Law, Policy, Expertise: Hallmarks of Effective Judicial Review in EU Competition Law

Abstract This article discusses the techniques—‘hallmarks’—that the EU courts have developed to ensure that judicial review remains effective in competition law. These hallmarks can be grouped into three main categories. Some of them concern the interpretation of substantive law, namely the definition of legal tests and their consistent use over time. Some, the need for administrative action to rely on the best available evidence (which comprises both reliance on the expert consensus and the careful examination of the economic and legal context). A third category relates to the scrutiny exercised over the policy statements through which the European Commission chooses to constrain its discretion.


I. INTRODUCTION
Judicial review has been central to the direction and evolution of EU competition law. Many of the substantive choices that define the aims and boundaries of the discipline were made in the context of a challenge against a European Commission ('Commission') decision under Article 263 of the Treaty on the Functioning of the European Union ('TFEU'). 1 In Continental Can, the European Court of Justice ('Court' or 'ECJ') clarified that Article 102 TFEU encompasses both exploitative and exclusionary conduct. 2 In addition, it held that a link between the dominant position and the practice is not a precondition for a finding of abuse. 3 In Airtours, the suppliers and distributors) must remain within the outer boundaries of the provision on which intervention relies. For instance, the application of Article 102 TFEU demands, inter alia, showing that the relevant undertaking enjoys a dominant position within a substantial part of the internal market. 14 Accordingly, a decision that fails to establish dominance to the requisite legal standard (or interprets the notion erroneously) would be quashed if challenged. Insofar as it would, the annulment of the decision, even if confined to issues of law, would interfere with the Commission's policy choices. In fact, it may seriously affect its enforcement priorities if action hinges on a legal interpretation that the EU courts find to be erroneous.
The challenge for the EU courts, against this background, is to ensure that judicial review does not unduly interfere with the policy choices made by the Commission and, conversely, to ensure that their scrutiny lives up to their remit, which is the control, in full, of the legality of all aspects of administrative action. Evaluating and/or measuring whether the EU courts fulfil this demanding duty is not any less difficult. Rather than ascertaining the extent to which judicial review is effective, this article tackles the question indirectly. More precisely, it identifies and discusses a number of indicators ('hallmarks') of meaningful scrutiny of administrative action. These indicators are drawn from the case law and seek to reflect the techniques developed by the EU courts over the years.
The article is organised as follows. Parts II and III discuss, respectively, the role of courts in the EU legal order and how law and policy are entangled, thereby complicating the task of controlling administrative action. Parts IV to VI address, as such, the hallmarks of effective judicial review. These hallmarks are grouped in three main categories. Part IV focuses on the interpretation of substantive law and includes the definition of legal tests and their perpetuation over time. Part V shows how the EU courts ensure that administrative action is based on the best available evidence, whether it relates to non-legal expertise or the careful consideration of the economic and legal context. Finally, Part VI considers the role of policy statements in which the Commission chooses to constrain its discretion.

II. JUDICIAL REVIEW AND THE EU COURTS
Administrative action is subject to full review in the EU legal order, at least as a matter of principle. 15 The scope of court scrutiny encompasses all issues of law and fact addressed in a decision. In terms of intensity, the ECJ and the GC give no deference to the administrative authority. In accordance with Article 263 TFEU, it is for the EU courts, not the Commission, to state what the law is. Thus, it would not be sufficient for the latter to claim that the interpretation of a given provision in a decision is 14 In accordance with the first paragraph of Article 102 TFEU: 'Any abuse by one or more undertakings of a dominant position within the internal market or in a substantial part of it shall be prohibited as incompatible with the internal market in so far as it may affect trade between Member States'. 15  reasonable (or, similarly, not irrational or manifestly erroneous); it has to be the correct one. If, by the same token, there is disagreement between the authority and the courts about how to make sense of an issue of law, it is the latter's understanding that prevails, even if the relevant concept could in principle be construed in more than one way. Several factors explain that full review of issues of law and fact is the rule in the EU legal order. The first is that the EU courts are modelled upon continental regimes, in particular the French system of administrative courts. 16 In these regimes, it is typically accepted as self-evident that it is for courts, not agencies, to state what the law is and that the latter do not enjoy any leeway when defining the scope of their powers. In this regard, continental systems are at odds with some aspects of the Anglo-American tradition of judicial review. In particular, the ideas of Wednesbury unreasonableness 17 (whereby administrative action is upheld unless it is found to be irrational) and the Chevron doctrine 18 (whereby courts defer to the agencies' interpretation of a provision where the legislature is silent or ambiguous about its meaning) are foreign to the continental approach.
A second factor that explains why full review is the norm in the EU legal order has to do with the fundamental rights implications of the institutional makeup of the Union's competition law regime. The Commission is said to be both 'prosecutor' and 'judge', in the sense that it has the decision-making power in the same cases it decides to investigate. 19 This makeup is far from unusual in Europe (and, indeed, the US). Given the scale and complexity of State intervention in contemporary societies, expert agencies with adjudicative power are a necessity. 20 This necessity, however, is not without theoretical and practical consequences. To the extent that these authorities combine investigative and decision-making functions, they may be said to deny firms access to an 'independent and impartial tribunal' within the meaning of Article 6 of the European Convention on Human Rights ('ECHR'). 21 16 21 Article 6(1) ECHR reads as follows: 'In the determination of his civil rights and obligations or of any criminal charge against him, everyone is entitled to a fair and public hearing within a reasonable time by an independent and impartial tribunal established by law. Judgment shall be pronounced publicly but the press and public may be excluded from all or part of the trial in the interest of morals, public order or national security in a democratic society, where the interests of juveniles or the protection of the private life of the parties so require, or the extent strictly necessary in the opinion of the court in special circumstances where publicity would prejudice the interests of justice'.
The case law of the European Court of Human Rights ('ECtHR') suggests that an institutional system that revolves around an integrated authority is in line with Article 6 ECHR provided that the decisions adopted by the said authority are amenable to full review on all issues of law and fact. This is a principle that can be inferred from the Menarini judgment, 22 which concerned the Italian competition law system (and more precisely the nature of the penalties imposed for breaches of the regime). Following that judgment, delivered in 2011, the EU courts have consistently signalled their commitment to full review. 23 In this sense, the Court has substantially limited the range of issues over which the Commission is granted deference and are thus subject to limited review.
