Quantum Antitrust – A Unified Exclusionary Abuse Theory

Abstract

Now that the judgments in Servizio Elettrico Nazionale and Unilever have made sense of previous case-law, a unified analytical framework can be predicated for all exclusionary abuses. It is made up of two limbs (artificiality/conduct deviating from competition on the merits and potential exclusionary effect/capability of foreclosing). The two limbs are two sides of the same coin, which consists of ascertaining whether the plausible rationale (in the sense of nature and economy) behind a dominant company’s conduct is to derive an advantage that equally efficient competitors cannot either derive by doing the same (first limb–artificiality) or offset by other means (second limb–potential effect). This exclusionary rationale, which is a cognitive state in the dominant company’s mind rather than an ontological reality, is objectivised by holistically judging the dominant company’s conduct against the backdrop of the relative efficiency of its competitors, which is the link between both limbs: if the inherent features of the company’s conduct and all the relevant circumstances surrounding it allow equally efficient competitors to either derive a similar advantage or to defeat its exclusionary effect by some other means, the conduct does not make anticompetitive sense (exclusion is not its plausible rationale). Owing to the very Darwinian nature of competition law, relative efficiency is the common “quantum” and, therefore, a useful yardstick for abuse from the historical, legal, teleological and practical perspectives, as well as from the points of view of causality and reality. Finally, plausibility would be the single standard of proof that could be rebutted by disproving either of the two limbs (in an Intel II-like fashion), either by providing an alternative non-exclusionary explanation that breaks the casual link or by putting forward an objective justification. All other tests and standards merely reflect the extent to which either the artificiality or the potential effects of the conduct can be presumed based on economic judgement or experience (as happened in merger control following CK Telecoms)–thus reconciling per se rules, consisting of an Art. 101-like cursory analysis, with a more economic approach.

1 Introductory Excursus

It does not take an Albert Einstein to realise that everything is of the same nature – one merely needs to take a magnifying lens and zoom in to make that discovery. Quantum mechanics is based on that premiss in that it aims to reduce all matter and energy to wave-particle entities transmitted in small packages (quanta). In this regard, Stephen Hawking’s leading candidate for a theory of everything/unified field theory (the M theory) aims to reduce all particles in the universe(s) to vibrations of strings of energy. It also seems fairly obvious that there are no such things as discrete variables but that everything is a continuum that we, inspired by Procrustes, chop or stretch into artificial categories for the sake of our extremely limited, and thus taxonomic, human understanding. Is this not at the heart of the foundational mathematical set theory, whose continuum hypothesis (whereby the infinite set of real numbers is larger than the infinite set of natural numbers) may even challenge our understanding of space-time in quantum mechanics as a continuum itself?¹ Applying this metaphoric “quantum antitrust” approach (if I may exercise some literary licence) to abuse of dominance cannot but lead to a single analytical framework based on a notion of relative efficiency for all exclusionary abuses, which (to continue my humble tribute to Hawking) I would call “unified exclusionary abuse theory”.

I would ask now the reader to forget all they know about competition law for one second except for one thing: its very essence, which is to prevent artificial advantage from undermining the Darwinian survival-of-the-fittest competitive process (you can also, more fancily, call this market failure or competition not based on the merits, depending on whether you are an economist or a lawyer). Against this background, the words “exclusion” or “foreclosure” refer to one dominant company’s use of artificial means to drive out of the market other companies that otherwise, in Darwinian terms, deserve to stay and prosper just as much as the former company, because they are equally efficient. Therefore, a certain measurement of relative efficiency comes across naturally as the quantum (or the “field” in M-theory terms) of exclusionary abuses. Please note at this point that this paper does not address the philosophical problems involved in establishing the nature and relevant magnitude of that measurement of relative efficiency. These problems are in their own right an obstacle to fully defining the contours of the exclusionary abuse framework, and may be bound to remain helpfully unclear in order to allow for some flexibility in the economic assessment of effects. One may think most notably of debates about the threshold of anticompetitive effects or the place for economic formalism in the more economic approach.²

The point of this paper is merely to explain why the key to interpreting the seeming variety of legal tests and standards for exclusionary abuses boils down to simply ascertain whether the plausible rationale behind a dominant company’s conduct is to derive an advantage that equally efficient competitors cannot either derive by imitating or offset by other means. In particular, I claim that this humble Occam’s razor or pretentious unified theory that I have nicknamed “efficiency-based criterion” (i) is inferred from recent case-law; (ii) is compatible with, and explains, older, apparently erratic, rulings; and (iii) has significant legal and practical advantages, which may provide added value, inter alia, when tackling new-look challenges to exclusionary abuses. In this latter respect, it could help protect the Darwinian essence of competition rules and reserve antitrust enforcement policymaking to the Digital Market Act’s³ regulatory offspring.

Such reasoning, one might think, is either tautological or superfluous. However, as theories of harm become overcomplicated, I believe that it is worth dusting off the original empiricist method, which consists of drawing on what comes clearly and distinctly to mind (to use Locke’s main criteria for ascertaining the reality of ideas). Now that we seem lost in a maze of jurisprudential legal tests and standards, and constrained by the tension between substance and form, going back to basics could allow us to approach with fresh unprejudiced eyes the new complexity challenging the traditional competition rulebook (think of all the fuss around digital markets for a prominent example). It took the Court of Justice of the European Union (CJEU) 60 years to reach this conclusion, whether intentionally or simply because it was the only possible conclusion. This surprises me, as Luxembourg judges have never had much difficulty asserting the unity of legal tests in merger control⁴ (although they have admittedly had to reiterate it recently, as the complexity of competition law assessment is becoming overwhelming).⁵

All in all, is shoehorning past case-law into new legal theories not an artificial exercise altogether? It is certainly an artificial, but not a barren, exercise and, in fact, it is what the CJEU does all the time when it rewrites precedents in new rulings in its own particular, relaxed, version of the principle of stare decisis.⁶ Hence, I propose the efficiency-based criterion as a compass to guide the CJEU in the rewriting of exclusionary abuse case-law that it has recently embarked on in an effort to piece together legal tests and standards, just as it did with merger control. And the momentum now seems to be that case-law is mature enough to allow for meaningful analysis, though not yet sufficiently consolidated to make reflection redundant or any attempt to keep it on track hopeless.

2 Proposition

Indeed, one can think of case-law as either the revealed truth that judicial oracles drip-feed us and prophetic commentators then make sense of, or as a constructive dialectic that strings together self-validating ex post facta and genuine criticism. Regardless, the CJEU’s jurisprudence on exclusionary abuses of dominance appears in recent years to have reached a point where we can already glimpse, or construct, a guiding thread that connects all seemingly separate legal tests and standards into one unified analytical framework. Its vocation is to be applicable across the whole spectrum, from essential facility holders that satisfy all the “exceptional circumstances” in the Magill-Microsoft test,⁷ through the much-hyped leveraging on digital platforms, to price and non-price abuses apocryphally labelled “per se” (or even “by object”). My assertion, that this unified analytical framework is based on the relative efficiency of competitors⁸ vis-à-vis the dominant company, is based on the following considerations.

The judgment in SEN,⁹ albeit reluctantly, in a convoluted manner and taking all the usual precautions not to break the link with the specific case, makes sufficiently clear that both limbs of the general exclusionary abuse test in Generics¹⁰ (i.e. (i) conduct that falls outside the scope of competition on the merits and (ii) conduct capable of producing exclusionary effects) can be distilled into a single efficiency-based criterion. In particular, the CJEU holds that

[t]he relevance of the material or rational impossibility for a hypothetical competitor, which is as efficient but not in a dominant position, to imitate the practice in question, in order to determine whether that practice is based on means that come within the scope of competition on the merits, is clear from the case-law on practices both related and unrelated to prices (para. 79).

