Chillin'Competition

The U.S. Supreme Court has issued today its Opinion in the AMEX case. It is available here. This is likely to be the most discussed antitrust opinion in the decade.

The lesson in a nutshell. Showing an apparent restriction on one side of a multi-sided platform is not sufficient to meet the burden of establishing prima facie anticompetitive effects. The plaintiffs’ argument “wrongly focusses on just one side of the market”. EU readers will wonder whether this was not already settled, well that is because…

EU Courts got there first. Back in December 2017 I outlined the main lessons flowing from EU case law regarding multi-sided platforms (see here). In that post I wrote “my bet is that when the SCOTUS rules on Amex (a case which some expect will provide us with quasi divine guidance), it will say nothing new as regards what we already have in EU case law”. And then in February, I insisted that the case “is important, but it could not be more straightforward in the light of the current EU case law. Indeed, the issues to be decided upon in that case are exactly the ones that EU Courts have perfectly understood”. Today’s Opinion fits exactly with the ideas developed in those posts. The merit obviously isn’t mine, it’s the ECJ that got it right and got there first. Actually, I think the issue continues to be better solved in EU law than in this case. In the EU we don’t need to hold the existence of a single product/ relevant market to factor in multi-sidedness; considering it under the relevant economic and legal context is arguably a more refined way to get to the right result.

Identifying multi-sided platforms. The Opinion makes sure to explain that not every market exhibiting some indirect network effects qualifies as a multi-sided platform. It relies on established economic literature and, in line with it, clarifies that a market is to be treated as multi-sided when indirect network effects are strong and don’t just go in one direction so that greater output depends on striking the balance between the different sides (“interconnected pricing and demand”).

Burden-shifting. At what stage is multi-sidedness relevant? All parties agree that the restraints should be assessed under a 3-step rule of reason analysis applicable to vertical restraints. In the first step, it is for the plaintiff to show a substantial anticompetitive effect in the relevant market. In the second, it is for the defendant to bring forward a pro competitive rationale. In the third, the plaintiff must show that there were less restrictive alternatives to attain the procompetitive effects. The transcript of the hearing reveals that all parties considered multi-sidedness to be relevant, the question was whether these considerations belonged to the first or second step. In other words, it was all about who bears the burden of proof (i.e who benefits from uncertainty). The SCOTUS makes clear that this is a question to be dealt with upfront by the plaintiff. As the CJEU had already established, multi-sidedness is part of the economic and legal context against which a restriction must be established in the very first place, when the burden still lies with the accuser.

Business model and inter-brand/platform competition. The Opinion notes that Amex competes with Visa and MasterCard “by using a different business model” (i.e. providing better rewards, which requires a continuous investment funded by charging higher merchant fees). The SCOTUS considers that this business model has “stimulated competitive innovations” in the industry, but creates frictions with merchants, who have incentives to “steer” consumers to pay with other cards. The restraints at issue were precisely Amex’s antisteering provisions used since the 1950’s and which the Court considers “necessary to maintain cardholder loyalty”.

On output and inter-platform competition as the relevant criteria. The Opinion confers great importance to the fact that “while these agreements have been in place, the credit-card market experienced expanding output and improved quality”. It also observes that Amex has made payment services “available to low-income individuals who otherwise could not qualify for a credit card and could not afford the fees that traditional banks charge”. It also underlines that the provisions have not stifled but “promote[d] inter-brand competition” and did in no way prevent rivals from promoting their broader acceptance. In my mind, those are indeed the right elements to look at, just like they are for other intra-brand vertical restrictions. In fact. this has a lot to do with what we recently discussed regarding McDonalds. In the words of the Opinion, to establish relevant anticompetitive effects the plaintiffs should prove that the provisions “increased the cost of credit-card transactions above a competitive level, reduced the number of credit card transactions, or otherwise stifled competition in the credit-card market”.