There is one doctrine that can be said to be at odds with the principle of full review of issues of law and fact in the EU. Early on, the Court held that, as far as 'complex economic' and 'technical' assessments are concerned, the Commission enjoys discretion. 24 As a result, such evaluations are only subject to limited review and would only be annulled if manifestly erroneous. The scope of the doctrine has never been entirely clear. 25 Irrespective of its boundaries, it would seem that the EU courts have, over time, played down its significance. Already in Tetra Laval, the ECJ held that the margin of discretion enjoyed by the Commission does not mean that the EU courts must refrain from assessing 'whether the evidence relied on is factually accurate, reliable and consistent but also 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'. 26 Since that seminal judgment, the scope and relevance of the doctrine appear to have declined, as noted by some members of the judiciary. 27 The GC judgment in CK Telecom exemplifies the trend perhaps better than any other, if only because 'complex economic assessments' were central to the case. 28 Evaluating whether a horizontal merger is likely to significantly impede effective competition demands, by definition, engaging in such evaluations. In spite of this fact, the judgment, when discussing the intensity of judicial review, does not mention the margin of discretion that preceding case law had recognised. What is more, the GC holds that the 'the scope of judicial review provided for in Article 263 TFEU extends to all the elements of Commission decisions taken on the basis of Regulation No 139/2004 which are subject to in-depth review by the GC, in law and in fact'. 29 It is worth noting, as a final point, that full judicial review is confined to the assessment of the legality of the Commission decision. The role of the EU courts, under Article 263 TFEU, is to ensure that the law is observed. One fundamental consequence is that it is not for them to substitute their assessment for that of the administrative authority. 30 In this sense, they do not have the power to decide whether the adoption of the decision was appropriate or desirable; their role is confined to considering whether doing so was lawful. In the same vein, legality review does not allow the EU courts to adopt or re-adopt a decision once it is found to be unlawful. The only area in which their role goes beyond legality control is in relation to fines, where they enjoy unlimited jurisdiction pursuant to Article 261 TFEU and thus the ability to increase or reduce the amount of the penalty. 31

III. DISENTANGLING LAW AND POLICY
A corollary to the final paragraph in the preceding section is that policymaking, as such, is not subject to full judicial review. To the extent that its lawfulness is scrutinised, court control is confined to manifest errors of assessment. It makes sense to discuss at greater length what falls within the definition of policymaking. It relates, primarily, to the definition of an agency's enforcement priorities. Competition authorities typically choose to devote their limited resources to some industries and/or practices. With the rise of the so-called 'more economics-based' approach, for instance, the Commission directed a substantial fraction of its capacity to the fight against cartels, which are understood to be the most pernicious practices for consumers and society. 32 In recent years, large digital platforms-the so-called Big Tech-have emerged as the clear enforcement priority. 33 Enforcement priorities can be defined positively or negatively. Starting with the former, the Commission may identify the industries or practices on which it will focus its resources. It may do so in several ways. The most obvious mechanism is by means of a decision to start an investigation or, rather, a string of investigations 29  in related activities or industries that, taken together, signal a policy direction. A systematic overview of the Commission's practice over the years shows that it is far from unusual to see clusters of decisions adopted in a short period of time. 34 Soft law instruments are another mechanism through which an administrative authority can express its policy choices. The Guidance on exclusionary abuses ('Guidance Paper') is a particularly eloquent example in this regard. In that document, the Commission spells out the criteria that it intends to follow when prioritising Article 102 TFEU cases. 35 The Draft Guidelines on horizontal co-operation agreements outline the Commission's policy in relation to arrangements driven by sustainability considerations. 36 The leeway that the Commission enjoys when choosing which cases to pursue entails, by definition, the ability to decide which not to investigate and, similarly, the discretion to close probes for lack of interest. The negative definition of policy choices frequently takes the form of complaint rejections. It has long been clear that the Commission cannot be compelled to start an investigation. 37 Because they remain within the boundaries of policymaking, complainants face an uphill battle when challenging rejections before the EU courts. As an expression of the authority's discretion, such decisions are only controlled, if at all, for manifest errors of assessment. 38 In this regard, the Court has consistently recognised that an administrative authority should be able to decide how its limited resources are best put to use.
Since the adoption of Regulation 1/2003, a key instrument through which policy has been expressed in a negative manner is the so-called commitments decision. 39 In accordance with Article 9 of that Regulation, the Commission can bring proceedings in a given case to an end without establishing whether or not Articles 101 and/or 102 TFEU have been infringed. Formally speaking, the role of this instrument is to make the commitments offered by firms binding upon them. Once the decision is adopted, the investigation is no longer a policy priority. 40  Commission's discretion (as opposed to an interpretation of the relevant provision), commitments decisions are only controlled for manifest errors of assessment. 41 Commitments decisions are peculiar in that they allow authorities to circumvent issues of law. This paper focuses on a different set of circumstances, namely those where the expression of a policy choice and the interpretation of a legal provision are entangled. These circumstances are, in fact, the default scenario in EU competition law. As already pointed out, the law is the vehicle through which policy is formulated. This is true, most obviously, where an investigation leads to the adoption of a decision establishing an infringement of Articles 101 and/or 102 TFEU 42 or declaring the incompatibility of a concentration with the internal market. 43 The fines that often accompany a formal finding of a breach are also a reliable indicator of an authority's enforcement priorities. 44 Penalties are deemed an effective mechanism to signal the seriousness of particular infringement 45 and as a deterrent steering firms' conduct. 46 When policy choices are wrapped in a decision formally construing the meaning and scope of the applicable provision, the EU courts are presented with a conundrum. On the one hand, legal interpretations are controlled in full; on the other, the policy choices underpinning an investigation are subject, if at all, to marginal scrutiny. As a result, the application of one standard of review might contaminate the other. The strict control of whether the Commission has correctly interpreted the relevant provision necessarily has consequences for the formulation of policy choices. To the extent that policy is expressed, first and foremost, by means of individual decisions, the annulment of each one of these decisions has an impact on the reach and effectiveness of enforcement in a given area, or in relation to certain practices.