This assertion by the CJEU was made in the context of ascertaining whether certain conduct was “capable of producing an exclusionary effect in respect of competitors that were at least as efficient as the undertaking in a dominant position”, which is deemed to be sufficient to characterise the practice as abusive (para. 71).¹¹

In fact, the artificiality of the conduct (the first limb of the analytical framework) is embedded in, and examined as part of, the Court’s analysis of the potential exclusionary effect (the second limb) because, according to settled case-law, if the potential exclusionary effect arose from a conduct that was not artificial (but constituted competition on the merits) there would be no abuse as it would be “Darwin-proof”. Advocate General Rantos expresses this as follows:

following the Commission’s revision in 2005 of competition policy with regarded [sic] to abuses, the Court has acknowledged that Article 102 TFEU does not prohibit an undertaking from acquiring, on its own merits, the dominant position in a market and that it does not, therefore, seek to ensure that competitors less efficient than the undertaking with the dominant position remain on the market. That approach was confirmed in the judgment in Intel, […]

from which he infers that “[t]he perception that Article 102 TFEU is not aimed primarily at protecting (less efficient) competitors, which the Commission also shares, seems now to be an accepted tenet of EU competition law.”¹²A contrario, he concludes that “exclusionary conduct of a dominant undertaking which can be replicated by equally efficient competitors does not represent, in principle, conduct that may lead to anticompetitive foreclosure and therefore comes within the scope of competition on the merits”.¹³

Although elevating replicability to the status of universal benchmark may be going too far, there can be little doubt about the universality of an analytical framework for exclusionary abuses made up of two elements that are indissociably intertwined rather than merely cumulative: (i) conduct that deviates from competition on the merits and (ii) a potential exclusionary effect. As a matter of fact, the CJEU had no reservations about stating that “a practice […] may, when implemented by an undertaking in a dominant position, be characterised as ‘abusive’ […] if it is capable of producing an exclusionary effect and if it is based on the use of means other than those which come within the scope of competition on the merits”.¹⁴ More eloquently, it ruled that,

in order to establish that an exclusionary practice is abusive, a competition authority must show that, first, that practice was capable, when implemented, of producing such an exclusionary effect, in that it was capable of making it more difficult for competitors to enter or remain on the market in question and, by so doing, that that practice was capable of having an impact on the market structure; and, second, that practice relied on the use of means other than those which come within the scope of competition on the merits.¹⁵

Particularly illustrative of the universal vocation of the dual-limbed exclusionary abuse framework is the scrapping of the concept of atypical abuse. Some claim that the general abuse test should be limited to atypical abuse.This was ones of the questions referred to the CJEU for a preliminary ruling in SEN, but was dismissed by the Advocate General as overly formalistic¹⁶ and simply ignored by the CJEU.

I propose the following interpretation of SEN: the CJEU is reflecting on the fact that the purpose of appraising potential exclusionary effects is to ascertain whether such effects arise from artificial conduct (that deviates from competition on the merits), because otherwise the exclusionary effects would be perfectly lawful. When following this reasoning, the CJEU stumbles upon the realisation that both aspects (artificiality and potential effects) ultimately attempt to answer the same question: whether or not the whole point of such conduct is for a dominant company to derive an advantage that equally efficient competitors, merely because they are not dominant, cannot either derive themselves by imitating (artificiality) or offset or replicate by other means (potential effects). In other words, the two limbs of the unified analytical framework (artificiality and potential effects) would be two sides of the same coin¹⁷ (or two phases, with the latter being a sanity check on the former). Both sides converge in a holistic assessment of whether the rationale of the dominant company’s conduct is to derive an advantage that (i) deviates from competition on the merits – owing to the fact that equally efficient competitors could never have gained an edge of the same magnitude had they engaged in the same conduct because they are not dominant (first limb); and (ii) is of such magnitude that those equally efficient competitors would not be able to replicate or offset it by other means and would then eventually (potentially) be excluded (second limb). One could even think of the first limb as qualitative and the second limb as quantitative.

Please note that by “rationale” I do not mean the “aim”, but rather the raison d’être, the “nature or general scheme” (la nature ou l’économie générale)¹⁸ or the essence, that is to say the Aristotelian final cause – a metaphysical end that is inherent in the conduct and transcends mere exclusionary intent, which is obviously not decisive for abuse.¹⁹ Therefore, by rationale, I mean not an ontological reality, but an epistemological or cognitive state in the dominant company’s mind, objectivised by holistically judging the dominant company’s conduct against the backdrop of the relative efficiency of its competitors. Indeed, in the references to the “abusive nature” of a practice in SEN²⁰ and Unilever²¹ (as well as in Google Android when referring to an overall assessment that includes the analysis of foreclosure effects),²² or to the “intrinsic capacity” of certain conduct to produce effects in Intel II²³ and Google Android,²⁴ I see a nod to this combination of inherent artificiality and exclusionary ability that is required for conduct to pass the general (rather than atypical), even universal, exclusionary abuse test. More specific tests for typical abuses are merely concretisations or particular ways of ascertaining a practice’s exclusionary rationale or nature, adjusted to the specificities of the practice at stake (see Section 6).

3 Vindication

It is true that SEN can be seen as not intended to enshrine an overarching principle or as having limited potential beyond the specific and particularly extreme case at stake (i.e. a former legal monopoly engaging in a quite clearly profiteering strategy). Nevertheless, I read it as allowing for the formalisation of a unified exclusionary abuse framework (rather than a test) that is based on relative efficiency and entails a number of advantages. It is not only from a systematic standpoint that the efficiency-based criterion appears to be the best means of ensuring the internal consistency of the dual-limbed general exclusionary abuse test, which it does by finding the common quantum or greatest common divisor of both prongs (i.e. the conduct is artificial because its rationale or nature is to derive an advantage that equally efficient competitors cannot derive merely because of their lack of dominance, and it is capable of producing exclusionary effects because those equally efficient competitors could be foreclosed as a result). It also provides a useful yardstick from the historical, legal, teleological and practical perspectives, as well as from the points of view of causality and reality, as explained below.

Preliminarily, care must be taken not to confuse the efficiency-based criterion with the “as-efficient competitor (AEC) test”. The latter is just one of the various tools allowing competition authorities on a methodological level to infer from “all the relevant circumstances” whether the conduct is capable of producing a foreclosure effect (second limb) or even whether it deviates from competition on the merits (first limb).²⁵ It is to this purely operational character that the CJEU is referring when it holds that

the significance generally given to that [AEC] test, when it can be carried out, shows that the inability of a hypothetical as-efficient competitor to replicate the conduct of the undertaking in a dominant position constitutes, in respect of exclusionary practices, one of the criteria which make it possible to determine whether that conduct must be regarded as being based on the use of means which come within the scope of normal competition.²⁶

Far from limiting the general applicability of the efficiency-based criterion to certain categories of abuse, and bearing in mind again that the CJEU is referring only to the AEC test here, the CJEU’s finding proves that both limbs of the framework for analysing general exclusionary abuse converge in the efficiency-based criterion. Indeed, it demonstrates that the AEC method, when applicable, can serve to prove both artificiality and potential exclusionary effect, because both are ultimately one and the same.

From the historical perspective, the universality of the single dual-limbed abuse framework has been implicit in the “objective” notion of abuse (and necessary for its objectivisation) ever since Hoffman-La Roche described it as

an objective concept relating to the behaviour of an undertaking in a dominant position which […], through recourse to methods different from those which condition normal competition […], has the effect of hindering the maintenance of the degree of competition still existing in the market or the growth of that competition.²⁷

Even twelve years before SEN, the efficiency-based criterion (incepted by AKZO²⁸ and reaffirmed by Post Danmark I²⁹) was already being used as a bridge between the two limbs of the single abuse framework in the margin squeeze cases – in particular, Deutsche Telekom (i.e. [Art. 102 of the Treaty on the Functioning of the European Union (TFEU)] “prohibits a dominant undertaking from, inter alia, adopting pricing practices which have an exclusionary effect on its equally efficient actual or potential competitors […], thereby strengthening its dominant position by using methods other than those which come within the scope of competition on the merits”)³⁰ and then TeliaSonera (i.e. “that would mean that competitors who might be excluded by the application of the pricing practice in question could not be considered to be less efficient than the dominant undertaking and, consequently, that the risk of their exclusion was due to distorted competition” and “[s]uch competition would not be based solely on the respective merits of the undertakings concerned).”³¹ This trend smacks of the pendular nature of case-law on the free movement of goods, illustrated by the fact that it took 35 years for the CJEU to get back, in the Trailers³² ruling, to Dassonville’s³³ market-access approach. Therefore, the first advantage of the efficiency-based criterion is that is rescues all previous case-law (see Sections 5 and 6).

Secondly, from a positivistic point of view, the efficiency-based criterion flows precisely from the legal core of Art. 102 TFEU: the “special responsibility.” Ultimately, the special responsibility/prohibition on abuse boils down to attaching to the factual finding of dominance the legal consequence of an ex-ante qualified legal status. That status prevents the dominant company from doing anything that may further weaken competition in a market where competition is already reduced (albeit lawfully) owing to the mere existence of the dominant position.³⁴ “Anything?” Of course not. Naturally, the (further) weakening of competition that results from the dominant company’s foreclosing less efficient competitors would be a perfectly acceptable (and desirable) outcome of the competitive process that competition rules are meant to protect.³⁵ Such outcome would be natural rather than artificial, so preventing it from naturally arising would be the artificial thing for enforcers to do.

