Chillin'Competition

The first two instalments of the series focused, respectively, on the institutional and substantive contributions of the Court of Justice judgment in Android Auto. This third (and final) entry moves to the consequences of the ruling. There are three areas where the ruling may have a significant impact.

Product design and business models after Android Auto

As explained a while ago here, dominant firms’ choices in terms of product design and business models were, for decades (before Android Auto, that is), largely insulated from competition law scrutiny. This is true, for instance, of the degree of openness of an infrastructure. Absent evidence of indispensability, the decision by the operator about which features to open to third parties and which to keep for itself would be generally compatible with Article 102 TFEU.

Similarly, business models (that is, the strategies followed by firms to monetise the value they generate) was typically (if no presumptively) in line with EU competition law. As a rule, a dominant firm could keep for itself a feature of a partially open model as an aspect of its monetisation strategy. Again, this choice would have come under scrutiny only if the conditions set in Magill or Bronner had been shown to be met.

Following Android Auto, however, indispensability is no longer an obstacle to intervention where infrastructure operators (or input providers) decide to open their assets to third parties only partially. As soon as access is given in relation to one potential use of the input or infrastructure, indispensability ceases to be relevant across the board.

This shift in the case law is particularly consequential in the current economic and technological landscape. It expands authorities’ policy space significantly. One implication is that the design of products and the choice of business models may have to be reconsidered more often than in the past.

The analysis of effects becomes particularly relevant in the new legal landscape

The retreat of the indispensability condition strikes a new balance between short-term and long-term competition. It may well be the case that, in dynamic industries, indispensability sets too high a bar for intervention. This, in fact, was the point made by Advocate General Kokott and Mariya Serafimova in a recent piece addressing this very issue.

Whenever a new balance is struck, however, there is always a risk that the pendulum swings too far in the opposite direction. There is a risk, in other words, that the design of products and the choice of business models move from being (almost) presumptively lawful to being (almost) presumptively unlawful. A radical shift may not be the optimal one, whether from a substantive or an institutional standpoint.

In this new legal landscape, the analysis of effects emerges as a particularly relevant step in the assessment of potentially abusive conduct. The effects test will, in practice, act as the primary legal filter to strike the right balance between the static and dynamic dimensions of competition. It is not a surprise, against this background, that the Court stresses its importance in Android Auto.

In line with the preceding case law, the judgment emphasises that the analysis of the actual or potential impact of a practice cannot be a mere formality. As the Court put it, the assessment cannot be grounded in a ‘mere hypothesis’ (para 57). Instead, it must be based on ‘tangible evidence’ informed by the specificities of the case. Similarly, the ‘existence of doubt’ about the capability of a practice to produce actual or potential effects must ‘benefit the undertaking’.

Testing the meaning and limits of Android Auto

It seems likely, if not inevitable, that the meaning and limits of Android Auto will be explored in the coming years at the national and EU levels. With private enforcement on the rise (in particular in the context of Article 102 TFEU), national courts may be confronted with the sort of issues (imposing access obligations, determining the price of access, monitoring compliance) that only arose sparingly before them prior to the ruling.

This, it seems to me, is a third major consequence of the substantive innovations introduced in the judgment. There are many issues that are likely to be tested before national courts. One question relates to whether this judgment is confined to digital platforms. I do not believe it is (the key substantive innovation applies to all sectors, as I understand it), but I am aware that there is no consensus on this point.

A second obvious question relates to what ‘open to third-party undertakings’ actually means. The limits of this concept are not immediately apparent. If an operating system is open to third-party application providers but is not licensed to third-party device manufacturers, does Bronner apply where the latter seek access? Again, there is room for interpretation. For the same reason, this issue (and similar ones) may come back before the Court in the next few years.

The Annual Conference of the Mannheim Centre for Competition and Innovation took place last week. I am very much grateful to the organisers for honouring the legacy and achievements of Professor Heike Schweitzer.

The script of the keynote I delivered can be found here. It focused on the Ordoliberal tradition, which Professor Schweitzer continued, and which I came to appreciate and study thanks to her. Its core insights come across as particularly relevant at this juncture. Ordoliberals’ concern with unchecked (private and public) power and its impact on democracy resonates more than it has in decades.

Paradoxically, the zeitgeist appears to be moving away from what I call the Ordoliberal compact, that is, an approach to policy-making that revolves around an independent authority that is constrained by law and is subject to judicial review (and which is behind the unlikely and resounding success of European competition policy since the 1950s).

This technocratic approach to competition policy is increasingly questioned. An emerging body of literature deems it excessively narrow and (unduly) insulated from the broader political landscape. Similarly, legal constraints on administrative action have come under critical scrutiny.

Counterintuitive as it may sound, moving away from the Ordoliberal compact may weaken competition authorities vis-a-vis powerful private (and public) actors. In addition, it may affect the effectiveness and credibility of enforcement.

I have just uploaded on ssrn (see here) a new paper on judicial review in EU merger control. I am delighted that the paper is forthcoming in European Law Open, a new, fully open access journal run by a group of leading EU law scholars.