On causality/attributability. The plaintiffs alleged harm was that the high price of Amex’s merchant fees. But the Court rightly observed that rivals’ merchant fees also increased “even at merchant locations where Amex is not present”, suggesting that the explanation may lie not so much in a restriction but “rather [in] increased competition for cardholders”. This is a sound analysis that again reveals how looking at inter-platform competition produces more accurate results than isolating silos.

The phrases that encapsulate it. “Amex business model has stimulated competitive innovations in the credit-card market, increasing the volume of transactions and improving the quality of the services. Despite these improvements, Amex’s business model sometimes causes frictions with merchants”. Once again, I find it quite hard to believe that Justice Thomas wrote something with which I can agree. It has become populistic popular to argue that multi-sided platforms are “gatekeepers”. In reality, all of the ones we typically think of have actually spurred output and competition. If third parties somehow depend on them, or if frictions exist, that is often because they have created opportunities for third parties in the first place. This doesn’t mean specific restrictions cannot be challenged, but one needs to show how they are severable from the system that enables the opportunities/the greater output in the first place (ah, the counterfactual! See point 6 here)

Why this doesn’t immunize multi-sided platforms. Proponents of the new antitrust revolution have been quick to criticize the Opinion for “immunizing multi-sided platforms from antitrust scrutiny”. EU case law again offers two killer arguments against this interpretation:

• Remember what happened after the CJEU ruled in Cartes Bancaires? Under the clarified/new framework the EC was still able to win after the case was sent back to the GC. This is therefore not about immunization; it’s about following the right analytical framework, without shortcuts.
• Vertical restraints (which is what anti-steering provisions are, according to all parties) have long been held to pretty much the same standard now applicable to multi-sided platforms, and for the very same reasons. It is not enough to show an apparent reduction in intra-brand competition if it spurs inter-brand competition. See cases like Pronuptia or Metro. And has this resulted in immunizing vertical restraints? No, it has only structured the analysis to adapt it to reality. Good cases remain possible.

A 5-4 vote. The 5-4 vote, and above all, the transcript of the oral hearing reveals that some misconceptions were not entirely dispelled. That is unfortunate, particularly because I generally tend to agree with the Justices who are on the dissenting side in this case. When time allows we will write another post commenting on the dissenting opinion and on why I think it gets it wrong, at least under EU law standards. Some may think the 5-4 vote suggests that this is a political decision. The convergence with EU case law would suggest this is not the case.

A very timely Opinion. For many reasons, this is a very timely Opinion. In 5 days I was supposed to hand in an article for The Antitrust Bulletin dealing precisely with case law lessons for antitrust enforcement in multi-sided markets! Fortunately, since my second son was born just last week (now I need to babysit another Pablo 😉 ) and summer looks busy,  the editors have graciously granted me an extension until late October. The thoughts in this blog post will be expanded then, also to incorporate this case and reflect any comments you might have.

Disclaimer. I had no role in this case and don’t work for Amex. I do work for a number of multi-sided companies that will like this Opinion, and also for a bunch of not multi-sided companies that may not. The views in this post are the same ones I have been holding for the past 4 years including here, here,here and here.

For many reasons, 2018 has been an eventful year. For the two of us. Some of these reasons are professional. You will not be surprised if I told you that I am immensely proud to have been promoted to a chair at LSE. And (it’s been in the making so long that it feels unreal) my book is coming out in a couple of weeks.

Alfonso and I felt these two landmarks gave us a wonderful excuse to have an event in London, at LSE Law (by Lincoln’s Inn Fields) – when the weather is still nice (well, fingers crossed) and classes have not yet started.

The idea is to bring some friends and colleagues and discuss some of the major themes I explored in my book. We will have three panels. And for a change, some non-competition lawyers will come along to provide a different perspective.

The event will take place in the afternoon of 14 September – in case you’re wondering, and you’re not based in London, it’s a Friday (and thus hopefully easier for those living or working elsewhere).

Marc van der Woude, Vice-President at the General Court and eminent competition lawyer, has been kind enough to accept our invitation to deliver a keynote speech.