Rather, there needs to be something intrinsically or “objectively” unlawful (as the Spanish Supreme Court puts it³⁶) about the dominant company’s conduct for the resulting weakening of competition to be an abuse. This is what case-law means by abuse being “an objective concept”, characterised by the dominant company’s weakening competition through “methods different from those which condition normal competition”³⁷ or “methods other than those that are part of competition on the merits”.³⁸ Consequently, since “[c]ompetition on the merits may, by definition, lead to the departure from the market or the marginalisation of competitors that are less efficient,”³⁹ the intrinsic or objective feature of the dominant company’s conduct that makes the resulting competitive advantage artificial (i.e. deviating from competition on the merits or normal competition) cannot but be a function of the dominant company’s relative efficiency vis-à-vis its competitors. This is the reasoning that led the CJEU in SEN to realise that, if ascertaining the artificiality of the conduct was a necessary step in an overall assessment of potentially exclusionary effects (which are not illegal if resulting from competition on the merits), the ultimate question was whether the dominant company’s benefitting from such exclusionary effects was the plausible rationale (or economy or nature) behind some conduct that equally efficient competitors could not derive a comparable advantage from (simply because they were not dominant).

Thirdly, the efficiency-based criterion could also solve the legal awkwardness in the jurisprudential construction of the causal link. According to case-law, the causal link must be shown between dominance and effects only,⁴⁰ rather than between dominance and conduct – although the latter would be a legally sounder and more objective device for attaching legal consequences to the dominant company’s conduct. In fact, we could reformulate “between dominance and effects” for “between dominance and dual-limbed artificiality” (which has foreclosure as both interdependent purpose and potential effect). In that case, the transition from legally qualified status of dominance to illegal consequence of abuse would not depend on something contingent (effects) but on something substantial (the inherent nature of the conduct, including its effects and “all other relevant circumstances”). Only in this way would the causality requirement, by linking dominance and conduct, meet the standard imposed by legal logic.

Put differently, there would be a causal link from the efficiency-based perspective only if dominance allowed for a certain conduct that, if engaged in by equally efficient competitors (which is sometimes possible)⁴¹ still could not give those competitors an advantage that was (i) qualitatively comparable to that derived by the dominant company (artificiality limb) or (ii) quantitatively sufficient (even if qualitatively different) to allow them to compete viably with the dominant company (exclusionary effect limb), merely because they were not dominant. Where we should set the threshold of foreclosure or whether there is a threshold at all is, as stated above, another controversial matter that I do not address in this paper.⁴² Indeed, the efficiency-based criterion can be formulated thus in causality terms: if competitors are less efficient, their exclusion cannot be causally linked to the dominant company’s conduct but only to their own lower efficiency.⁴³

By the same token, the efficiency-based criterion avoids making the concept of competition on the merits dependent on a subjective judgement call of (extra-legal) reprehensibility, which would reverse the burden of proof. An undefined legal concept like competition on the merits could be seen as rather being on the morality side of the Dworkinian house, which would give enforcers ample leeway to double-think (from an extra-legal or moral angle) whether the dominant company’s (potentially exclusionary) conduct is either normal or artificial. This seems to be the approach validated by the General Court in Google Shopping, where judges had to double-think Google’s business model in order to opine that the use of criteria not strictly based on relevance was something abnormal for a search engine to do and thus outside competition on the merits.⁴⁴ This shifted the burden onto Google to justify its conduct.⁴⁵

In this regard, it is my opinion that the efficiency-based criterion may even be the only chance of saving competition on the merits from being scrapped in favour of other ways of objectivising the artificiality of conduct. One alternative may be the heuristic commonness test, which reduces subjectivity by placing the emphasis on the market context (working out what is normal for a number of companies to do) as a yardstick for the design of the practice,⁴⁶ instead of focusing on the economy of the practice (the context being just part of “all the relevant circumstances”⁴⁷ in an Ortegian logic). The danger of commonness, which seems to have been endorsed to some extent by the CJEU in Google Shopping,⁴⁸ is that competition authorities may be tempted to abuse it in digital markets where they are desperately in need of deviations from competition on the merits for imposing level-playing-field solutions as a matter of policy. The clash between the efficiency-based and the commonness criteria, and whether the General Court’s allegiance to the latter is still compatible with Unilever, was very present in the questions posed at the hearing on the Google Shopping appeal.⁴⁹

Another alternative criterion, the American profit-sacrifice test, is, in my view, fundamentally flawed because it depends too much on a subjective stance by the enforcer regarding the hypothetical market outcome.⁵⁰ Finally, consumer welfare would not even be a competing benchmark for artificiality, because, even in structuralist European Union competition law,⁵¹ it has never been an actual yardstick but rather just an aspiration for the competitive structure of markets. Consumer welfare plays a direct role only if there is direct harm to consumers (exploitative abuses),⁵² or is at most a criterion for efficiencies to be considered at the separate stage of justification (as attempts to introduce a rule of reason have been fragmentary and sibylline).⁵³

A fourth upside to the efficiency-based criterion is systematic: it brings homogeneity across the continuum of the exclusionary abuse framework, as it is consistent with the essential facilities doctrine. Indeed, the economic and objective essence of the efficiency-based criterion is the same as that of the indispensability requirement under the essential facilities doctrine. For that matter, the efficiency-based criterion provides a spectrum in which the various tests and standards of proof can be pinned down – from traditional per se rules, now identified with the “abusive nature” of some practices, at one end, to essential facilities at the other (see Sections 5 and 6 below). Arguably, as explained below, by reconciling per se rules with the more economic approach, the efficiency-based criterion ensures homogeneity with the analysis of agreements under Art. 101 TFEU: the establishment of an abuse on the basis of a cursory analysis of the anticompetitive nature of the practice, taking into account both the objective rationale of the conduct and the surrounding context (“all the relevant circumstances”), would, subject to rebuttal, be equivalent to an analysis of the degree of harm in the light of the content, purpose and legal and economic context of the agreement concerned.⁵⁴

The efficiency-based criterion could also make it easier to find a single and continuous infringement of Art. 102 TFEU, which may otherwise be more challenging than under Art. 101 TFEU.⁵⁵ All in all, founding the common exclusionary rationale on whether it is plausible that the various instances of a certain conduct contribute to an advantage that equally efficient competitors cannot derive and that is capable of foreclosing them requires that these instances “complement each other inasmuch as each of them is intended to deal with one or more consequences of the normal pattern of competition and, by interacting, contribute to the realisation of the objectives intended within the framework of that overall plan”.⁵⁶

Therefore, fifthly, the efficiency-based criterion is suitable also from a teleological point of view, as it helps strike the balance, underpinning the general scheme of Art. 102 TFEU, between the legal interests in fostering competition and those in limiting intervention to the minimum necessary to prevent market failure. Such balance is inherent in the concept of competition on the merits because, depending on whether or not conduct deviates from competition on the merits, either the general exclusionary abuse framework (requiring only proof of a potential exclusionary effect) or the much higher bar set by the essential facilities doctrine will apply. Not surprisingly, competition on the merits is of pivotal value as a hinge between the just (non-artificial) refusal to supply, subject to the essential facilities doctrine, and the unjust (artificial) conduct that deviates from competition on the merits. In margin squeeze cases, in which that distinction emerged, indispensability was no longer a requirement,⁵⁷ precisely because the potential exclusionary effect necessarily affected equally efficient competitors (as they would be undercut regardless of how efficient they were). This fact alone made the practice deviate from competition on the merits.⁵⁸ Therefore, the unified exclusionary abuse framework was born already with its two limbs (deviation from competition on the merits and potential exclusionary effect) bound together in the efficiency-based criterion.⁵⁹ Then, it matured (in a non-pricing version of margin squeeze) in Slovak Telekom/Deutsche Telekom II,⁶⁰ and was finally reflected in the Commission’s amendment to the 2009 Guidance Paper.⁶¹

That function of the efficiency-based criterion is extremely important because it would prevent enforcers from using their own discretion to evade the application of the essential facilities doctrine by subjectively reproaching the dominant company for its business decision not to share its advantage in the exact form in which it benefits from it itself, without proving that equally efficient competitors cannot replicate or otherwise defeat that advantage. As indispensability is very hard to prove in digital environments (“competition is one click away”⁶²), conferring such discretion on enforcers would provide them with a powerful tool to consider abusive any business model by digital platforms entailing a conflict of interest between them and their competing business users and not ensuring absolute neutrality (the “level playing field”⁶³). A solution could be for the Commission’s forthcoming exclusionary abuse guidelines to state, as I suggested in the Spanish competition association’s contribution to the consultation,⁶⁴ that the efficiency-based criterion is the connecting thread that ensures consistency between the two limbs of the single exclusionary abuse framework. Thus, the assessment of deviation from competition on the merits would become objective and would limit the discretion of enforcers to qualify business models as “abnormal” in order to bypass the indispensability requirement under the essential facilities doctrine.