In light of recent case law, the paper addresses an eternal debate in EU merger control: whether, and if so how, one can disentangle issues of law (which are subject to full review in the EU legal order) from complex economic assessments (controlled for manifest errors).

The question (which has both a constitutional and a competition law dimension) is whether the two standards of review can co-exist or whether, in practice, one will inevitably exercise its gravitational pull over the other.

It seems to me that this debate is particulaly relevant at a time of technological and economic change. It has been suggested that, in this new reality, the administrative authority would benefit from greater leeway to experiment and take risks to tackle unprecedented substantive issues. The paper discusses the impact that such rebalancing of powers could have on the merger control regime.

I look forward to your comments on it (as usual, nothing to disclose).

The first instalment of this series on the judgment in Android Auto discussed the metamorphosis of the case before the Court: a preliminary reference concerning the interpretation of the indispensability condition was turned into an analysis of whether the Bronner criteria were applicable in the relevant circumstances.

Beyond institutional issues, the judgment is notable in that it introduces a number of substantive innovations that represent a move away from Magill (which was the first providing a structured framework to refusals to deal), Bronner (which extended this framework to tangible property) and IMS Health. One may add to this list Slovak Telekom and Google Shopping, which did not depart from (in fact, they embraced) the same approach.

One can identify three main areas where Android Auto carves its own path and reframes the case law. To begin with, the judgment departs from the logic underpinning the legal framework introduced in Magill. Second, it changes the interpretation of a crucial point of law, which has the effect of significantly reducing the scope of Magill and Bronner. Third, it allows for more far-reaching intervention.

Android Auto and the logic underpinning the refusal to deal case law

The rationale underlying the refusal to deal case law and, specifically, the indispensability condition had been clear from the outset, and were effectively summarised by Advocate General Jacobs in his Opinion in Bronner. If anything, the Court has been more explicit about this logic in recent years.

There are two main reasons why Magill, and then Bronner, set the bar so high. The first relates to the fact that imposing a duty to deal on firms amounts to a major interference with their right to property and their freedom of contract. It should therefore only happen, the Court has consistently explained, in genuinely ‘exceptional circumstances’.

The second reason has to do with the consequences that such interference has on firms’ incentives to invest and innovate (that is, on dynamic competition). The fundamental premise behind Magill and Bronner is that dynamic competition is more beneficial for citizens and society than its short-term counterpart. Against this background, requiring evidence of indispensability was deemed to be a reasonable means to ensure that the latter does not impact negatively the former.

The Court moves away from this logic in Android Auto. Indispensability is not an element of the legal test even though intervention in the case amounts not just to compelling a firm to deal with a third party, but even to changing the operation of its platform.

The indispensability condition, therefore, is no longer universally applied as a filter to ensure that intervention does not affect dynamic competition. By the same token, the judgment strikes a new, different balance between static rivalry and firms’ incentives to invest and innovate.

How Magill and Bronner are reinterpreted in Android Auto

The Court, in Android Auto, attaches a great deal of importance to the fact that the platform had not been solely developed for the ‘needs of [Google’s] own business but with a view to enabling third-party undertakings to use that infrastructure’ (paras 44 and 49). Where the platform is a partially open one, the ECJ concludes, the indispensability condition is not an element of the legal test. This is so, the judgment suggests, even when access to ‘third-party undertakings’ has been given for purposes other than the one for which access is requested.

It is sufficient to take a cursory look at Magill to realise how this position departs from the preceding three decades of case law, including Bronner. The TV operators in Magill had enabled ‘third-party undertakings’ to use the programming listings at issue in the case. As explained by the Commission in its decision, RTE, BBC and ITV had not kept their inputs for their ‘own use’. They were already licensing their intellectual property to newspapers, free of charge, when the Commission started its investigation. What TV operators kept for the ‘needs of [their] own business’ was not the input generally speaking but a particular application of the input (the publication of weekly listings, as opposed to daily ones).

In spite of the above, indispensability was found to be an element of the legal test in Magill. What mattered was not whether the TV operators had licensed the listings to ‘third-party undertakings’ in a market unrelated to the one at issue in the case (daily listings), but whether the dominant firm had kept for the ‘needs of its own business’ the activity for which access was requested (weekly listings).

Android Auto introduces a new interpretation of Article 102 TFEU. It suggests that if access is requested for application A, indispensability will not be an element of the legal test when access is requested for application B. Once the platform is partially open to ‘third-party undertakings’, Magill and Bronner will no longer be the relevant framework to assess the abusive nature of a refusal, irrespective of the adjacent market concerned.

The implications of this rereading of the case law are significant. It necessarily reduces the scope of the Magill and Bronner doctrines, which are now confined to instances where the integrated dominant firm has not dealt with rivals in any market or in relation to any actual or potential use of the input or infrastructure to which access is requested.