In addition, the event will feature the following luminaries:

• Paul Daly (University of Cambridge)
• Michal Gal (University of Haifa)
• Damien Gerard (European Commission)
• Eric Gippini Fournier (European Commission)
• Carol Harlow (LSE)
• Andriani Kalintiri (LSE, and soon City Law School)
• Bill Kovacic (George Washington University and King’s College London)
• Alfonso Lamadrid (Garrigues)
• Ioannis Lianos (University College London)
• Joana Mendes (University of Luxembourg)
• Jorge Padilla (Compass Lexecon)
• Denis Waelbroeck (Ashurst and Université libre de Bruxelles)

Please watch this space for news about the programme and about how to register. We look forward to seeing many of you there for this occasion.

On 20 April, the Florence Competition Programme organised a workshop on EU competition law after Intel and Cartes Bancaires. The organisers managed to put together a great (and fun) line-up. The keynote speech was delivered by Svend Albaek, Deputy Chief Economist at DG Comp. He shared his personal views on the notion of restriction by object. The speech was really interesting and thought-provoking. I asked whether it would be possible to publish it on Chillin’. Alfonso and I were really pleased that the answer was yes. We leave you with Svend:

In this short comment I will explain why I think that after working more than 18 years in DG Competition I still don’t really understand what the case law says on object restrictions. I probably have to give a bit of warning first. When I once tried to get help on this topic from a well-known law professor, his answer was something along the lines of “oh, you economists always want everything to be so logical; that’s not necessarily the way things work”.

So yes, I may be damaged by my training, but I really would like to see a clear logical construction underlying the characterisation of object restrictions. What I will do in this comment is to present one possible such logical construction and then point out where I think this construction does not entirely fit the way object restrictions are presented in the case law.

So what is this construction? Well, we could call it “per se light”, in the sense of something similar to the US per se concept but with the possibility – and I stress possibility for reasons that will become clear later – of an efficiency defence under Article 101(3).

The idea would be that an object restriction is a practice that (in an admittedly slightly loose formulation) looks like it (1) could plausibly be anti-competitive and (2) could never – or almost never – have pro-competitive countervailing effects. That is, it is not a plausible source of efficiency gains. If this is the case, we would (almost never) condemn a practice that has a net pro-competitive effect. In the worst case, we might attack something that only has a negligible anti-competitive effect.

Probably many readers will know an article (“Defining ‘By Object’ restrictions“) that my good friend and colleague Luc Peeperkorn published in 2015. What I’m thinking of is pretty similar to what Luc suggests, and I was pleased to find out recently that also Pablo Ibáñez and Alfonso Lamadrid think along similar lines (“On the notion of restriction of competition: what we know and what we don’t know we know“). I apologise if there are more people out there that should have been mentioned but whom I just don’t know about because I haven’t had the time to make a proper literature search. However, as you can understand, I’m certainly not trying to claim originality here when I say “my” construction. You will find much of what I write here stated in more detail – and probably better argued – in the articles of Luc, Pablo and Alfonso, so I encourage you to read them if you are interested in the topic.

We can, of course, find something pretty similar to this view in certain Commission documents. And, importantly, we can find traces of it also in Cartes Bancaires, most clearly in Wahl’s Opinion’s in paragraphs 55-58. However, it is, in my view, not spelled out as clearly as Luc, Pablo and Alfonso do.

This construction sounds very much like the US approach to per se infringements, only that there the parties cannot invoke efficiency arguments – or at least not in the lower courts – while in Europe this is still possible under Article 101(3). But the Commission does say in its Guidance on object restrictions that “practice shows that restrictions by object are unlikely to fulfil the four conditions set out in Article 101(3)”. So in reality, it sounds like we would be pretty close to a per se approach. Thus my expression “per se light”.