From that practical point of view, the efficiency-based criterion features a sixth advantage. It is not only key to reducing enforcers’ discretion, and to maintaining, in the abstract, the balance of interests between the dominant company’s freedom to conduct business and economic incentives to innovate, on the one hand, and public intervention to make good market failures, on the other. The efficiency-based criterion also facilitates the effective judicial review of the administrative finding of a deviation from competition on the merits. It is clear that, “in areas giving rise to complex economic assessments, the Commission has a margin of discretion with regard to economic matters”, which does not relieve the CJEU of its obligation to “establish, among other things, whether the evidence relied on is factually accurate, reliable and consistent” and “whether that evidence contains all the information which must be taken into account in order to assess a complex situation and whether it is capable of substantiating the conclusions drawn from it”.⁶⁵

In this regard, the efficiency-based criterion would be equivalent to the Kantian concepts that need to be filled up with experience or intuition to become quantificational objects – concepts are empty shells without experience, and experience is a “swarm of appearances” without concepts.⁶⁶ Similarly, the efficiency-based criterion only becomes meaningful through case-specific economic analyses of effects to determine whether an advantage that competitors cannot derive or otherwise defeat is inherent in a particular conduct. This is the complex economic assessment (more or less formalistic but always factual) in which the competition authorities’ discretion should play out, taking into consideration all the circumstances and evidence, including those provided by the defendant (as a matter of procedure or substance, depending on how we read Intel II), and using whatever tools and indicia prove useful (AEC test, exclusionary intent, market position, market coverage, etc.). Then, it is for the reviewing court to form a firm conviction⁶⁷ that the plausible story told by such circumstances and evidence is that the dominant company derives an advantage that equally efficient competitors cannot derive or offset by other means and that therefore could lead to their departure from the market.

Coming back to the Google Shopping example, let us imagine that the Commission focused on whether the essence of Google’s practices was to derive from its search engine (and then potentially materialise) a decisive advantage for its comparison service over equally efficient competitors, instead of merely guessing whether those practices were the normal business strategy for a search engine to adopt. In that case, the General Court for its part could have limited its review to verifying that the evidence provided gave it a firm conviction that deriving such an advantage was the objective “nature” of the conduct, instead of getting dragged into making a subjective (almost moral) value judgement about Google’s business model.⁶⁸ Therefore, by providing the reviewing court with a single legal framework based on the objective features of the conduct in question and the surrounding circumstances, the efficiency-based criterion allows that legal plane of review (including any legal test, standard of review and standard of proof) to be delineated from the factual gathering of circumstances and evidence that must be left to the competition authority’s discretion for complex economic assessments.⁶⁹

The efficiency-based criterion not only allows for the judicial review of administrative decisions but also helps enforcement comply with the principle of legal certainty by providing dominant companies with an objective benchmark for assessing the “abusive nature” of their conduct. It is not a coincidence that the case-law on legal certainty, whereby, since the list of abuses is open-ended, a dominant company must be able to “assess the lawfulness of its own conduct”, also originated in the context of margin squeeze cases.⁷⁰

Seventhly, I claim that the inseparable (rather than merely cumulative) nature of the two limbs of the general analytical framework, which hinges on the efficiency-based criterion, is also required if the legal plane is to mirror the reality of the conduct. If either of the two limbs fails (i.e. either the conduct does not allow the dominant company to derive an advantage that equally efficient competitors cannot obtain by doing the same, or equally efficient competitors are still able to offset the advantage by some other means), it no longer makes anticompetitive sense for the dominant company to engage in that conduct. Although this reasoning assumes certain anticompetitive intelligence (rather than will) on the part of the dominant company, it is this cognitive state, in which the company can reasonably foresee the unlawfulness of its conduct and still takes the risk of embarking on it, that underpins the aforementioned jurisprudential construction of legal certainty and the whole (quasi-strict) liability attribution regime in competition law.⁷¹

4 Refutation

Notwithstanding the above, some authors have seen recent case-law as struggling with the cumulative or alternative nature of the means-effect binomial (a different way of regarding the two limbs of the general exclusionary abuse framework – methods deviating from competition on the merits and potential foreclosure effect).⁷² This brings to my mind the debate on monophysitism – I hope this paper does not come across as heretical in the competition law orthodoxy when it contends that both limbs of the unified test are of the same nature. In reality, problems arise only when means and effects are artificially stripped apart. By contrast, we can make sense of seeming contradictions in case-law about whether the two limbs are alternative or cumulative if we consider that it all depends on the side of the coin on which the CJEU placed the emphasis in casu. In particular, there may be a temptation to focus on effects where the artificiality of the conduct is more controversial – see, e.g. Google Shopping, where the weakness of the reasoning as to why conduct deviated from competition on the merits lay precisely in its unnatural split from the foreclosure effect limb and the disconnect from the efficiency-based criterion.

On the other hand, the CJEU logically focuses on means departing from competition on the merits, and sidelines the assessment of effects in conventional per se abuses (if such a thing were to exist) with the archetypal example of exclusivity obligations and payments. In those cases, the CJEU has stuck to the compromise solution of Intel II. The analysis of foreclosure effects is then factored in as a procedural rule, with competition authorities being obliged to examine the defendant’s rebuttal of such otherwise presumed effects. However, even this seems to be turning into a formalistic substantive obligation to assess specific “relevant circumstances”, as demonstrated by Intel renvoi⁷³ and Qualcomm⁷⁴). Indeed, even the traditional per se approach to exclusivity prior to Intel II has been reconciled with SEN’s general dual-limbed, efficiency-based, exclusionary abuse framework by the CJEU in Unilever⁷⁵ (as eloquently shown by the oxymoronic joint reference in para. 40 thereof to SEN and Tomra⁷⁶). Some authors have seen in Unilever a certain ambivalence as to whether the two limbs of the general exclusionary abuse framework are cumulative or alternative (see above), but I see in that ruling clear confirmation that, rather than being only cumulative, the two constitute one single element revolving around the efficiency-based criterion.

It is precisely the monophysis of means and effects that may make some paragraphs of Unilever, if taken in isolation, wrongly point to the two limbs being alternative rather than cumulative (meaning that only either a deviation from competition on the merits or a potential exclusionary effect would need to be proven). However, the way in which the reasoning in Unilever is structured leaves no doubt (if Generics⁷⁷ or even Deutsche Telekom⁷⁸ or TeliaSonera⁷⁹ left any) as to the need to prove both the artificial means of the conduct and its ability to produce foreclosure effects, which are indissociably blended together in the “abusive nature” of the conduct. The fact that the referring court asked whether potential anticompetitive effects needed to be proven and the CJEU answered in the affirmative made this even clearer. In particular, the CJEU asserted the need

to establish, in the light of all the relevant circumstances and in view of, where applicable, the economic analyses produced by the undertaking in a dominant position as regards the inability of the conduct at issue to exclude competitors that are as efficient as the dominant undertaking from the market, that [the conduct in question is] capable of restricting competition” (para. 62).

This is no wonder, because any other conclusion would have run counter to the combined reading of Intel II and SEN.

How can we reconcile this conclusion with the finding that “the use by a [sic] undertaking in a dominant position of resources other than those governing competition on the merits may be sufficient, in certain circumstances, to establish the existence of such an abuse” (para. 57)? Quite easily: just by accepting that establishing a deviation from competition on the merits inherently includes establishing a potential foreclosure effect, as part of those “certain circumstances” or “all the relevant circumstances”, because it requires proving that the conduct not only generates an advantage that equally efficient competitors cannot generate (thus deviating from competition on the merits) but also that it is capable of driving such equally efficient competitors out of the market (otherwise the conduct would not make anticompetitive sense, and it would not be legally reprehensible since it could not restrict competition). Of course, as explained in Section 6 below, for some abuse tests the anticompetitive object is so evident that the relative efficiency of competitors becomes moot in showing the artificiality of the conduct. However, it is still instrumental in ascertaining the overall foreclosure rationale by allowing for the assessment of potential anticompetitive effects, failing which there would be, epistemologically (on the dominant company’s drawing board) and thus legally, no abuse.

The same conclusion is reached by the finding that an

abuse of a dominant position could be established, inter alia, where the conduct complained of produced exclusionary effects in respect of competitors that were as efficient as the perpetrator of that conduct in terms of cost structure, capacity to innovate, quality, or where that conduct was based on the use of means other than those which come within the scope of ‘normal’ competition, that is to say, competition on the merits (para. 39 of Unilever).

In this paragraph, the disjunctive “or” does not separate means and effects but actual and potential effects. Note that it reads “produced exclusionary effects”, which means “actual” effects, which of course do not need to be proven (para. 41), while potential effects are implicit in the reference to means departing from competition on the merits, because both are part of a whole. That whole, objectifiable by reference to equally efficient competitors, is the anticompetitive rationale (or “abusive nature” in per se lingo) of the conduct, which is to be assessed “in the light of all the relevant factual circumstances” (para. 40). I claim above that the reference to “abusive nature” taken from SEN suggests that the unified exclusionary abuse framework ultimately boils down to determining whether the exclusion of equally efficient competitors is the overarching rationale or economy of the conduct (including both means and aptitude – if one fails, either there is no foreclosure or foreclosure is lawful). The reference to an “intrinsic capacity” to produce effects in Intel II is even more articulate, because it means that, by virtue of a rule of experience, some practices cannot reasonably be assumed not to give rise to at least potential effects (in a similar logic to that lying behind restrictions by object under Art. 101 TFEU or the presumption of harm in cartel cases).⁸⁰

This softer and (from the “more economic” perspective) more palatable version of per se abuses explains the rather cryptic para. 39 in Unilever: sometimes the artificial means are enough, provided that such artificiality is established by comparison with equally effective competitors. This is because, if equally effective competitors cannot derive a comparable advantage, they can be assumed (res ipsa loquitur) to be unable to offset this advantage either and, thus, the capability to foreclose is presumed (is intrinsic), in a cartel-like fashion. Therefore, para. 39 in Unilever provides all the more evidence that blending together both limbs of the exclusionary abuse into an efficiency-based criterion is the solution to die-hard per se rules (see Section 6 below). It also underpins the object-effects dichotomy: if the plausible purpose of the conduct is anticompetitive, a rational dominant company would only engage in it if it expected, on the basis of experience or economic judgement, to benefit from anticompetitive effects. Therefore, at least the capability of the conduct to produce such anticompetitive effects can be (rebuttably) presumed.⁸¹

Effectively, following SEN, the analysis of competition on the merits in Unilever is part of a broader assessment of competitor foreclosure based on an efficiency-based criterion: it must be proven that foreclosure is both capable of occurring (potential effect as opposed to actual effects) and unlawful (because the means through which it occurs depart from competition on the merits). This is confirmed in the assertion that, “although, by reason of their nature, exclusivity clauses give rise to legitimate concerns of competition, their ability to exclude competitors is not automatic” (para. 51, which makes reference to the 2009 Guidance Paper⁸² in a nod to the “more economic” approach). The CJEU then devotes the ensuing paragraphs to discussing the potential of the AEC test (which is not mandatory but not limited to pricing abuses either) and whether Intel II means that demonstration of the potential effect, which is always required, is either a procedural rule, with the burden being somehow reversed onto the defendant to raise the issue, or is a fully fledged substantive requirement that must be examined ex officio.⁸³

Moreover, while acknowledging the usefulness of the efficiency-based criterion as a proxy, it has been asserted that the search for an all-encompassing abuse test is futile⁸⁴ (obviously, before the judgments in SENand Unilever were pronounced; afterwards, the author took a more nuanced view).⁸⁵ The reasons stated revolve around the fact that the main attempts so far, which focus on the absence of an alternative explanation to an exclusionary end, lack explanatory value beyond inherently (obvious) anticompetitive conduct (naked abuses), and are not useful for specific cases but rather remain abstract.⁸⁶ In particular, it has been added that a “no economic sense” test, focused on whether the only plausible purpose is anticompetitive, would only capture practices that blatantly deviate from competition on the merits.⁸⁷ I fully agree with this backdrop, against which the merits of the efficiency-based criterion stand out due to its added value, (i) as an abstract (objective) benchmark, in warding off dangerous subjective appraisals of competition on the merits and, (ii) in the specific case in question, in accommodating (and requiring) effects-based economic analyses to establish whether the conduct is apt (more or less obviously, intrinsically or per se) to give rise to an artificial advantage that equally efficient competitors cannot derive or offset.

In any event, once again, the efficiency-based artificiality proposed here must not be construed as a requirement that an exclusionary end or purpose be the only plausible explanation for the conduct. Such a construction, I concede, would only capture inherently exclusionary practices; and, anyway, under case-law,⁸⁸ intent can only provide an indication of abuse. The proposed efficiency-based artificiality is not to be found in the means through which the conduct is executed either. On the contrary, it refers to the nature, economy or essence of the conduct being such as to give the dominant company a competitive edge over equally efficient competitors that could not replicate or offset it, and would consequently be foreclosed. Then, methodologically, it would work as a hypothesis that such a state of mind of the dominant company, consisting of the expectation that both means and potential effects would lead to successfully foreclosing equally efficient competitors, is the plausible explanation for the conduct. It does not need to be the only explanation, but merely a plausible one, which is to be more or less presumed depending on a rule of experience or an economic judgment. Then, the dominant company can always rebut the explanation by using the procedural, or even substantive, avenues that accommodate the more economic benefit of the doubt in traditional abuse tests, or just by providing an alternative explanation serving as objective justification. This would be the single standard of proof proposed below (Section 5), and the quality of the evidence required would depend on how the unified framework were specifically manifested in the form of a legal test adjusted to a particular theory of harm (Section 6).

At the other end of the spectrum, claiming that the efficiency-based criterion is an overarching objective of competition can (rightly) come in for criticism as an overstatement. Indeed, the proposed efficiency-based criterion does not entail that economic efficiency from the perspective of consumer welfare should be the goal of competition law, especially where European Union competition law is concerned about the market structure, which may also be made up of less efficient competitors.⁸⁹ Hence, less efficient competitors must also be factored into the assessment.⁹⁰ This does not call into question the efficiency-based criterion, because competitive constraints from those less efficient competitors are part of “all the relevant circumstances” that can reduce the capability of the conduct to foreclose equally efficient competitors and, thus, the plausibility of its exclusionary rationale. In short, the efficiency-based criterion is just an operational tool required by the nature of the prohibition on abuse in Art. 102 TFEU, which, after Intel II (and then SEN and Unilever), nobody can doubt is preventing the dominant company from excluding equally fit competitors.⁹¹ Other policy-driven interests, such as the (in)famous level playing field infiltrating Google Shopping, must be inserted into the abuse framework by means of the efficiency-based criterion: by undermining equality of opportunities, Google is making sure that equally efficient competitors cannot replicate its advantage or offset it by other means. Otherwise, those interests can be better tackled by regulation through concepts such as fairness and contestability in the Digital Markets Act that are alien to competition law enforcement in that they (fairly enough) aim to increase the chances of competitors regardless of their (lower) efficiency.

5 Efficiency-Based Criterion and Standard of Proof

In the light of the above, the efficiency-based criterion provides a homogeneous continuum in which the different legal tests and standards of proof are coordinates for a factual analysis of the economic rationale of a dominant company’s conduct, which is in the end what the assessment of abuse boils down to.⁹² The objective of the assessment is to determine whether a dominant company’s conduct is explained by its ability to artificially foreclose equally efficient competitors. Then, the applicable test (whether a per se rule, a significant artificial advantage that competitors cannot offset, a reasonably likely foreclosure effect, indispensability, etc.) and the applicable standard of proof (which probability of actually excluding competitors needs to be proven) depend on how much anticompetitive sense the conduct makes considering the conduct itself and the context (“all the relevant circumstances”). This is shown by the reference in Unilever, when setting the standard for the demonstration of potential effects at “actual capability”,⁹³ to Woodpulp,⁹⁴ the landmark tacit collusion case in which it was held that parallel conduct only proved concertation if concertation was the only plausible explanation for parallel conduct. This is made clearer by the fact that the benefit of the doubt always allows the dominant company to break the causal link by providing an alternative plausible explanation for the conduct⁹⁵ and, more generally, that an objective justification always trumps the abuse.

Therefore, it is all about separating mere hypotheses from plausible explanations, while the different standards of proof and legal tests are just milestones to provide legal certainty and administrability⁹⁶ by chopping continuous reality into manageable discrete variables. As mentioned above, the CJEU seems to have already come to this realisation in merger control, where it has held that the standard of proof is one and the same, as is the legal test, no matter how outlandish the theory of harm. Rather, in merger control, what varies from case to case is just the quality of the evidence that is required to meet that single standard: the question is always whether a significant impediment to effective competition (test) is more likely than not (standard). However, naturally, the evidence that is required to give a satisfactory answer to that question must be stronger for more hypothetical theories of harm (such as conglomerate effects) than for more obvious ones (such as horizontal unilateral effects).⁹⁷

The conclusion that there is only one standard of proof can already be inferred from abuse case-law.⁹⁸ However, I would like the CJEU to affirm this as clearly as in merger control along the following lines: “the firm conviction needs to be reached that the plausible explanation for the conduct (standard), subject to rebuttal, alternative explanation or objective justification, is for the dominant company to derive an advantage (i) based on means either specifically designed to foreclose equally efficient competitors or that equally efficient competitors cannot use to derive a comparable advantage (means part of the test) and (ii) that equally efficient competitors cannot offset by other means (effects part of the test)”. I contend that the same rewriting that CK Telecoms undertook of Tetra Laval can be undertaken for exclusionary abuse case-law, where differences in standard are as apparent as change seemed to Parmenides: merely illusion. This would help put all the arguably different standards in a continuum from presumption of lawfulness, through elimination of competition and likely anticompetitive effect, to capability of producing an exclusionary effect.⁹⁹

A paradigmatic example is the apparently chaotic and fragmentary treatment of anticompetitive effects in Google Android. Ultimately, the discussion in that case was whether equally efficient competitors are doomed by the fact that they cannot replicate the conduct (which is therefore artificial). Depending on the type of practice, this cause-and-effect relationship is more or less plausible or hypothetical, which entails that the benchmark, and the intensity of the evidence required, is different, while the standard remains the same: foreclosure being the plausible explanation for the conduct. For antifragmentation agreements (AFAs), whereby device manufacturers that wished to pre-install Google apps could not sell devices running versions of Android that were not approved by Google, the General Court associates the anticompetitive effects with the actual exclusion of non-compatible forks.¹⁰⁰ Nonetheless, the reasoning bears more resemblance to the per se cursory analysis that used to be applied to exclusivity rebates before Intel II to prove the capacity to foreclose and that used to be limited to verifying that the contestable share was not enough to allow equally efficient competitors to compete viably.¹⁰¹

By contrast, when it comes to the mobile application distribution agreements (MADAs), whereby device manufacturers had to pre-install Google’s general search application (Google Search) and browser application (Chrome) to be able to obtain a licence to use its application store (Play Store), the foreclosure effect is equated to a significant competitive advantage that competitors are not able to offset. Concerning the revenue-sharing agreements (RSAs), whereby Google granted a percentage of its advertising revenues to device manufacturers and mobile network operators that agreed not to pre-install a competing general search service on any device within an agreed portfolio, a similar logic applies by virtue of the Intel II jurisprudence, as the amounts granted are exclusivity payments.¹⁰²

The reason, one might think, is that MADAs are classic tying practices, traditionally subject to a per se rule, which the General Court is more reluctant to move away from, while RSAs are exclusivity payments that have already been “de-per-se-alised” by Intel II. However, an alternative explanation could simply be that for some theories of harm, depending on their complexity, uncertainty or less clear cause-and-effect relationship, it is more difficult to meet the (one and only) requisite standard (i.e. plausibility of foreclosure of equally efficient competitors as the explanation for the conduct) than for others. This is precisely why we have per se rules, which are merely rules of experience (res ipsa loquitur). Experience tells us so clearly that foreclosure is the rationale behind a certain conduct (the object of the conduct in Art. 101 TFEU terms) that there is a priori no need to examine the plausibility of other “conceivable chains of cause and effect”, to use the words of Advocate General Kokott in CK Telecom.¹⁰³ This logic, which is also behind the restriction by object of Art. 101 TFEU (see Section 2), likewise underlies para. 39 in Unilever (see Section 4). The same applies to practices where the anticompetitive foreclosure has already materialised, or has not materialised yet despite the fact that the conduct has long been implemented – in those cases retroactive analysis instead of counterfactual analysis could make it easier to establish the plausibility of the foreclosure rationale, thus creating the false impression that the standard of proof is lower.

However, for less straightforward theories of harm, the various “conceivable chains of cause and effect” need to be examined. Counterfactual analysis might even be required to establish that the plausible explanation for the conduct is foreclosure of equally efficient competitors rather than any other alternative explanation. This is because, if only one alternative explanation were plausible, the benefit of the doubt would break the causal link by operation of the presumption of innocence of the dominant company (see para. 42 in Unilever, citing Woodpulp, and paras. 78, 80, 363 and 500 of Google Android), which, moreover, would be able to provide an objective justification. Plausibility (“more likely than not”) seems to always be the standard of proof.¹⁰⁴ This is consistent with the possibility of providing a plausible alternative explanation in order, on the grounds of lack of causality, to evade the prohibition on abuse, since the presumption of innocence would entail that any anticompetitive effects be attributed to the alternative lawful cause.¹⁰⁵ To what extent the artificiality of conduct and its potential anticompetitive effects can be presumed and how difficult they are to rebut depends on how clear the economic theory or rule of experience underlying the chain of cause and effect is – this is where the different legal tests come into play, which can be classified in the four categories described in Section 6.¹⁰⁶

That is why, in Google Android, the General Court considered that counterfactual analysis was necessary for RSAs (exclusivity payments), and annulled the Commission’s decision on that point,¹⁰⁷ but did not deem it necessary for MADAs (tying), instead placing the emphasis on the “anticompetitive nature of the objectives pursued”. I argue that the latter should be identified with a scenario in which an alternative explanation for conduct other than foreclosure of equally efficient competitors is hardly conceivable on the basis of a rule of experience or an economic axiom. However, the objective features of the conduct and its context (“all relevant circumstances”) still require a cursory analysis, similar to the assessment of the degree of harm for the restriction by object of Art. 101 TFEU, in order to show that the presumed effects are not hypothetical (subject to an Intel II-type rebuttal).¹⁰⁸

The different treatment of AFAs (more intense tying if we assume the Commission’s reasoning that they entail an outright exclusion of all contestability for non-compatible forks) and MADAs (lower-key tying) is also illustrative of how the seeming difference in standards (which in reality is just different levels of difficulty in meeting the same plausibility standard) depends on whether a cursory analysis of the degree of harm or anticompetitive nature is enough to reveal a plausible exclusionary explanation. While, for MADAs, the General Court refers to the specific facts of the case requiring “close examination of the actual effects or further analysis” (picking up on Microsoft) and to the relationship between “restriction” and “capacity to restrict”,¹⁰⁹ for AFAs it has little difficulty in dispensing with the counterfactual analysis and referring to the need to simply demonstrate that the practice was “capable of restricting competition”.¹¹⁰ Hence, there are three different intensities of assessment required to meet the standard (arguably there would be as many as there are theories of harm) for RSAs (which are clearly subject to stricter effects analysis after Intel II and the renvoi), MADAs (where a significant competitive advantage not able to be offset by competitors needs to be proven), and AFAs (where anticompetitive intent and the capacity to restrict competition based on the absence of contestability are enough).

6 Efficiency-Based Criterion and Legal Tests

As in the case of the standard, the mosaic of legal tests for specific abuses only proves the human need for discrete categories, and the universality and inseparability of the two limbs of the efficiency-based criterion. If one dissects any of those tests, the efficiency-based criterion underpins them all, in the sense that conduct that is artificial (meaning that there is no alternative explanation to the foreclosure of equally efficient competitors) cannot be offset by equally efficient competitors even by other means. Both limbs need to be considered jointly because, without either one, the conduct makes no anticompetitive sense (see Section 3). Indeed, the objective of all exclusionary legal tests is to ascertain the anticompetitive sense beyond mere intent, but rather meaning the overarching rationale or economy of the conduct that makes foreclosure the epistemological explanation for both means and effects, as shown in  Fig. 1. Such tests are, nonetheless, necessary for proving or ruling out specific theories of harm or hypotheses in the light of that unified methodology. In my view, this is what makes some authors rightly assert that different tests are necessary.¹¹¹

Fig. 1

Unified exclusionary abuse test mapping the traditional tests in a decision tree

It is true that the artificiality limb (competition on the merits) does not always require a check of whether the foreclosure rationale comes from the dominant company’s realisation or expectation that equally efficient competitors cannot obtain a comparable advantage by the same means (replicability), as this can also be clearly inferred from a self-evident or demonstrated exclusionary strategy. These two categories of artificiality or two ways of satisfying the first condition in the unified exclusionary abuse framework were suggested in Generics¹¹² and unequivocally stated in SEN.¹¹³ However, the fact that artificiality can sometimes be detected from the clear exclusionary object does not gainsay the universal applicability of the efficiency-based criterion as a connecting thread between the two limbs. Effectively, if there is a clear foreclosure strategy, it is presumed that the dominant company intends to evict equally efficient competitors without there being any need to prove that it can derive a comparable advantage by the same means (first limb). However, the anticompetitive rationale of the conduct still requires proof that equally efficient competitors are not able to offset the effects (second limb). That can be more or less difficult depending on the theory of harm. Conversely, if the anticompetitive rationale is not the clear object of the conduct, artificiality depends on equally efficient competitors not being able to replicate the advantage that the dominant company derives from that conduct – this reconciles the opinion of Advocate General Rantos with the conclusion of the CJEU in SEN.¹¹⁴

Actually, although categories are by nature despicably artificial, the classification in SEN may prove useful for the sake of explanation. Then, both categories can be segmented into two subcategories, depending on how clear the exclusionary effect under the second limb is. The resulting taxonomy could be as follows, as summarised in Fig. 2:

Fig. 2

Map of traditional tests according to the presumption of the artificiality (y) and effects (x) limbs

6.1 Cases Where Both the Exclusionary Object and the Foreclosure Effect Are Clear

The quintessential instance of a legal test that does not require proof of equally efficient competitors being unable to derive a comparable advantage by the same means or to offset the ensuing foreclosure effect by other means is selective price cuts by super-dominant companies.¹¹⁵ Their overwhelming strength vis-à-vis competitors and the clear foreclosure strategy allow for a rebuttable presumption that gaining a competitive edge over equally efficient competitors is the plausible rationale behind their conduct. The clear anticompetitive rationale can also stem from the fact that their conduct violates Art. 101 TFEU – this would be the case of an agreement among all incumbents holding a collective dominant position to control the parameters of competition in the market and avoid entry by fixing prices, setting quotas and restricting imports.¹¹⁶ The same goes for tying, where by definition the dominant company is leveraging its strength in one product to force customers into buying another product (regardless of its higher price, superior quality, greater innovation or, in general, higher relative efficiency).¹¹⁷ Another scenario would be the test for multi-product rebates in the Guidance Paper, according to which predation is to be assessed for each of the products, if competitors do not offer the whole range (because in that case the products are clearly being imposed regardless of efficiency), or for the whole package, if there are competing packages.¹¹⁸

Actually, the predatory pricing test itself also provides a rather illustrative example. It is obvious that, if the dominant company sells below its variable costs, the advantage can neither be replicated nor offset. Therefore, the economy of the conduct cannot be other than deriving an advantage. On the other hand, if the dominant company sells below only total costs, it does not necessarily lose money on any unit it markets (and nor would equally efficient competitors). In that case, the lack of replicability is not as clear as is the exclusionary effect. Thus, the exclusionary strategy becomes relevant: it must be ascertained whether the object of the conduct is to derive an artificial advantage that equally efficient competitors will not be able to beat in the longer run.

Cases where companies abuse their intellectual property or other exclusive or special rights would also fall under this first category. If a company were to exercise such rights with the intention of implementing a foreclosure strategy, this would be sufficient for the resulting advantage to be considered artificial (otherwise, the exercise of those rights would not be legally reproachable). Since competitors cannot measure up, it can be more safely presumed that any aggressive action by a dominant company has an exclusionary rationale (a sort of “qualified special responsibility”).¹¹⁹ This would explain the outlying principle of equality of opportunities in DEI. In this case, the quasi-monopolistic position in the electricity wholesale market that state-granted lignite exploitation rights gave the dominant company placed it beyond the efficiency of competitors. This inequality automatically made artificial (i.e. deviating from competition on the merits) any advantage to be derived therefrom (which obviously could not be replicated by competitors regardless of their efficiency). Only the other limb (potential effect) had to be proven and it could be presumed (“potential or actual anti‑competitive consequence liable to result”),¹²⁰ in a similar manner as for restrictions by object under Art. 101 TFEU, that is, following a cursory analysis of the degree of harm (or “abusive nature” or “intrinsic capacity”). Such a cursory analysis, traditionally identified with per se rules, needs to be more or less exhaustive depending on how clearly super-dominant the position is, which is, in turn, a function of market definition.¹²¹ The fact that exclusive or special rights are conferred on a dominant company by public authorities instead of being built by it from scratch also plays a role in presuming the artificiality of their exercise.¹²²

Finally, there are “artificial” refusals to deal. The lawful non-artificial decision not to deal can only be subject to Art. 102 TFEU if the indispensability of dealing makes the general interest in effective competition prevail, from a legal perspective, over the dominant company’s freedom to conduct a business and right to property under Arts. 16 and 17 respectively of the Charter of Fundamental Rights of the European Union, and, from an economic perspective, over its incentives to invest and innovate.¹²³ This balance to be struck under the essential facilities doctrine becomes unnecessary where the refusal to deal is “artificial”. That is the case when the exclusionary rationale behind the dominant company’s refusal to deal results from the fact that the dominant company is no longer able to decide not to deal. There are several possible reasons for this: one is that the company had already made the decision to deal (in quantum mechanics terms, the wave function of the decision to deal or not to deal had already collapsed) and then stops dealing¹²⁴; another is that a Union-compatible regulatory framework forces it to deal,¹²⁵ or entrusts it with an asset (to which the decision to deal refers) through exclusive or special rights or public finance (as opposed to Bronner-like self-developed and self-funded infrastructure);¹²⁶ another is when the company acts so outrageously that foreclosure is the only logical economic explanation for the conduct;¹²⁷ the reason could also be that the company does decide to deal but in terms that are de facto equivalent to refusal (margin squeeze¹²⁸ or non-price constructive refusal,¹²⁹ which are both to be distinguished from self-preferencing, where a company’s exclusionary rationale behind favouring its business in a related market is not so clear – see Section 4).

In those cases, a cursory analysis of the exclusionary effect, including a mechanical price-cost verification under the AEC test, along with consideration of the legal and economic context,¹³⁰ should be sufficient. This is because dealing is important (if not indispensable) for enabling equally efficient competitors to defeat foreclosure, which is again the result of applying the unified exclusionary abuse framework: if dealing is not so important, then the conduct does not make “exclusionary sense”.¹³¹ Note the references to the abusive nature of conduct across the judgments in Deutsche Telekom, TeliaSonera, Telefónica and Slovak Telekom/Deutsche Telekom II. Also, revealingly, the counterfactual analysis is put on the back burner.¹³²

Authors that suggest that the applicable legal test depends on the intensity of the envisaged remedy seem to come to a similar conclusion via a different avenue. In cases where the dominant company does not really have a choice not to deal, a reactive remedy (cease and desist) is sufficient, while imposing an ex novo duty to deal is a proactive remedy the intensity of which demands proof of indispensability under the stricter test of the essential facilities doctrine.¹³³ Advocate General Saugmandsgaard Øe’s reasoning in Slovak Telekom seems leaner to me: the only difference between a refusal to supply subject to the essential facilities doctrine and an “artificial” refusal to supply is simply that, in the latter, the dominant company has not refused to deal¹³⁴ and, therefore, the general efficiency-based dual-limbed abuse framework applies in accordance with Deutsche Telekom and TeliaSonera¹³⁵. It is true that we can use a remedy narrative to tell the same story:¹³⁶ since the dominant company has not refused to deal with the victims of the abuse, the remedy would not impose any duty on it to deal ex novo with companies that it had not yet selected as partners.¹³⁷

6.2 Cases Where the Exclusionary Object Is Clear but the Foreclosure Effect Is Not

The most prominent example of a case where the dominant company’s conduct is “objectively unlawful” (as the Spanish Supreme Court eloquently describes conduct that deviates from competition on the merits – see Section 3) is probably AstraZeneca.¹³⁸ One might think that the relative efficiency of competitors in this case was irrelevant because the practice, which consisted in making misleading representations to patent offices in order to extend patent protection beyond its expiry date, as well as deregistering marketing authorisations and withdrawing and launching presentations of drugs, could not have any rationale other than foreclosure. However, as that the efficiency-based criterion connects both limbs of the unified exclusionary abuse framework, it plays a fundamental role in causally linking potential foreclosure with conduct (see Section 3). Any company that holds a patent could lie about that patent’s expiry date but only a dominant company could reap the advantage that AstraZeneca reaped in the form of the potential foreclosure of equally efficient competitors, this being sufficient to make the conduct abusive.

The remaining categories of abuse under this class traditionally follow a logic of restriction by object because, although the foreclosure effect is not necessarily clear, the exclusionary purpose compensates therefor. The assumption used to be that exclusivity arrangements and tying and bundling (at least when competitors did not offer all the products of the bundle)¹³⁹ had no economic explanation other than that they deprived competitors of access to customers.¹⁴⁰ The assessment of foreclosure effects could therefore be overlooked.¹⁴¹ However, after Intel II, the narrative is that of “abusive nature” or “capability of restricting competition” (to be either presumed and then questioned by the dominant company or even cursorily proven in the first place by enforcers).¹⁴² Indeed, as already mentioned, in Intel renvoi and Qualcomm, the procedural rule in Intel II seems to have turned into a rather formalistic substantive obligation on enforcers to take into account all the relevant circumstances¹⁴³ and, more specifically, the five factors mentioned in para. 139 of Intel II (market position, market coverage, conditions for granting of rebates, duration and amount of rebates, and exclusionary strategy).¹⁴⁴

In particular, in Qualcomm, the efficiency-based criterion was consolidated as the key to deciphering the exclusionary rationale behind Qualcomm’s conduct by linking deviation from competition on the merits and foreclosure effect. Qualcomm’s exclusionary payments to Apple were found not to deviate from competition on the merits because no other competitor could have provided chipsets that fulfilled Apple’s requirements (i.e. no competitor was equally efficient). Thus, those payments were found to give rise to no foreclosure effect because the anticompetitive capability of the conduct was purely hypothetical.¹⁴⁵ Similarly, in Google Android, the General Court sticks to the dual-limbed efficiency-based exclusionary abuse framework both for the tying¹⁴⁶ and the exclusivity¹⁴⁷ theories of harm, by referring to a competitive advantage derived by the dominant company that equally efficient competitors are not able to (i) derive or (ii) offset. Whether such advantage needs to be significant or what the specific threshold is, if there is any, is not clear and seems to vary from one type of abuse to the other depending on how close they are to traditional per se categories.

6.3 Cases Where There Is no Obvious Exclusionary Object but the Foreclosure Effect Is Clear

For practices where the exclusionary purpose is not clear, the effects limb of the framework bears the burden of establishing the foreclosure rationale behind the conduct. The typical scenario would be one of leveraging the dominant position to favour the dominant company’s business in a locked-in neighbouring market. Although such self-preferencing theory of harm already existed in analogue markets, it often flew under the radar of Art. 102 TFEU, aside from cases where the exclusionary purpose was clearer (tying, margin squeeze or constructive refusal to supply). However, this practice has specific characteristics when engaged in by digital platforms, making it more likely to be abusive under the “leveraging” tag. Aside from dominance being more likely, as digital markets are by nature concentrated (owing, inter alia, to network effects and economies of scale and scope),¹⁴⁸ the difference from analogue world leveraging lies in the ambiguity of two elements: (i) digital platforms’ incentives to drive off competing business users, which is rational economically only if users are locked in, so that price and non-price gains from self-preferencing outweigh the foregone commission fees and the ripple effect on the ecosystem; and (ii) the effects of digital platform leveraging, which can entail short-term efficiencies resulting from vertical and conglomerate integration (avoidance of double mark-up, synergies and competitive pressure in neighbouring markets), but also a long-term reduction in innovation as a result of network effects and economies of scale and scope, leading to quicker and more irreversible market tipping.¹⁴⁹

Foreclosure incentives and effects can be likened to the two limbs of the exclusionary abuse framework. Therefore, the efficiency-based criterion, which could be implemented through a non-price version of the AEC test,¹⁵⁰ is the silver bullet to digital platform leveraging. Taking Google Shopping as an example,¹⁵¹ the General Court subjectively judged as “abnormal” Google’s decision to reserve prominent placement of results to its comparison service Google Shopping and to apply an algorithm that systematically demoted rival comparison services’ results in its general search result page by penalising their intrinsic features. Once the deviation from competition on the merits was thus established, the General Court upheld the Commission’s demonstration of potential exclusionary effects based on the importance of Google Search traffic for comparison services, the decisive user bias towards top results in the general search result page, and rival comparison services’ inability to effectively replace Google Search traffic. I claim that, in the absence of a clear exclusionary strategy and to avoid the subjective reproach against Google’s business model, the Commission and then the General Court could have applied the efficiency-based criterion to work out whether those three elements that pointed towards rival comparison services not being able to derive a similar advantage or offset it by other means were the plausible explanation why Google engaged in self-preferencing (subject to Google’s rebuttal or objective justification). In this way, there would have been no need to reverse the burden of proof.

The efficiency-based criterion would also solve the made-up¹⁵² problem that some commentators complained of,¹⁵³ namely that Google’s conduct did not fit into any established abuse category, because the test and the standard were one and the same – it was all about objectively ascertaining the exclusionary or legitimate rationale behind the conduct. Some authors, and even I, before SEN and Unilever, (futilely) argued that the best fit for digital platform leveraging was a non-price version of discrimination by super-dominant companies under Art. 102(c) TFEU, as in Irish Sugar¹⁵⁴ – referred to in para. 239 of Google Shopping).¹⁵⁵ Such reasoning could also lead to digital platform leverage being categorised under category 0 above, but the exclusionary rationale is less clear, given the above-mentioned ambivalence of incentives and effects. An example of abuse from the analogue world that falls under this class could be the test for predatory pricing where the price is above the variable cost but below the total cost, so that an exclusionary purpose must be shown.¹⁵⁶

6.4 Cases Where There Is no Obvious Exclusionary Object or Foreclosure Effect

Finally, the remaining abuse categories can be placed along the path of the continuum running from grey-zone abuses, such as loyalty rebates,¹⁵⁷ through conduct that is presumed lawful, such as quantity rebates,¹⁵⁸ to, ultimately, a legitimate refusal to supply. The latter only falls under Art. 102 TFEU if the exceptional circumstances under the essential facilities doctrine make the general interest in competition prevail over the dominant company’s right to property and freedom to conduct a business, and the economic incentives to innovate and invest.¹⁵⁹ It seems that the potential of the essential facilities doctrine is limited to those cases where the dominant company can lawfully take but has not yet taken the legitimate decision not to deal, and arguably where that company has developed and funded the indispensable resource itself (rather than the indispensable resource being a windfall entrusted by public authorities).¹⁶⁰ It also appears that the Commission is welcome to tie itself in knots applying the stricter essential facilities test in cases that, despite perhaps meeting the indispensability requirement clearly enough, might well have been subject to the regular exclusionary abuse framework.¹⁶¹ However, the essential facilities doctrine is not only about indispensability. It also raises the bar for the quality of evidence needed to prove more hypothetical chains of cause and effect to the same plausibility standard (to use the CK Telecoms narrative), entailing the more mundane, and thus neglected, requirement that a credible request to deal and a firm (if not outright) refusal be shown.¹⁶²

7 Conclusion

In this paper I posit that case-law on exclusionary abuses can be construed (or rewritten as in merger control) as setting out a single standard and a unified dual-limbed framework test by finding the “quantum” that is common to all seemingly different standards and tests: the relative efficiency of competitors. I suggest wording it along these lines: that the plausible rationale (nature or economy) behind a dominant company’s conduct (the standard) is for the dominant company to derive an advantage (i) based on means that are either specifically designed to foreclose equally efficient competitors or that equally efficient competitors cannot use to derive a comparable advantage (the means part of the test); and (ii) that equally efficient competitors cannot offset by other means (the effects part of the test). The whole is subject to rebuttal by the dominant company, which may do so by offering an alternative explanation (thus breaking the causality link), putting forward an objective justification, or simply casting doubt on the competition authority’s assessment in an Intel II-like fashion.

The underlying efficiency-based criterion, indissociably linking both limbs of the framework test, is not only clearly asserted by recent CJEU judgments and required by the very Darwinian essence of the competitive process that competition rules are meant to protect. It also features a number of advantages: (i) it makes sense of previous case-law; (ii) it gives the notion of abuse the objective nature required to fit with the legal content of Art. 102 TFEU (i.e. the special responsibility); (iii) it solves the causality problem by allowing the causal link to be between dominance and conduct, instead of between dominance and effects, (iv) it makes deviations from competition on the merits objective, thereby reducing competition authorities’ arbitrariness in this regard and enabling judicial review, (v) it bridges the gap between the assessment of abuses and agreements, as well as between the more economic approach and per se rules (to be applied in the form of a cursory analysis inspired by the analysis of the degree of harm necessary to establish restrictions by object under Art. 101 TFEU), and facilitates recourse to the single and continuous infringement device; and (vi) it helps strike a balance between, on the one hand, the general interest in competition and, on the other, the dominant company’s rights to property, freedom to conduct a business and economic incentives to innovate and invest, which are ensured by the essential facilities doctrine.

It is in this light that I respond to the main opinions voiced against the possibility of a unified theory of exclusionary abuses or the universality of the efficiency-based criterion. I then apply the unified standard of proof and the efficiency-based framework to the classic categories of exclusionary abuse. In this way, I show how all seemingly different standards and legal tests are just methods of adapting that one unified standard and framework to specific factual scenarios, so that the first or the second limb can be presumed to a greater or lesser extent, based on experience or economic theory, depending on the specific theory of harm. To make this point, I sort exclusionary abuses into four categories. Of course, any views from “equally efficient commentators” are more than welcome.