Crucially, it is not obvious that this new interpretation is limited to digital platforms. It is easy to think of instances where an input or platform can be used for a variety of purposes, and nothing suggests that the new reading of Bronner will vary based on the relevant sector.

This interpretation was not expressly acknowledged or contrasted with the facts at stake in Magill, whether by the Court of by Advocate General Medina in her Opinion. It was treated as a novel point of law in the judgment. This technique reminded me of some recent refinements of the case law in the context of Article 101 TFEU (and discussed here).

How Android Auto goes beyond Magill and Bronner

It is worth mentioning, finally, that the judgment in Android Auto goes beyond Magill and Bronner, in the sense that it allows for a deeper and more significant interference with a dominant firm’s right to property.

In both Magill and Bronner, the Court ruled that there may be circumstances where a firm can be compelled to deal with rivals. Android Auto suggests that a firm may be forced not just to deal with third parties, but to change the operation of its infrastructure to accommodate rivals (and this, without the need to establish that the said infrastructure is indispensable).

The Commission had already suggested expanding the scope of Article 102 TFEU to require firms to invest to adjust their property and accommodate rivals (in particular in energy markets, as in ENI). Its approach is now vindicated. In a sense, in fact, the shift heralded in Android Auto has been decades in the making. I will elaborate on this issue in the next instalment. In the meantime, your comments would be very much welcome.

The “revolution”, “counterrevolution” and most debates around Article 102 TFEU over the past couple of decades have revolved around the notion of an “effects-based approach”. Yet, there remains considerable uncertainty as to the meaning of “effects” in abuse of dominance cases. This is remarkable given the Court’s clarification that a finding of abuse presupposes that the conduct at issue “at the very least, is capable of producing exclusionary effects” (Superleague, para. 130). Indeed, establishing exclusionary effects/foreclosure is the bottom line in every case. If the notion of effects is too narrow or demanding, effective enforcement may be compromised. If, on the contrary, there are no discernible limitations to the notion of effects, then this paves the way to discretionary enforcement and legal uncertainty. We should arguably avoid either extreme.

In the Policy Brief launching the process for the future adoption of Guidelines on Article 102 TFEU (titled “A dynamic and workable effects-based approach to abuse of dominance”) the Commission acknowledged that “[t]he effects-based approach promoted by the Commission is clearly reflected in [case law developments]” (p.2), but also raised “the question of whether the heightened substantive legal standard that the Union Courts have accorded to it may inadvertently lead to undesired outcomes”, expressing concern that “the bar for intervention [could be set] at a level that would render enforcement against practices that restrict competition unduly burdensome or impossible” (p.4).

The desire to “consolidate the case law” in such a way as to make the effects-based approach more “workable” also inspired the March 2023 amendments to the Guidance Paper on exclusionary abuses. With those amendments, the Commission replaced the notion of “anticompetitive foreclosure” that the Commission had publicly committed to follow since the original 2008 version of the Guidance Paper, and which was consistent with that used under merger control (see e.g. the Commission’s non-horizontal merger guidelines, para. 18). This significant amendment, which implied not only signposting for the future, but arguably also an element of goalpost-moving, received scarce attention at the time.

The draft Guidelines now take several additional steps to alleviate the Commission’s burden of having to establish exclusionary effects:

First of all, it is noteworthy that under the approach set out the Commission would not need to establish anticompetitive effects in cases involving “naked restrictions” (Draft Guidelines, para. 60.c) or conduct otherwise “presumed to lead to exclusionary effects” (including exclusivity agreements, conditional rebates, predatory pricing, margin squeeze and certain forms of tying; Draft Guidelines, para. 60.b)). This reliance on presumptions has already been the subject of much commentary (see, e.g., here, here, here, or here).

But for the purposes of this post I will focus on the analytical framework that the Draft Guidelines propose to follow in relation to the only category of practices that the Commission itself describes as “conduct for which it is necessary to demonstrate a capability to produce anticompetitive effects” (Draft Guidelines, para. 60.a):

If one connects what the Draft Guidelines say regarding capability, causality and appreciability together with their definition of exclusionary effects, my reading is that even in the subset of cases where the Commission would be required to establish effects, the Commission would only need to show that conduct by a dominant firm is capable, considered together with other factors, of potentially increasing the likelihood that rivals (including in some cases less efficient rivals) may find it somewhat more difficult to compete against a dominant undertaking (e.g. by increasing barriers to entry), even if the added difficulty is minimal or not appreciable.

Query: Can anyone think of any conduct outside the scope of competition on the merits that would not meet this test?

I can’t. And if every conduct deviating from competition on the merits (itself a controversial and vaporous filter, as discussed here or here) would meet this test then, arguably, the requirement that conduct must be capable of having exclusionary effects to fall within the scope of Article 102 TFEU would constitute an empty filter.

There are, I suppose, three possible counterarguments to my observation, namely (a) that my reading of the Draft Guidelines is an oversimplification; (b) that the Draft Guidelines are simply and objectively based on established case law; and/or (c) that, in practice, the Commission would in any case exercise self-restraint. So let’s address those: