, con importantes implicaciones para su aplicación. Presentamos también el esquema inicial de una prueba jurídica adaptada para evaluar la esencialidad.

Palabras clave

Self-preferencing, mercados digitales, plataformas, essential facilities, negativa a contratar

___________________________________________________________________________

INTRODUCTION

The concept of self-preferencing is key to recent discussions about the regulation and promotion of competition in digital platforms. It appears in a series of studies and reports that have sought to understand potential changes, proposals, and new contours of competition in these markets. And it could not be otherwise. The strategy called self-preference could be considered a relatively customary practice in the activities of platforms operated by large players in digital markets.1

With the interest in the topic, we also witnessed a dizzying growth in the search for the term in the academic field and in the practical activity of antitrust authorities, especially since 2015 and, more intensely, since 2019, when European Union reports were released.2 The figure below illustrate this scenario worldwide:

Figure 1. Google books Ngram viewer: self-preferencing (2009-2019)3

Source: Authors, based on Google Ngram Viewer tool.

Figure 2. Google Trends: self-preferencing (2009-2022)

Source: Authors, based on Google Trends tool.

However, a deeper analysis shows that the expression is used to refer, at the same time, to multiple - and distinct - practices, without its use being linked to the construction of a specific theory of competitive harm or a specific legal test. The result is negative for the development of the concept if self-preferencing can mean “everything,” in practice it can also mean “nothing”.

This is not at all positive. Scholars have long criticized attempts to develop a universal legal test for unilateral conduct.4 The perils of a test that is overly wide are multiple: besides the lack of legal certainty and potential restrictions to the right to proper defense within investigations, as they may create imbalances on the burden of proof and contrafactual assessment; unclear standards lead to obstacles in applying adequate and proportional sanctions, as well as designing remedies that effectively revert the potential anticompetitive effects deriving from the conduct. The pitfalls unfold into ineffective competition advocacy, with the misguidance of market players, weakened deterrence, ambiguous interpretation of legal and regulatory provisions that may apply to the activity or market in question, and reducing the incentives for innovative business strategies with potential pro-competitive effects.

The movement around the topic of self-preferencing has seemed to follow exactly that path. The issue becomes even more relevant in the context in which proposals are being discussed in foreign jurisdictions, aiming at prohibiting self-preferencing practices completely, from ex-ante regulation, as is the case of the Open Innovation Act and the Digital Markets Act.5 In other cases, such as the recent change in the German competition law, the practice receives a presumption of illegality when adopted by agents with transversal market power (i.e., in more than one related market).6 Competition authorities in Japan,7 Korea,8 Australia,9 and others are also considering regulations on the subject.

If this is the case, some innovative and disruptive business models that eventually rely on self-preferencing mechanisms, but are still procompetitive, can be discouraged or become unviable, due to unclear standards for what can and what cannot be done in relation to product placement, design of promotional incentives, categorization of sellers and users, algorithmic pricing, targeted advertising, and third-party partnerships - to name a few.

In light of this scenario, this article assesses (i) how the expression “self-preferencing” is being interpreted and applied in the Brazilian context, especially by the Administrative Council for Economic Defense (“CADE”); (ii) the possible contribution that the resort to the new term brings to competition analysis, comparatively to practices and theories of harm already well known and tested by the antitrust community, and (iii) whether pre-existing economic and legal foundations, in the present study namely the essential facilities doctrine (EFD), can be used in current discussions - without, of course, disregarding the particularities of the context and dynamics of digital markets.

Following this introduction, Section II shall focus on examining the practices currently encompassed in the definition of self-preferencing, abroad and in the Brazilian context, in an attempt to systematize the modalities that have been referred to in recent cases, to bring more clarity on the competition analysis applicable to conducts framed as self-preferencing. Section III explores the elements that are present in theories of assessment of refusal-to-deal, as to identify to what extent they can draw near to “pure” self-preferencing. The last section presents the conclusions of this study.

IN SEARCH OF A USEFUL DEFINITION FOR SELF-PREFERENCING PRACTICES

The main reference to a conduct of self-preferencing appeared in the European Commission’s decision that condemned Google for favoring, on its web search feature, results from its price comparison site - Google Shopping - to the detriment of other companies that competed with the latter service.10 In that case, the conduct was treated as an unfolding of leveraging, i.e., as a strategy to abuse the dominant position in one market (the organic search market) to gain power in another adjacent market (the price comparison market). Specifically, the European Commission characterized the conduct of self-preference based on the identification of three elements: (i) the existence of discrimination against competitors in an adjacent market, with significant impacts on their capacity to compete; (ii) the ability of the conduct to generate anticompetitive effects in the markets involved; and (iii) the absence of objective justification for the difference in treatment between the economic agents.11 The decision was criticized for allegedly confusing a form of abuse (i.e., leveraging as a standalone and independent conduct) with a category of abuse (i.e., leveraging as a theory of harm that could unfold into different conducts, such as tying, margin squeeze or refusal to supply),12 an aspect that was later partially enlightened by the General Court’s decision.