The consequences of an annulment for policymaking would be particularly acute where enforcement is premised on a particular interpretation of the applicable provision. Suffice it to mention an example in this sense. Suppose that Commission action against former monopolies in recently liberalised network industries is based on the assumption that, once it is established that they enjoy a dominant position, it is not necessary to show that incumbents' behaviour is capable or likely to have anti-competitive effects. Under this interpretation of Article 102 TFEU, it would only be necessary for the authority to establish that the contentious practice has been implemented. If the Court concludes that this understanding of the notion of abuse is unlawful, the correction of the error of law would reduce, by definition, the reach of policymaking, if only because it would set the bar of intervention higher 41 Commission v Alrosa Company Ltd, C-441/07 P, EU:C:2010:377. 42 That is, a decision adopted in accordance with Regulation 1/2003, note 31 above, Art 7. 43  and would make it more difficult for the authority to discharge its burden of proof in individual cases. Because of the impact that a strict standard of review may have on the reach and scope of policymaking, the EU courts might show deference to the Commission. There is, as a consequence, a risk that marginal review becomes the de facto standard in order for the judiciary to avoid interfering with the definition of the authority's enforcement priorities and, in effect, substituting their assessment for that of the latter. This concern is not a theoretical one. It has been expressed, in the past, by EU judges and law clerks. 47 At the same time, this concern (and the deference that derives from it) is at odds with the EU courts' duties pursuant to Article 263 TFEU. Under the architecture of the EU treaties, the impact of intense court scrutiny on the reach and effectiveness of intervention is not a cause for concern but an inevitability, and arguably a necessary reminder that policymaking must remain within the boundaries of what is allowed by law.
It is not infrequent for commentators to argue that, precisely because of the potential impact of the annulment of a decision on policymaking, the EU courts tend to be deferential to the Commission and thus fail to fulfil their obligations under the Treaty. 48 These claims are, however, not easy to substantiate. The rate of annulment of administrative action is an unreliable proxy for the effectiveness of judicial review: it is not necessarily because a challenge is dismissed that court scrutiny does not meet the requisite threshold of scope and intensity. 49 Conversely, the fact that the EU courts declare their commitment to full review does not necessarily mean that the control is in keeping with their pledge: as the ECtHR pointed out in Menarini, what matters is what judges do, not what they say they do. 50 This paper does not seek to ascertain whether judicial review is effective or not. Instead, it discusses a number of techniques that the EU courts have developed and that make meaningful scrutiny of administrative action possible, thereby ensuring that the Commission does not enjoy-de facto if not de iure-unwarranted discretion over issues of law. The hallmarks of effective judicial review addressed hereinafter pursue three main objectives. First, to craft well-defined legal boundaries to administrative action. Second, to make sure that the said action is grounded on the best available evidence (as opposed to unsubstantiated hypotheses, untested theories or hypotheticals). Third, to provide legal certainty and allow firms to reasonably anticipate the likelihood and outcome of intervention. The hallmarks of effective judicial review can be categorised along three main themes. A first theme concerns the way in which the EU courts engage with substantive EU competition law. In this regard, it makes sense to pay attention to the techniques they follow when crafting legal tests and how they deal with consistency in the application, over time, of the said tests. A second theme relates to the use of the best available evidence, and more precisely the role of non-legal expertise (in particular economic analysis) and of the relevant context of which the practice is a part. A final theme has to do with the role of policy statements made by the Commission. It appears, in this sense, that while policy choices may only be subject, if at all, to limited review, the EU courts expect the administrative authority to behave in a manner that is consistent with the positions to which it commits.

A. The Definition of Legal Tests
EU competition law provisions are famously broad and vague. They prohibit certain practices without defining the key concepts. Just to mention two examples, nowhere does Article 102 TFEU construe the notion of abuse, and nowhere does Regulation 139/2004 explain what amounts to a significant impediment to effective competition. At most, primary and secondary law provide some high-level (and non-exhaustive) examples of behaviour that could lead to an infringement, 51 or identify a number of criteria that need to be considered when ascertaining whether the conditions for intervention are met. 52 What is more, the notions used in primary and secondary law would not be sufficiently specific, even if they were fleshed out. Stating what an abuse or a restriction of competition is provides, at best, some broad principles.
The content and scope of competition law provisions can only be defined through the case-by-case administration of the law. The sort of questions that individual instances raise include, for example, whether below-cost prices charged by a dominant firm are abusive and whether setting up a joint venture to produce an input for a finished product is restrictive of competition. Legal tests are the tool that, in effect, shapes the law and bridges the gap between the competition law provisions and the peculiarities and demands of particular cases. Crafting legal tests demands, first, the categorisation of practices and, second, the definition the conditions under which the various categories of practices give rise to a prima facie infringement (or, in the case of a merger, to prima facie incompatibility). Once crafted, 51 For instance, Article 101 TFEU refers to agreements and concerted practices that, inter alia '(a) directly or indirectly fix purchase or selling prices or any other trading conditions' and '(b) limit or control production, markets, technical development, or investment'. 52 See in this sense Article 2(1) of Regulation 139/2004, note 43 above.
the legal test will be the template against which the lawfulness of behaviours or transactions will be evaluated. The scope of competition law provisions has been defined, to a significant extent, through the dialogue between the Commission and the EU courts. An analysis of this interaction hints at a pattern in the approach followed by each of the players when developing legal tests. Review judges have consistently favoured clear and welldefined criteria, the respect of which can be easily controlled when administrative action is challenged. Conversely, the Commission has displayed a tendency to craft relatively unstructured legal tests that can be adjusted to the circumstances of each case. From the authority's perspective, the factors that are relevant (or dispositive) in one instance are not necessarily so in another one. Against this background, it is not surprising that this disparity of views about what an optimal test is has given rise to occasional frictions and, similarly, to the annulment of some Commission decisions.
Examples abound in the case law. A relatively early one can be found in the AKZO ruling of 1991. 53 This case addressed the legal status of predatory pricing in the EU legal order. In its decision, the Commission had proposed a legal test that could be described as holistic, in the sense that it was grounded on a broad range of considerations, which might or might not be relevant in the context of a specific case. 54 Under this interpretation, most notably, a finding of abuse would not be contingent on the authority showing that the practice would lead to below-cost prices. 55 The ECJ judgment, on the other hand, crafted a cleaner legal test identifying two instances in which an aggressive pricing strategy would give rise to an infringement within the meaning of Article 102 TFEU. 56 Under this legal test, evidence of below-cost prices would be a necessary precondition for the application of the provision.
As far as Article 102 TFEU is concerned, a recent example can be found in the Intel saga, which will be discussed at greater length below. In its 2017 judgment, the Court identified a number of indicators that the Commission must consider once a firm provides evidence showing that a system of loyalty rebates is incapable of restricting competition. 57 These factors include the extent of the dominant position and the coverage of the practice. 58 When the case was sent back to the GC, the latter found that the Commission had erred in law by failing to assess some of these factors, and the decision was accordingly annulled. 59  that the administrative authority had failed to evaluate the coverage of the practice when ascertaining whether it had the potential to negatively impact competition. 60 In line with the discussion above, the GC understood the judgment in Intel as demanding from the authority an evaluation of all five factors identified by the Court. The failure to consider one of them (coverage) would necessarily entail the annulment of the decision. The Commission appealed the renvoi judgment. 61 As one would expect given the way in which it has approached the definition of legal tests in the past, the authority argues in its first ground of appeal that the anti-competitive effects of the practices at stake in the case can be established without necessarily taking into account its coverage. 62 From this perspective, it would be possible to compensate the failure to assess this factor if the 'overall assessment' is indicative of a restriction. 63 Another recent example illustrating this difference of views between the agency and the review courts can be drawn from merger control, and more precisely the CK Telecoms judgment of 2020. 64 This case provided the first occasion for the EU courts to define the conditions under which a merger that does not create or strengthen a dominant position (whether individual or collective) nevertheless gives rise to a significant impediment to effective competition. 65 This, in other words, was the opportunity to deal with the so-called 'gap cases', 66 which would not have fallen within the scope of Regulation 4064/89. Because transactions falling below the threshold of dominance were not previously caught by the merger control regime, there were no obvious precedents on which to rely. In particular, the EU courts had to identify the instances where an impediment to effective competition is significant and thus leads to a prima facie finding of incompatibility.
In its decision declaring the incompatibility with the internal market of the merger between Hutchison 3 G UK and Telefónica UK, the Commission argued that a significant impediment to effective competition can be established on the basis of an unstructured test. More precisely, it explained that, in its view, there are a number of factors that are 'potentially [but not necessarily] relevant' in the context of a 60  particular transaction. According to this interpretation of Article 2 of Regulation 139/ 2004, the criteria to evaluate the concentration would be decided by the authority on a case-by-case basis. This approach to the definition of legal tests is not fundamentally different from that defended in the AKZO decision, discussed above. In this vein, the Commission argued that a finding of incompatibility does not necessitate evidence showing that one of the parties is an 'important competitive force' within the meaning of its own Guidelines. 67 The GC did not uphold the Commission's interpretation of Regulation 139/2004. The unstructured test suggested by the authority would have made it difficult, if not impossible, to anticipate the outcome of an investigation. In fact, as noted by the GC, the Commission had construed Article 2 of the Regulation in such a way that it would have enjoyed, de facto, the power to declare the incompatibility of virtually any horizontal transaction with the internal market. 68 The decision was indeed based on the premise that a significant impediment to effective competition could be established without showing that the transaction would lead to the elimination of 'important competitive constraints' and that such constraints necessarily stand out relative to other. 69 To the extent that concentrations between actual or potential rivals lead to the reduction of some competitive constraints, it is difficult to see how the holistic test developed by the Commission would not have been met in every horizontal merger.
Instead, the CK Telecoms judgment introduced a test based around two criteria. Accordingly, the first question is whether the transaction would eliminate important competitive constraints between the parties. The second, whether the transaction would lead to a reduction of competitive pressure faced by the remaining rivals. In addition, the GC interpreted these criteria in a manner that would not amount to giving, in effect, discretion to the Commission. In particular, it noted that the competitive constraints that would be eliminated as a result of the merger must stand out relative to those coming from rivals. In other words, the first criterion demands evidence showing either that the parties were particularly close competitors (that is, that rivalry between them was distinctly intense) 70 or that one of them acted as a 'maverick' on the relevant market. 71  69 Ibid, para 158. 70 Ibid, paras 227-50. 71 Ibid, paras 155-226. criteria to establish the abusive nature of exclusive dealing obligations were permitted to fluctuate and the Commission were allowed to change the applicable criteria from one case to another. In such circumstances, the test laid down in Intel would not provide legal certainty in any meaningful way, and firms would not be able to anticipate the outcome of administrative action. The Commission would enjoy, in effect, the discretion to define the scope of legal provisions. There should be little doubt, against this background, that effective judicial review is only possible where there is inter-temporal consistency in the definition and administration of legal tests. By the same token, checking whether Commission decisions are in line with the relevant case law (and, similarly, whether there are reasons justifying a departure from the case law) is an indispensable hallmark of effective judicial review. The need to ensure consistency through judicial review is all the more important if one takes into account that administrative authorities, including the Commission, may be tempted to depart from the legal tests already laid down in the case law. Examples illustrating this point can be easily found. From the very early days, the Court held that a limitation of a firm's freedom of action is as such insufficient to establish a restriction of competition. As explained in detail below, the actual or potential effects of a practice are assessed by reference to the relevant market, not their impact on individual firms. 72 For years, the Commission did not follow this aspect of the case law and continued to infer a restriction of competition from a limitation to a firm's freedom of action. 73 However, the case law remained consistent over time. 74 Another example of this same tendency is provided by the case law dealing with refusals to deal. The conditions under which a firm can be compelled to conclude an agreement with an actual or would-be rival are notoriously difficult to meet. A refusal only amounts to an abuse of a dominant position where it relates to an input or infrastructure that is indispensable for competition on an adjacent market and where, in addition, it would lead to the elimination of all competition in the latter market. 