Over the years, however, the same name has been given to conducts imputed to other agents that adopt commercial practices that differ from those of Google.13 In this proposed framework, for example, it is not trivial to distinguish what would be the “new” conduct of self-preferencing from a “traditional” conduct of anticompetitive discrimination, refusal to deal, tying, bundling, cross-subsidization, among others. This is, in fact, the criticism that some make of this new “type”.14

In foreign jurisdictions, examples are the investigations on (i) the discriminatory treatment allegedly adopted by Amazon for the use, by advertisers present in the marketplace, of a prominent space, the so-called “Buy Box”, conducted by the European Commission15 and by the CMA (United Kingdom);16 (ii) the charging of fees for the use of third party payment systems and the distribution of third party applications in Google’s and Apple’s app stores, before authorities such as the European,17 North American, Australian,18 Dutch,19 among others;20 (iii) the differentiated treatment given to certain commercial partners that contract more than one service offered by the platform, as is the case of Amazon’s conviction before the Italian antitrust authority.21

Before CADE, this scenario is no different, as the present study will show. Based on keyword research for the expression “self-preferencing” in the general search tool of CADE’s Electronic Information System (SEI), one can already notice an increasing number of results in the last 4 to 5 years.

Figure 3. CADE Research - Occurrence of the term “self-preferencing” (2013-2022)

Source: Authors. Based on CADE’s Electronic Information System (SEI/CADE). Last updated September 2022.

The rising incidence of the expression in CADE’s files does not however account for a growing body of case law or decisional practice on the subject. If anything, it indicates a scenario of great uncertainty related to the topic. Analyzing the cases currently under examination or recently decided by the authority in which the reference to “self-preferencing” has appeared, besides the Brazilian Google Shopping case, when it was first applied in a more substantial way (but still far from a structured legal test), there is only one other case that can be highlighted as a unilateral conduct investigation in a digital market. It refers to the complaint filed by the Brazilian Association of Workers’ Benefit Companies (ABBT), accusing iFood, an online food ordering and food delivery leader, of privileging its own benefit voucher (iFood Benefícios) as a means of payment in its delivery platform.22 It is important to note that, in this case, the claimant tried to specifically sustain that the delivery platform would constitute an essential activity for the development of the benefit voucher market, exactly in line with the suggestions presented in this article.23 The case is currently under investigation by the General Superintendence (“SG”).

But for what do the other occurrences of the expression “self-preferencing” in CADE’s library stand? Self-preferencing is mainly being referred to outside the context of investigations of anticompetitive conduct, in merger review cases, both in CADE’s decisions and in documents filed by the parties and other market players. There are indications that certain concentrations can increase incentives to adopt self-preferencing: the main one involved the verticalization of a digital marketplace with a seller (Nike/SBF-Centauro).24 In that case, the SG had dismissed the competition concerns related to the practice of self-preferencing,25 suggesting the approval without restrictions. However, when provoked by rival Netshoes, CADE’s Tribunal conditioned the approval of the transaction to the adoption of a Merger Control Agreement (ACC) that, among other things, provided behavioral remedies specifically aimed at mitigating the risks of self-preferencing.26

Since CADE did not sanction the Google Shopping case, the Nike/SBF-Centauro merger case is perhaps the most relevant intervention conducted by CADE within the self-preferencing discussion, with remedies designed to specifically address that concern. Notably, merger review cases that involve different assessments and legal tests applied to evaluate if a merger enhances the risk of potentially anticompetitive concerns differs from a reactive investigation of a material conduct that has effectively occurred. Therefore, the Nike/SBF-Centauro merger case is not a strong precedent and does not present a clear or sufficient legal ground to assess self-preferencing, despite representing a relevant step toward the outlining of the conduct under Brazilian law.

It is important to note, however, that, more surprisingly, in other recent merger cases, the term has been used to refer to concerns unconnected to digital platform markets. In the transaction between Sulamerica, the leader in the healthcare insurance market, and Rede D’Or, a large hospital chain, for example, market players pointed at the risk that verticalization would induce self-preferencing of Rede D’Or hospitals by Sulamerica, for instance through “imposing difficulties for the reimbursement process of competing hospitals, or even the possibility of Sulamerica making it difficult for competing hospitals to use its healthcare plans”.27 In this case, however, although one can fully discuss the potential competition risks resulting from verticalization - and there are numerous theories of harm that have been tested for years in this regard - it is clear that this is not an indispensable digital platform, and the mention of the practice of self-preferencing is detached from the context in which it was originally used.

The same is true of the transaction between CMA CGM and Maersk,28 in the container shipping market,29 where the issue arose at the request of the Brazilian Association of Port Terminals (ABTP). The discussion of the topic led the SG to include in its opinion a general definition of the practice of self-preferencing as: “the adoption, by a given company or economic group, of actions aimed at favoring its own products or services over those of its competitors. Such practice is usually associated with the conduct of leverage, in which a company takes advantage of its dominant position in a given market to strengthen its position in another market.” As one can see, the definition is broad, confusing, as leverage is treated as a conduct in itself, and not accompanied by a specific contextualization, either in relation to digital markets, or in relation to the platform structure (e.g., multi-sided market), or in relation to the characteristics that would be necessary for the practice to be framed as anticompetitive (e.g., following assessment of efficiencies that may outweigh restraints to competition).

Thus, the use of the expression “self-preferencing” as a generic legal category may not benefit competition law and, more specifically, CADE’s punitive action. Quite the contrary: framing conducts with different structures, economic rationality and effects under a single umbrella tends to create confusion both for the antitrust authority, which will not have a starting point, methodology and clear criteria to be applied when analyzing a concrete case, and for economic agents, who loose in legal certainty, and may be induced to refrain from adopting pro-competitive practices.

The use of the expression to qualify conducts that fall outside the scope of the digital platforms’ realm may also ignore some of the consensual peculiarities of such markets, which are not commonly found in non-digital platform markets. Most relevantly, besides the connection of different players and stakeholders in multi-sided markets; indirect externalities of network effects and distinct market power and incentives to discriminate of the vertically integrated owner of the platform,30 non-digital platform markets are, generally, not marked by some of the features that distinguish self-preferencing. Indeed, digital markets are subject to unique forms of market tipping, which benefit from unmatched returns to scale and economies of scope, including specific network effects that weaken multi-homing. They are also subject to low marginal and distributional costs and global reach.31

Although we do not necessarily reject the possibility of self-preferencing practices in non-digital platform markets, we are critical of (i) the unrestricted appeal to the expression in a free-ride of its rising importance to deal with typical vertical integration and vertical restraints cases; and (ii) as a “catch-all” for conducts in digital markets that does not clarify central elements of a type. Both lead to uncertainty of foundational factors, which are essential to the application of administrative sanctioning law, especially the regime of presumptions, burden shifting, and weighting and evaluation of evidence, issues that will thus be defined by the authorities on a case-by-case basis.

Precisely because of that, Pablo Colomo suggests that a segmentation be made according to different variables, as the author traditionally indicates for unilateral conduct analysis:32 first, separating practices involving horizontal relations (that is, that analyze adjacent markets or complementary products) from those involving vertical relations; second, in terms of the degree of preference given to affiliates, which can be total (with a complete closing of the market to third parties), or partial (with favoritism).33

There are also other differences that can be observed, especially important for designing the appropriate forms of antitrust intervention, when necessary (whether reactive or proactive, behavioral, or structural). When the practice is related to the company’s core business, remedies determined to overcome the competition distortions would require a more comprehensive change in the business model (e.g., the monetization strategy, the contractual relations with users, the portfolio of services offered, the API etc.). On the other hand, when the practice involves aspects or activities that are only ancillary to the main activity in a given market, the remedies required to reverse its effects are possibly simpler and may involve only fine-tuning adjustments, such as changing the display formats of the content made available to the buyers of the services and/or products offered in the platform. Other authors have already made similar proposals.34 Based on these variants, it would be possible to build a typology of practices usually included in the universe of the expression self-preferencing and bring them closer to the existing classification for the analysis of unilateral conduct.

In the case of vertical practices (i.e., those involving vertically related markets), the effects on activities may, in principle, approach refusal to deal, or anti-competitive discrimination, depending on the degree of favoritism. On the other hand, horizontal practices, related to complementary products or adjacent markets, also depending on the level of favoritism, may equate to tying or bundling: the first, when favoritism is absolute, and the second, when it is not. Issues such as the pre-installation of complementary products in operating systems are well-known in antitrust, and refer to paradigmatic long-standing foreign cases, such as the IBM and Microsoft cases.35

There are also transversal strategies that aim at entering other markets not previously explored, in which a dominant company bears financial losses to finance new activities (a “cross-subsidy”).

Despite being potentially applicable to other jurisdictions - as we do not necessarily envision particularities related to the strategies of implementation of self-preference in Brazil compared to other countries - we center our assessment in the Brazilian context, given the specific development of the jurisprudence, particularly case law involving the essential facilities doctrine (“EFD”), as we further discuss below.

We agree that certain categories of conduct should not be confused with the underlying theory of anticompetitive harm, in that sense, much in line with criticism that followed the Google Shopping decisions. We suggest that a legal test to separate lawful and unlawful self-preferencing practices may benefit from distinguishing between the different strategies that may be included under the self-preferencing umbrella, considering both an approximation to traditional and specific categories of abuse and the elemental underlying business strategy. The summary table below presents a first attempt to systematize the modalities of anticompetitive self-preferencing:

Chart 1. Different strategies of anticompetitive self-preferencing36

Suggested classification	Underlying business strategy	Description	Relationship between markets	Possible Approximations	Example in case law
Pure Self-preferencing	Platform foreclosure	Platform that competes with its users and favors its own products/services offered	Vertical	Refusal to deal	Amazon BuyBox Google Shopping
Leveraging self-preferencing	Entrance support	Platform that taps its customer base to gain prominence in another relevant market where it was not active before	Horizontal Vertical	Predatory pricing Indirect discrimination (subsidy)	iFood Benefits (Brazil)
Defensive self-preferencing	Market power maintenance	Platform that uses cross- market strategies between adjacent markets to strengthen and protect its market power. It may even frame favoritism to certain partners37	Horizontal	Bundling and tie-in sales	Google vs. Epic Games Apple App Store Google App Store Dating Apps (Netherlands) Amazon FBA (Italy)

Source: Authors.

Indeed, for different unilateral conducts, antitrust commentators and authorities, CADE included, have developed different tests, or “legal criteria of illegality”.38 And, as much as the lack of harmonization and consistency of these criteria39 can be criticized, one cannot ignore them completely, in the aim of developing a new and comprehensive theory of damage. On the contrary, they should serve, at least, as a starting point for the analysis of such “new” conducts. It is precisely because these criteria often construct themselves as “confusing legal shortcuts and presumptions of legality”41 that the unbridled framing of different conducts under a single category tends to be problematic.

In this article, we shall focus on the first category defined in the chart above, called the “pure” self-preferencing, which relates to the practice of self-preferencing in the context of vertical restrictions and foreclosure of an essential platform. We argue that this is a necessary first step, to better understand to what extent the parameters of analysis already established to investigate conduct occurring in traditional markets can be useful to the development of one - or multiple - legal theses to support cases involving self-preferencing.

“PURE” SELF-PREFERENCING AND GROUNDS FOR REFUSAL OF ACCESS

What we refer to as a “pure” strategy of self-preferencing is the one when a platform is essential or indispensable to market players that compete with the platform’s operator. As we shall claim below, the assessment of essentiality is a key aspect of the legal test suggested to identify the “pure” form of self-preferencing that is abusive and unlawful. Pure self-preferencing may involve strategies, including explicit refusal of access, favoring the platform’s own products or services, and other forms of discrimination. Indeed, those strategies may also be identified within the other categories we laid out in Chart 1; but what singles out “pure” self-preferencing is the fact that third-party competitors are left with no viable alternative than that of the essential platform and are therefore dependent. We understand that antitrust authorities - and CADE in particular - can revert to existent case law to identify the hypothesis of the pure form of self-preferencing.

As discussed above, a portion (and only a portion) of what is currently called self-preferencing may come close to practices linked to a concern for competitors’ access to a platform. The fact is that the conformation of markets (which is fundamental for the concept of self-preferencing) in which players that dominate relevant technologies are also competitors of those that depend on them, is not new to competition law.40 On the contrary, except for the inclusion of the term “technology”, the characteristic of bottleneck infrastructure domination, whose access is essential for agents active upstream, downstream, or in adjacent markets has already been the subject of different theories by the antitrust scholarship, notoriously the essential facilities doctrine, or the EFD.41

The EFD is applied to assess the conditions in which a dominant player (rectius: monopolists) may be required by competition law and antitrust enforcement to share a certain asset or infrastructure. The doctrine, as developed more recently,42 limits its incidence to situations in which there is an economic capacity and incentive for the holder of the infrastructure to exclude a competitor who wishes to access it - that is: when there are potential anticompetitive effects arising directly from a total or partial refusal-to-deal conduct. This formulation is precisely reflected in CADE’s case law involving EFD, where the doctrine is discussed, and recognized, but applied with limitations.

This restriction is particularly important given the fact that the Brazilian Constitution grants a general principle of freedom to act, so that the State may intervene, interfering in the private contractual freedom, if and only to address practices that exceed the regular exercise of the economic activity. More than that, antitrust interventions should be residual and eventual, in order not to cause undesirable market distortions, as the mandatory opening of access to competitors may reduce the incentive to innovate. If this is true for antitrust discipline in general, it is especially true for practices that involve behavioral remedies.43 For this very reason, leading Brazilian scholar Calixto Salomão Filho suggests that the EFD be used “for cases of extreme economic concentration”44, involving markets with natural or structural monopoly formats. Indeed, we do not intend to overlook the scholarly criticism of a comprehensive and extensive resource to the EFD, both in Brazilian45 and international literature.46 Neither do we suggest that one should apply EFD under its traditional formulation. Rather, we propose the EFD could be used as a starting point for a novel theory of harm for self-preferencing for digital platforms whose features are akin to those of an essential facility.47

In the case of essential infrastructures, CADE’s traditional case law usually requires four elements to establish anticompetitive conduct, namely: the existence of an essential good held by a monopolist; legal or economic impracticability of its duplication; the refusal to grant access to the good to a competitor in the target market; and the practical ability of the monopolist to provide access to the infrastructure without compromising the access previously granted.48

More recently, however, the first requirement - the essentiality of the good - seems to be called into question. In this context, it is relevant to mention the B3 case decided by CADE,49 which raised discussions about the nature of the services that were the object of accusations of restriction on the freedom to contract and the applicability of the EFD. In that case, CADE discussed services with economic characteristics of an essential facility, without, however, having fixed a strict understanding about its qualification.50 What was noticed, with CADE’s decision in the B3 case, was a broadening of the concept to cover situations in which the asset in question has a “similar economic function” to the one performed by essential facilities.

A relaxation of the concept has also been suggested in the investigation arising from the fintech Nubank’s claim reporting a refusal to contract with major Brazilian financial institutions. Although the claimant pointed to automatic debit services as “essential facilities” for certain aspects of its activities in the financial market,51 CADE SG chose not to examine this issue further, at least preliminarily.

It seems, therefore, that for some time CADE has been developing a theory of harm to deal with cases involving refusal to contract outside the classic context of natural monopolies and infrastructure sectors. This effort does not ignore the rigidity of the parameters for intervention in private relationships, it only adapts the criteria developed for the EFD to other circumstances. This accumulation cannot be lost with the introduction of a new “type” in the list of potentially anticompetitive conducts.

As a starting point for the exercise proposed in this study, it is important at least to try to apply the above-mentioned criteria for the characterization of EFD to the practices of “pure” self-preferencing in digital markets. As particular weight was given to the assessment of the essentiality of the good or service refused, a revamped EFD for self-preferencing should then be focused on the analysis of the essentiality of the platform.

Not without reason, some attempts have been made at bringing more concreteness to the definition of essentiality, as is the case of the regulation of the concept of gatekeeper platforms.52 The issue of essentiality is indeed the main challenge in relation to the burden of proving the likelihood of the anticompetitive effects of the practice particularly in digital environments. This is why the European Commission decided to depart from and discharge this burden in the Google Shopping case.53 Although the decision was later criticized, it was, to some extent, reaffirmed when the General Court of the European Union upheld the condemnation.54

Nevertheless, as defended in this paper, this does not mean that one should completely disregard the doctrine, but rather adapt it, by adopting other similar criteria discussed and deepened by scholarship and case law.

In this sense, Calixto Salomão Filho indicates the existence of two assumptions underlying the legal reasoning behind the EFD: the existence of an economic situation of dependence and the impossibility of overcoming such dependence by alternative means of access to goods or services dominated by third parties – this would be more important than the identification of the infrastructure itself, or the qualification of the good. Extreme dependence would therefore be the essential element for EFD.

Following this rationale, in the Brazilian context, the EFD should be analyzed under the interpretation of the principle of the social function of property – especially the ownership of production assets and, notably, of the so-called “necessary access assets” for competitive entry and permanence in a market.55 Approaching this reasoning from the construction of a damage theory for self-preferencing, it is clear that not any platform has the capacity to generate the anticompetitive effects linked to the practice. Even platforms with market power may not be able to generate “extreme dependence” on their users, especially when there are available alternatives.

Similarly, in the context of the European Commission, the case law developed from paradigmatic cases, such as the Bronner case in 1994,56 the criterion of indispensability of the product or service, which would be central to segregate the refusal to deal that is an expression of private autonomy and freedom to contract assured to the economic agents and the one that may generate anticompetitive effects and, therefore, be considered illegal.57 This criterion is considered an important indicator of the cost-benefit of intervention by the antitrust authority in the private contractual relationship. Once again, the standard applied is high. The relevance or importance of a platform as a sales channel (for example) does not necessarily imply its indispensability for its users. That is, this is another key factor limiting the potential anticompetitive effects of a self-preferencing practice.

The choice of reinterpreting “essentiality” as indispensability and dependence may work as an intermediate standard between the EFD and the total absence of a parameter to assess self-preferencing cases.

Going further with the application of the three other (even though less significant) requisites of the traditional EFD to a pure self-preferencing case, the framing of the conduct as a competition offense may encounter more obstacles, especially in relation to the requirement of non-duplication of the good. At this point, it is not merely the inexistence of technology to develop a similar digital product - which seems a rare hypothesis to be verified in practice - but also a circumstantial analysis of whether any competing platform entrance or maintenance is probable, timely and sufficient based on the other factors that made the platform a relevant channel for the two (or more) sides connected by it, especially data, client attraction and network effects - which, although more feasible, are often difficult to measure. The actual refusal to access should also be reinterpreted to include situations in which the competitor’s presence in the platform leads to a zero or near zero economic results. Also, the ability to share access to the platform may depend on technical issues linked to the architecture of the microsystems that support it, such as integration and algorithms, and commercial issues related to private agreements.

The chart below attempts to summarize and compare the rigorous traditional EFD to its adapted version to digital platforms and self-preferencing antitrust assessment, as proposed in this article:

Chart 2. Adapted EFD test for the assessment of pure self-preferencing

EFD	Adapted EFD for pure self-preferencing
Existence of an essential good held by a monopolist.	Essential, indispensable and/or non alternative platform (also referred to as bottleneck or gatekeeper structures).
Legal or economic impracticability of its duplication.	Circumstantial unviability of the platform duplication, based on the assessment of existing barriers to entry (whether any competing platform entrance or development is probable, timely and sufficient) also considering data and client (two sides) attraction (network effects).
Refusal to grant access to the good to a competitor in the target market.	Assessment of whether the conduct is prohibitive in practice (i.e., competitor entrance and maintenance in the platform leads to a zero or near zero gains).
Practical ability of the monopolist to provide access to the infrastructure without compromising the access previously granted.	Assessment of the access to the platform in light of technical and/or commercial aspects, such as systems and microsystems integration, algorithms, and private agreements.

Source: Authors.

One should interpret the above suggestions within the general framework for investigating unilateral conducts as anticompetitive. Indeed, the antitrust authority will usually evaluate market power (e.g., dominance), the potential of anticompetitive effects or restrictions to competition (e.g., exclusionary effects), and possible counterbalancing efficiencies. When it comes to the enforcement implications of our proposals, we anticipate a matter of degree of rigorousness and burden of proof. Traditionally, under a “by effects” or “rule of reason” assessment, or even considering a “by object” methodology, there are distinct levels of the burden of proof undertaken by the authority or room for efficiencies arguments by the parties to reach a final understanding on the (il)legality of conducts. That said, we agree that the essentiality test adapted to pure self-preferencing would allow CADE to benefit from a stronger presumption of the anticompetitive effects of such conduct. On the other hand, leveraging and defensive self-preferencing by dominant platforms should be subject to further investigation of the potential anticompetitive (e.g., exclusionary) effects.

We do not have to look far to see evidence of the practical distinction between pure and leveraging or defensive self-preferencing. Indeed, as mentioned before, not all platforms are essential within their relevant markets, even if they create relevant (or even dominant) channels or commerce. While Google stood alone as the sole provider of search engines in most parts of the world, and Apple runs the only app store for iOS mobile operating systems; Amazon and other consumer goods marketplaces, whilst undoubtedly dominant, are not the only channel for online shopping (at least not in Brazil). From the start we can draw a line that distinguishes between the conducts - and thus between the applicable legal tests - that are currently under investigation in Brazil and/or internationally, as indicated in Chart 1.

CONCLUSIONS

The concept of “anticompetitive conduct” and each of its “types” is neither clear nor exhaustive. It is no wonder that the competition legislation has foreseen an open type, precisely to be able to encompass practices that arise from the creativity of an infringing agent. In fact, one must recognize the “impossibility of defining a priori all the hypotheses of infringement to the economic order, especially in face of the variation and the constant evolution of corporate practices”.58 But it is also true that, whenever a concrete conduct is being investigated, an effort is made to fit it into one of the types already known and used in the literature and decisions on antitrust matters. And this occurs for a simple reason: naming and classifying the most common practices helps to identify which are the potential anticompetitive effects and the damage theories that apply to them. That is, even if such practices do not exhaust the universe of violations to the economic order,59 a certain convention around the main types of anticompetitive conducts observed in practice is beneficial and can also be useful for a systematization of the analyses undertaken by the antitrust agency.

Before CADE, there are already indications by private agents and even decisions that use the expression “self-preferencing” to designate conducts in so-called “traditional” markets (that is, outside the context of digital markets and platforms) that are expressly typified and established in Brazilian competition law, such as anticompetitive discrimination, refusal to deal, tying, bundling, among others. This highlights the risk of this expression being used as a mere rhetorical device, following the increasing attention it has been receiving.

For this reason, putting different conducts in the same boat, under the guise of one single category of self-preferencing, results in disregarding tests that have already been formulated, developed, and tested by case law. In this article, we have focused on one of the possible practices that today internalize the idea of self-preferencing, what we call the “pure” version of this conduct. It seems to us that the distinctions suggested in the text are central to the correct and consistent formulation and application of appropriate legal tests to analyze the illegality of practices framed as self-preferencing, which, as we have discussed, are diverse and distinct among themselves. Therefore, it is impossible to adopt a single and watertight legal test, but several different ones and, each of them, can be subject to different presumptions, and burden of proof regimes.

Thus, based on the study proposed here, we do not see sufficient reasons for such practices to completely abandon the lessons learned throughout CADE’s extensive practice, just because they take place in platform markets. It is clear that the competitive dynamics of these markets have particularities that should be considered in the competitive analysis, but any derogation of previous foundations and theories should at least start from them and not from the creation of a new conduct so broad that, by trying to be everything, it ends up having its meaning emptied.

BIBLIOGRAPHY

1. ACM, “ACM: Apple changes unfair conditions, allows alternative payment methods in dating apps,” available at: https:wwallows-alternative-payments-methods-dating-apps, accessed 08.09.2022.
2. AGCM, “A528 - Italian Competition Authority: Amazon fined over €1.128 billion for abusing its dominant position”, available at: https://en.agcm.it/en/media/press-releases/2021/12/A528, accessed 08.09.2022.
3. Ahlbon, Christian; Leslie, Will; O’riley, Eoin, Self-Preferencing: Between a Rock and a Hard Place, Competition Policy International Antitrust Chronicle, v. 3, n. 2, 2020.
4. Athayde, Amanda. Direito da concorrência e supermercados: como essas plataformas de dois lados podem trazer riscos aos consumidores? Revista Direito GV, v. 16, n. 1, jan./abr. 2020.
5. Binotto, Anna; Kastrup, Gustavo, Old tools for new problems? - what can be learned from recent decisions in the Google Shopping case, Revista de Direito e Novas Tecnologias, v. 10, 2021.
6. Bostoen, Friso. The General Court’s Google Shopping Judgment Finetuning the Legal Qualifications and Tests for Platform Abuse. In: Journal of European Competition Law & Practice, Volume 13, Issue 2, March 2022, Pages 75–86, 2022.
7. Bougette, Patrice; Budzinski, Oliver; Marty, Frédéric M., Self-Preferencing and Competitive Damages: A Focus on Exploitative Abuses, Forthcoming in the Antitrust Bulletin, GREDEG Working Paper No. 2022-01, available at: https://ssrn.com/abstract=4028770 or http://dx.doi.org/10.2139/ssrn.4028770, accessed 31.08.2022.
8. Bouzoraa, Yasmine, Between Substance and Autonomy: Finding Legal Certainty in Google Shopping, Journal of European Competition Law & Practice, v. 13, n. 2, p. 144 - 153, 2022.
9. CMA, “Investigation into Amazon’s Marketplace,” available at: https://www.gov.uk/cma-cases/investigation-into-amazons-marketplace, accessed 08.09.2022.
10. Colomo, Pablo Ibanez, Self-Preferencing: Yet Another Epithet in Need of Limiting Principles, World Competition, v. 43, n. 4, p. 417, 2020.
11. ___ Pablo Ibanez, Anticompetitive Effects in EU Competition Law, Journal of Corporation Law & Economics, v. 17, n. 2, p. 309 - 363, 2021, p. 315 - 323.
12. Cordeiro, Alexandre. Essential facility doctrine: A dificuldade de enquadrar casos na doutrina diante da assimetria de informação. In: Jota, 2017.
13. Cotter, Thomas F. Essential Facilities Doctrine. University of Minnesota Law School Legal Studies Research Paper Series Research Paper No. 08-18, 2008.
14. Coutinho, Diogo R.; Kira, Beatriz, Adjusting the lens: new theories of harm for digital platforms, Revista de Defesa da Concorrência, v. 9, n. 1, 2021.
15. Crémer, Jacques; Montjoye, Yves-Alexandre de; Schweitzer, Heike, Competition Policy for the Digital Era: Final Report.
16. Da Silveira, Paulo Burnier; Fernandes, Victor Oliveira, Google Shopping in Brazil: Highlights of CADE’s Decision and Takeaways for Digital Economy Issues, 2019.
17. Deutscher, Elias, Google Shopping and the Quest for a Legal Test for Self-Preferencing Under Article 102 TFEU, European Papers, v. 6, n. 3, p. 1345 - 1361, 2021.
18. Eben, Magali, Fining Google: a missed opportunity for legal certainty? European Competition Journal, v. 14, n. 1, pp. 129 - 151, 2018.
19. Epic Games, “Australia Free Fortnite,” available at: https://www.epicgames.com/site/en-US/news/epic-games-files-legal-proceedings-against-google-in-australia
20. Fernandes, Victor Oliveira. Direito da Concorrência das Plataformas Digitais: entre abuso de poder econômico e inovação. Revista dos Tribunais, 2022.
21. Frazão, Ana de Oliveira. Direito da Concorrência: Pressupostos e Perspectivas. São Paulo: Saraiva, 2017.
22. Gonçalves, Priscila Brolio, A obrigatoriedade de contratar como sanção fundada no direito concorrencial brasileiro, PhD, Universidade de São Paulo, São Paulo, 2008.
23. Heinz, Silke, Self-Preferencing - Some Observations on the Push for Legislation at the National Level in Germany, Competition Policy International Antitrust Chronicle, v. 3, n. 2, 2020.
24. Hoffman, Bruce. Shinn, Garrett, Self-Preferencing and Antitrust: Harmful Solutions for an Improbable Problem, available at: https:wwhoffman--final-pdf.pdf, accessed 07.09.2022.
25. Brazilian Institute for Studies on Competition, Consumption, and International Trade - IBRAC. Unilateral Conduct Guide, 2021, available at: https://ibrac.org.br/UPLOADS/Eventos/497/IBRAC_-_Guia_de_Condutas_Unilaterais.pdf.
26. G7. Compendium of approaches to improving competition in digital markets, 2021, available at: https://www.jftc.go.jp/houdou/pressrelease/2021/nov/Compendium.pdf, accessed 08.09.2022.
27. Lee, Sangyun, Main Developments in Competition Law and Policy 2021 - Korea, Kluwer Competition Law Blog, available at: http://competitionlawblog.kluwercompetitionlaw.com/2022/02/14/main-developments-in-competition-law-and-policy-2021-korea/, accessed 03.09.2022.
28. Lombardi, Claudio, The Italian Competition Authority’s Decision in the Amazon Logistics Case: Self-preferencing and Beyond, available at: https:wwin-the-amazon-logistics-case-self-preferencing-and-beyond/.
29. Mason, Max. “Apple and Google hit with class actions by Australian app users,” available at: https://www.afr.com/technology/apple-and-google-hit-with-class-actions-by-australian-app-users-20220629-p5axpb, accessed 08.09.2022.
30. Padilla, Jorge; Perkins, Joe; Piccolo, Salvatore, Self-Preferencing in Markets with Vertically-Integrated Gatekeeper Platforms, Centre for Studies in Economics and Finance, v. Working Paper No. 582, 2020.
31. Pargendler, Mariana, The Corporate Governance Obsession, Journal of Corporation Law, v. 42, 2016.
32. Petit, Nicolas. Theories of Self-Preferencing under Article 102 TFEU: A Reply to Bo Vesterdorf, Competition Law & Policy Debate, 1, 2015, available at: https://ssrn.com/abstract=2592253, accessed 05.09.2022.
33. Picker, Randy, Structural Separation and Self-Preferencing: What are the Right Lessons of History? Pro Market, 2020.
34. Salinger, Michael A., Self-Preferencing, The GAI Report on the Digital Economy, 2020.
35. Salomão Filho, Calixto. Direito Concorrencial. São Paulo: Malheiros, 2013.
36. Vesterdorf, Bo. Theories of Self-Preferencing and Duty do Deal - two sides of the same coin?, Competition Law & Policy Debate, Volume 1 (1), 2015, p. 4-9, available at: https://ssrn.com/abstract=2561355, accessed 08.09.2022.

---