75 The conditions are event stricter where the input in question is protected by intellectual property rights. 76 It is not surprising that the Commission has sought to 72 Stergios Delimitis v Henninger Bräu AG, C-234/89, EU:C:1991:91, para 15: 'It is necessary to analyse the effects of a beer supply agreement, taken together with other contracts of the same type, on the opportunities of national competitors or those from other Member States, to gain access to the market for beer consumption or to increase their market share and, accordingly, the effects on the range of products offered to consumers'. 73 For a discussion, see Chapters 3 and 6 in P Ibáñez Colomo, circumvent this legal test by applying a different one or by arguing that the behaviour under consideration should be distinguished from a refusal to deal. In Microsoft, for instance, the Commission argued that it could establish the abusive nature of a refusal to deal against a different set of criteria, 77 and claimed, before the GC, that it would be 'problematic' to apply mechanically the conditions laid down in the relevant precedents. 78 In addition, it argued that there were several factual differences between this case and those at the origin of the case law (namely the fact that the practice amounted to a disruption, as opposed to a refusal to deal; and the fact that information was not protected by intellectual property rights). In spite of these claims, the GC did not deviate from the relevant precedents and assessed the lawfulness of Microsoft's behaviour against the criteria laid down in Magill. 79 The first-instance court displayed the same commitment to inter-temporal consistency in the interim measures in IMS Health. 80 A third example comes from the treatment of predatory pricing after AKZO. As explained above, this judgment laid down a clear set of criteria to establish when a policy of aggressive pricing by a dominant firm amounts to an abuse. The fundamental implication of AKZO is that above-cost prices, no matter how aggressive, would not be caught by Article 102 TFEU. 81 In two subsequent cases-Irish Sugar 82 and Compagnie Maritime Belge 83 -the Commission concluded that aggressive pricing strategies can amount to an abuse of a dominant position even when prices do not fall below cost. The inconsistency between AKZO and these two decisions was not corrected by the EU courts when the latter were challenged. The resulting legal uncertainty-acknowledged by the Advocates General in their Opinions 84 -was not addressed until the Court judgment in Post Danmark I. 85 This last example shows, a contrario, how important inter-temporal consistency is for the purposes of ensuring that judicial review remains effective and the law is predictable. 77 Microsoft Corporation v Commission, EU:T:2007:289, paras 304-09. 78 Ibid, para 316. 79

V. EFFECTIVE REVIEW AND THE BEST AVAILABLE EVIDENCE
A. The Role of Non-legal Expertise 1. Expert consensus as a constraint on administrative action Competition law cannot be interpreted or enforced without engaging in economic analysis, be it implicit or explicit, formal or intuitive. Contrary to what is occasionally suggested, reliance on economic concepts cannot be circumvented. The choice for courts and authorities, in this regard, is not whether to rely on these tools, but whether to make legal notions explicitly revolve around formal concepts or, instead, trust informal (and often implicit) analysis. Think of the substantive test in EU merger control. When assessing whether a transaction is likely to lead to a significant impediment to effective competition, an authority may decide not to rely on the mainstream consensus when shaping the relevant provision. This choice does not mean, however, that economic analysis will have been avoided. The decisions issued by the authority will still be built around an (informal, heterodox) analytical framework.
For instance, the consensus among economists is that conglomerate mergers are generally unproblematic and, to the extent that they can be presumed to lead to efficiency gains, they are also prima facie pro-competitive. 86 Accordingly, it is accepted that intervention would have to rely on a theory of harm specifying the mechanism through which competition would be affected and showing why the merged entity would have the ability and incentive to engage in that strategy. Administrative action may rely on this consensus. Alternatively, it may disregard it and claim that the accumulation of power across a number of adjacent markets is problematic in and of itself, irrespective of whether the leveraging of market power is a viable and likely scenario in the circumstances of the case. 87 The latter approach does not rely on economic analysis any less than the former; it is just a different strand thereof.
The question of whether legal provisions should be construed in light of the established body of non-legal expertise has important consequences from the perspective of judicial review and the powers of an administrative authority. If the EU courts gave deference to the Commission in relation to the economic apparatus on which its assessment relies, the latter would have the discretion, in effect, to decide whether or not to be constrained by the mainstream consensus. If the authority enjoyed leeway in this regard, the very choice of the theoretical framework (and the assumptions underpinning that framework) would be, from this perspective, a 'complex economic assessment'. Alternatively, the EU courts could treat this question as an issue of law that is subject to full judicial review: under this approach, the mainstream consensus would constrain administrative action.
A careful overview of the case law shows that the EU courts do not give deference to the Commission when it comes to the non-legal foundations of the assessment.
Administrative action, in the EU legal order, must be the expression of the best available evidence. Thus, if the body of economic expertise suggests that a given practice can only produce anti-competitive effects in certain scenarios, the administrative authority does not have the freedom to ignore this consensus, or to rely on heterodox, intuitive or discredited views suggesting a different conclusion. Similarly, where the consensus suggests that there are several potential explanations for a given phenomenon, the Commission must acknowledge, when interpreting the relevant legal provision, that a given line of conduct could be explained in more ways than one. For instance, the Commission would have to acknowledge that tying conduct does not necessarily serve an exclusionary purpose. 88 These ideas are more effectively illustrated by reference to some concrete examples drawn from the case law. Consider Woodpulp. 89 One of the fundamental questions at stake in the case related to the notion of concerted practice under Article 101 (1) TFEU, and more precisely whether parallel conduct provides evidence, in and of itself, of explicit collusion. In its decision, the Commission had not produced any direct proof of such collusion. 90 The Court, in light of the available body of expertise, concluded that the adoption of a common course of conduct is as such insufficient to trigger the application of Article 101(1) TFEU, and this insofar as explicit collusion is not the only plausible explanation for it. Accordingly, the Commission, to discharge its burden of proof, would have had to show not only that the relevant undertakings had adopted a system of parallel price announcements, but that the said system can only be rationalised as the outcome of a concerted practice. 91 Merger control provides several additional examples revealing the importance of the expert consensus as a constraint on administrative action. The most obvious one comes from the Airtours judgment of 2002. 92 The case concerned the meaning and scope of the notion of collective dominance within the meaning of Regulation 4064/89. In its decision, the Commission had construed the notion in a manner that departed, in several fundamental respects, from the mainstream. In particular, the authority suggested that a concentration could be deemed to create or strengthen a position of collective dominance whenever it increased firms' incentives to collude. 93 What is more, the definition of the notion would not necessarily have to be defined in light of the relevant economic concepts.
The GC, by contrast, brought the legal notion of collective dominance in line with the mainstream (and, in the same vein, with the concept of tacit collusion only that the remaining players would have the incentive to collude, but also that they would have the ability to do so in a sustained manner. The insights from economic theory were turned into an operational legal test revolving around three cumulative conditions aimed at establishing whether the transaction would create the circumstances in which the new entity, together with any remaining rivals, would find it both feasible and profitable to adopt a common course of conduct on a lasting basis. 94 The expert consensus has played a central role in other areas of EU competition law. A relatively recent example can be drawn from the Cartes Bancaires judgment of 2014. 95 The case concerned a series of measures implemented by the members of a cooperative joint venture. One of the key questions raised before the Court had to do with the object of the agreements at stake (and more precisely whether, because of their purpose, the said measures were inherently anti-competitive). When assessing the nature of the arrangements put in place by the joint venture, the ECJ noted that the Commission could not ignore the fact that the relevant market had the features of a two-sided platform. 96 This concept, borrowed from economic theory, 97 had not been duly considered in the decision. 98 Ensuring that administrative action is aligned with the expert consensus is consistent with the EU court's general approach to judicial review. Given how intertwined legal and economic concepts are in EU competition law, full judicial review would not be possible if the Commission were allowed to disregard the accumulated body of knowledge developed by experts over the years. If, for instance, the authority were allowed to disregard the mainstream consensus when defining the notion of collective dominance, or when assessing the object of an agreement, it would, in effect, enjoy discretion to define the scope of legal provisions. In a case like Cartes Bancaires, for instance, the nature of the agreement could only be appropriately grasped by relying on the appropriate theoretical framework.
Embracing mainstream expertise is also an effective mechanism to ensure that firms can exercise their rights of defence and effectively challenge the arguments raised by the administrative authority. If the Commission were not bound by the best available non-legal expertise, it would simply not be possible for undertakings to disprove its claims. The authority could argue, for instance, that evidence of parallel conduct is sufficient, in and of itself, to establish a concerted practice. Absent the safeguard provided by the accumulated body of knowledge, a firm would not be able to meaningfully argue that there are several plausible explanations for parallel conduct and that inferring explicit concertation from that behaviour is inconsistent with consensus views. Similarly, the parties to a merger would not be in a position to argue that tacit collusion is unlikely to arise and be sustained absent retaliation mechanisms.

Administrative action absent consensus
The judgments discussed above do not address the separate question of whether an administrative authority can rely on economic analysis where no consensus has emerged about the nature and purpose of a practice and/or its potential pro-and anti-competitive effects. This situation may arise where the practice is relatively novel and is not yet fully understood. The Special Advisers' Report on Digital Markets 99 is illustrative in this sense. Its authors acknowledge that there is much that is not entirely grasped about conduct implemented in digital markets, in particular in relation to the efficiencies to which they may lead. 100 Where there is no consensus, the question-from the perspective of competition law enforcement and its review-is whether the Commission can rely on the expertise that is aligned with the outcomes it favours from a policymaking standpoint. On this point, the case law suggests that the Commission cannot rely on a body of expertise that is contested or about which there is no consensus. By the same token, any conclusions grounded on such expertise will be insufficient to establish an infringement to the requisite legal standard. This conclusion can be inferred from Budapest Bank. In that judgment, the Court held that a finding that an agreement restricts competition by object presupposes that there is 'sufficiently general and consistent experience' about the nature of the practice and its effects. 101 Where the expertise is ambivalent about the rationale behind the conduct (for instance, because some of the expertise suggests that there is a potential pro-competitive rationale for it), the authority cannot be dispensed from the need to consider its effects. 102 While Budapest Bank addresses a relatively narrow issue, it is difficult to see how it would not be relevant in other instances, and in particular where the contested expertise relates to the likely (pro-and anti-competitive) effects of a practice or transaction.
The position expressed by the Court in Budapest Bank is also consistent with the general principles of EU law (and, for that very reason, also relevant beyond its limited scope). In Intel, for instance, the GC noted that competition cases are quasi-criminal in nature. As a result, the presumption of innocence applies. This fundamental principle is necessarily relevant where the administrative authority relies on expertise that does not reflect consensus positions. In such circumstances, any doubts about the reliability of the said expertise must benefit the defendant. In addition, Expedia mentions that allowing the Commission to depart from the policy statements declared in a soft law instrument would be inconsistent with the principle of equal treatment and the protection of legitimate expectations. 103

B. The Relevant Economic and Legal Context 1. The relevant economic and legal context and the object of practices
The EU courts have consistently held that a restriction of competition cannot be established without paying due regard to the relevant economic and legal context. The fundamental implication is that the question of whether a practice amounts to an infringement of Articles 101 and/or 102 TFEU (or whether a transaction is likely to significantly impede effective competition) is a case-specific exercise. Accordingly, abstract categories (such as 'price-fixing' or 'market sharing') cannot and do not determine, in and of themselves, whether an agreement is restrictive of competition by object. An agreement featuring one or more such categories is not necessarily caught by Article 101(1) TFEU, let alone by its very nature (in fact, it may sometimes escape the prohibition altogether).
The EU courts have frequently annulled Commission decisions due to their failure to fully take into consideration the legal and economic realities of which contentious practices are a part. As far as the evaluation of the object of agreements is concerned, Cartes Bancaires provides an eloquent example. The Commission decision was found to be erroneous in law insofar as it did not acknowledge the peculiar features of the relevant market, and more precisely the need to put in place balancing mechanisms to prevent free-riding. 104 The idea that the evaluation of the object of an agreement demands a case-by-case, context-specific inquiry (and that failing to meaningfully account for the circumstances surrounding the practice leads to the annulment of administrative action) is confirmed even when one considers cases where the Commission decision was not quashed. 'Pay-for-delay' cases illustrate this idea well, if only because of the uniqueness and complexity of the relevant context.
On the surface, a 'pay-for-delay' arrangement (whereby a pharmaceutical company -the 'originator'-makes a transfer of value to another one-the 'generic producer'and the latter, in turn, agrees to delay its entry into the market and not to challenge the validity of the patent held by the former) seems suspicious. Sharing profits with a would-be rival as consideration for the latter not competing looks like a blatant 'market sharing' cartel. However, the Court-in Generics, and later in Lundbeck-emphasised the importance of the relevant economic and legal context. Thus, it did not hold that 'pay-for-delay' agreements are, in the abstract, restrictive of competition by object. 105 Whether or not they amount to a breach of Article 101(1) TFEU by their very nature necessitates a careful assessment of, inter alia, the regulatory regime 106 (considering that intellectual property legislation is central in these cases) and of the specific terms of the agreement, 107 including its potential to yield pro-competitive gains. 108 The legality of subsequent administrative action has been controlled in accordance with these principles. 109

The relevant economic and legal context and the effects of practices
The same can be said about the assessment of effects. The Court has consistently held that the evaluation of the potential impact of a practice must not be purely hypothetical. 110 It must be grounded on the concrete realities in which it is implemented. Inter alia, anti-competitive effects must be assessed by reference to the relevant counterfactual, that is, by reference to the conditions of competition that would have existed in the absence of the practice. 111 Thus, there is no infringement where the conditions of competition would be the same with and without the practice. 112 In addition, the Court has identified a number of factors to evaluate, and in particular the degree of market power enjoyed by the firm(s), the coverage of the behaviour, the nature of the product or service, and the features of the relevant market. 113 The authority's failure to consider the counterfactual has frequently resulted, over the years, in the annulment of Commission decisions. In Nungesser, for instance, the Court concluded that, given the investments made in the development of a new plant variety, and given the risks undertaken by the licensee when introducing it in a new geographic market, some of the territorial restrictions included in the technology transfer agreement at stake in the case fell outside the scope of Article 101(1) TFEU. 114 Because the Commission decision had failed to acknowledge the legal implications of these features, it was quashed. Administrative action was also annulled in subsequent cases, including European Night Services. In that case, the GC noted, in light of a careful evaluation of the relevant regulatory and economic context, that the Commission had failed to show that the parties to the joint venture qualified as actual or potential competitors. 115 O2 led to a similar outcome. 116

Procedural dimension
The duty to consider the relevant economic and legal context not only has a substantive dimension (described above), but also a procedural one. In this sense, the Commission is under a duty to consider the arguments pertaining to the specific circumstances of the case when raised by the parties. Accordingly, arguments to the effect that a practice does not restrict competition in the relevant economic and legal context cannot be brushed aside by the authority. Where these arguments are capable of substantiating doubts about the existence of a restriction, the Commission must engage with them and show to the requisite legal standard why the practice amounts to a prima facie breach of Articles 101 and/or 102 TFEU. Crucially, these arguments are relevant irrespective of whether the conduct under examination would otherwise qualify as a 'by object' or 'by effect' infringement.
The duty, incumbent upon the Commission, to evaluate the arguments raised by undertakings during the administrative proceedings, is one of the key lessons to draw from the Intel ruling, mentioned above. 117 The case is notable in that the practice at stake-a system of 'loyalty rebates'-is, in principle, abusive by its very nature. 118 Accordingly, the application of Article 102 TFEU was not contingent on evidence of actual or potential anti-competitive effects. In spite of this fact (and even though it confirmed that Hoffman-La Roche remains good law 119 ), the Court held that, once the dominant undertaking provides evidence showing, in light of the relevant economic and legal context, that the practice is incapable of having restricting effects, the Commission must engage with this evidence and evaluate its potential impact in light of the factors mentioned above. 120 Failing to do so would inevitably result in the annulment of administrative action.
The principles laid down in Intel were a central aspect of the assessment of the legality of administrative action in Qualcomm. 121 The Commission decision in the latter, adopted shortly after the Court judgment in the former, was annulled in its entirety on procedural and substantive grounds. The dominant firm in that case, just like Intel, had advanced arguments pertaining to the relevant economic and legal context. It is on the basis of these arguments that the GC came to the conclusion that the Commission had failed to establish an abuse of a dominant position, and this insofar as it had not considered all the relevant circumstances. In particular, the judgment notes that the authority failed to take the relevant counterfactual into account. 122 The GC's analysis reveals that, even in the absence of the practice, the relevant customer would not have been able to find an alternative supplier for 90% of its requirements. Insofar as this is the case, it is not easy to see how the practice (at least in relation to these requirements) could restrict competition, whether by object or effect.

A. The Binding Nature of Policy Positions
It has been explained above that policymaking is only controlled, if at all, for manifest errors of assessment. This fact does not mean that policy statements made by the Commission do not have any binding effects. From the perspective of an undertaking subject to Articles 101 and 102 TFEU, it is reasonable to expect that, when an administrative authority announces that it is going to exercise its discretion in a particular way, subsequent behaviour is consistent with it. Consistency between policy as declared and policy as exercised would be a requirement of the principle of good administration. 123 The fact that policy positions are not a statement of the law is not a crucial factor in this regard. The ability to reasonably anticipate the outcome of administrative proceedings is not confined to instances in which the authority is interpreting or applying a particular legal provision. Thus, a decision deviating from a stated policy position may be annulled.
The above appears to summarise effectively the legal status of the Commission's declared policy positions in the case law. In Expedia, 124 the Court considered the binding nature of the De Minimis Notice, 125 which signals how the Commission intends to filter out cases of minor importance. 126 It has never been seriously argued that this Notice is a binding interpretation of Article 101(1) TFEU (and more precisely of the Völk ruling 127 ). Because it is not, national courts and authorities are entitled to deviate therefrom. 128 The Commission, on the other hand, is expected to follow the approach enshrined in it. Accordingly, undertakings can expect that the Commission will not start proceedings against an undertaking in instances where the De Minimis Notice suggests that the practice in question is unlikely to appreciably restrict competition.
The binding nature of the Guidance Paper, mentioned above, has been frequently discussed in recent years. As already pointed out, there should be little doubt that this instrument is not a statement of the law. Its role is confined to the definition of the Commission's enforcement priorities when applying Article 102 TFEU. The orientation it provides, on the other hand, is particularly conducive to bringing legal certainty to dominant undertakings. Generally speaking, the Paper reflects a commitment to intervention when potentially exclusionary conduct is likely to foreclose equally efficient competitors. 129 What is more, it provides relatively detailed indications regarding how it intends to go about assessing the anti-competitive effects of a number of practices. In particular, the Guidance commits to prioritising cases involving the use of conditional rebates in light of the so-called 'as efficient competitor' (or 'AEC') test. 130 Accordingly, it would be reasonable for a dominant undertaking to conclude, following a self-assessment of the potential impact of its behaviour based on the criteria laid down in the Guidance Paper, that action by the Commission cannot be expected. If the administrative authority decides to depart from the soft law instrument, one would assume that it must at least set out the reasons why it is justified to do so in a particular case. By the same token, failing to explain the rationale behind its decision to deviate from a stated policy position would entail the annulment of administrative action. The Intel judgment appears to provide support for this interpretation. The Court noted that the Commission had chosen to apply the AEC test enshrined in the Guidance Paper to evaluate whether intervention was consistent with an 'effectsbased' approach to Article 102 TFEU. As a result, a review court controlling administrative action cannot simply ignore the choice made by the authority. 131 In its renvoi judgment, the GC also suggested, in line with the above, that the Commission is bound, at least in principle, by the policy statements it makes in soft law instruments like the Guidance Paper. As already explained, the ruling noted that, as a matter of law, the administrative authority was required to consider the coverage of the practice, and, to the extent that the Commission had failed to do so, its decision was vitiated by an error of law. Crucially, the GC added that this omission by the administrative authority was inconsistent not only with the Court's interpretation of Article 102 TFEU, but with the policy statements enunciated in the Guidance Paper. 132 Having applied the principles enshrined in that document, the Commission cannot subsequently argue that they are not relevant in the assessment.

B. Policy Positions and Non-legal Expertise
The cases that followed the Court's judgment in Intel hint at another way in which soft law instruments containing policy statements may play a role when administrative action is reviewed. Even though it is not a binding statement of the law, the EU courts have cited, with approval, the Guidance Paper. In Servizio Elettrico Nazionale, for instance, the Court referred to the document in support of the claim that evidence of the actual impact of the practice that has been carried out over a certain period (or its absence) is a relevant consideration when assessing its potential anti-competitive effects. 133  made several references to the Guidance Paper, including in relation to the notion of anti-competitive effects 134 and the AEC test. 135 Even when not cited expressly, it would appear that the principles enshrined in the Guidance Paper have progressively found their way into the law. Whenever the case law was at a crossroads, the Court has expressed a preference for the approach advocated by the Commission in its policy instrument. In particular, the Court has consistently held, since Post Danmark I, that Article 102 TFEU is only concerned with equally efficient rivals. 136 The departure from the market of rivals that are less attractive in terms of, inter alia, prices, output, quality and innovation would be a natural expression of competition on the merits. 137 By the same token, any anti-competitive effects must be attributable to the dominant firm for Article 102 TFEU to come into play. 138 The influence of the Guidance Paper (and the ideas contained therein) should not come as a surprise. Often, soft law instruments encapsulate the mainstream consensus and adapt it to the needs of legal enforcement. The Guidance Paper is not an exception in this regard. This document is, in fact, the culmination of a process that started with the preparation of an expert report outlining the key elements of an economic approach to the enforcement of Article 102 TFEU. 139 Since the Court has always relied on this sort of expertise to define the meaning and scope of legal provisions, it is only natural that soft law instruments directly inspire its lawmaking function and are relied upon when controlling administrative action. In this sense, documents such as the Guidance Paper are effective to keep the Commission's commitment to expertise under check.

VII. CONCLUSIONS
Effective judicial review is an indispensable ingredient of any competition law regime. It is, first, a necessity from a fundamental rights perspective. In addition, it fulfils a precious function to ensure that administrative action is clear, predictable, consistent and based on the best available evidence. Directly establishing whether judicial review is effective-or has been effective in a particular case-is, however, a difficult task. This article approaches the question in an indirect way, that is, by identifying the techniques that the EU courts have developed over the years and that allow for the meaningful review of administrative action. These techniques can be grouped into three main categories: those that relate to the interpretation and application of the law; those that relate to the role of expertise and context; and, finally, those that have to do with the Commission's policy positions.
The first conclusion to draw from the exercise is that the Court has a marked preference for clear legal tests that are structured around a well-defined set of criteria. It is an approach to the interpretation of provisions that ensures that judicial review can take place. This technique would be devoid of meaning, however, if the legality of administrative action were not checked against the same set of criteria over time (and the Commission were thus allowed to depart from established tests). Inter-temporal consistency is, accordingly, a pre-requisite for the effectiveness of judicial review. The stability of the law is all the more important if one considers that administrative authorities may be naturally inclined to depart from established tests where these significantly reduce the scope of policymaking. The legal uncertainty created following Irish Sugar and Compagnie Maritime Belge illustrates this point well.
Second, effective judicial review demands that intervention be grounded on the best available evidence. There are two dimensions to this hallmark. One dimension has to do with the role of the mainstream consensus. Administrative action can only be meaningfully controlled if it follows the established body of expertise in relation to non-legal matters, in particular economic analysis. Allowing the Commission to choose the expertise on which it relies would amount, in effect, to giving it discretion on issues of law. The second dimension has to do with the economic and legal context. Administrative action that fails to consider it, in full, will be quashed. Similarly, the Commission is required to engage meaningfully with the arguments raised by the parties and pertaining to the context of which the practice is a part.
Finally, the policy positions expressed by the Commission in Guidelines and similar instruments are not devoid of consequences. If an administrative authority declares that it will exercise its discretion in a particular way, it is reasonable for undertakings to expect that subsequent action will be consistent with its policy positions. One would expect, at the very least, an explanation of why it is appropriate to deviate from the stance previously expressed. Where the Commission departs from its policy statements without providing reasons as to why it is doing so, administrative action can be expected to be quashed. An analysis of the case law suggests that soft law and other instruments are valuable for another reason. As embodiments of consensus positions, soft law instruments have been cited by the EU courts in support of their conclusions.