Now, this may sound fine and logical – perhaps also to the reader – but, unfortunately, I don’t think it reflects reality. Or at least not perfectly. Or at least not yet, if one thinks that the ECJ is in a – probably very slow – process of changing the case law.

I see two major problem areas, which I should say that Pablo and Alfonso also identity in their paper where they call them “outliers”.

The first is with respect to what we could call “internal market partitioning cases”. It is generally recognised – both in economics and in competition law – that territorial restrictions in distribution agreements can bring efficiency gains. If that is true, probably also absolute territorial protection can bring efficiency gains, and it should then not be classified as an object restriction according to my “construction”. The distinction between active and passive sales used in the case law does not really fit with this way of thinking. Something else is clearly going on. Pablo and Alfonso put it this way:

“The peculiar legal status of absolute territorial protection and other practices aimed at limiting trade between member states is explained by the fact that market integration is an autonomous objective of EU competition law”.

So basically, I think I can forget about trying to fit this into my neat logical construction.

The other outlier is Resale Price Maintenance. Many economists think there can be efficiency explanations for almost all – if not all – vertical restrictions and that it is therefore wrong to treat these as object restrictions. And many economists would without much hesitation include RPM in this way of thinking. Now, you may know that Luc Peeperkorn strongly believes that RPM should be an object restriction. So when I had read his article on object restrictions, I immediately went to his office to tease him by saying that he had just argued very eloquently for why RPM should not be an object restriction, since obviously there could be efficiency reasons for imposing RPM; for instance, avoiding free riding by retailers on service provided by other retailers. Luc’s answer was that he does not agree that RPM can be expected to lead to appreciable efficiencies and fulfil the conditions of Article 101(3) since (1) although RPM provides the retailers with a higher margin, it does not take away the free riding incentive; (2) it can thus not be expected that the retailers will voluntarily, without detailed monitoring and disciplining, spend this extra margin on free-rideable services such as product promotion; which (3) implies that there will normally be more effective and less restrictive means of achieving the same efficiencies. I’m not entirely convinced that this is true; consequently, I’m not convinced that RPM being an object restriction fits into my construction either.

Next I would like to ask a couple of questions. First, can the finding of an object restriction depend on the market power of the parties? The ECJ seemed to say so in Allianz Hungária, only to backtrack in Cartes Bancaires. And it would indeed not make sense according to my construction. Because my construction is based on categories of behaviour; I think this may be what the ECJ means when it uses the expression “types of coordination” in Cartes Bancaires (e.g. in paragraph 50). In my view, Wahl’s Opinion also strongly hints at this when he writes that “only conduct whose harmful nature is proven and easily identifiable” should be regarded as an object restriction. So the idea would be that one can easily see that a certain conduct falls into an “easily identifiable” category (e.g. price fixing or market sharing) and therefore conclude that something is an object restriction. But, in my view, that a restriction is a plausible source of efficiency gains cannot depend on market power. It may well be that the result of the ultimate balancing will depend on this, but that would imply that we may expect an effects analysis to end up with a negative (net) result if the parties have (a lot of) market power and with a positive (net) result if they don’t. The conclusion should not be that we think a category of conduct is an object restriction if the parties have market power but not if they don’t. In parenthesis, this way of thinking also made it hard for me to understand the way the General Court in paragraph 89 of its Intel judgment dismissed the relevance of Delimitis. But it seems that the ECJ has now cleared this up in its own Intel judgment.

Finally, a question related to efficiencies and object restrictions. I have a couple of times heard Richard Whish argue that competition agencies and courts should show more willingness to use Article 101(3). I have a lot of sympathy for that. But Prof Whish then pointed to decisions by the Singaporean competition authority concerning airline alliances as a kind of model to follow. As far as I understand – and I must admit that I have not read the decisions – the authority found that the alliances were object restrictions; I guess because the alliance partners would be coordinating on several parameters of competition. But the authority then apparently went on to recognise efficiencies under the equivalent of Article 101(3) and in the end found that the alliances were not anticompetitive. As I wrote, I understand that Prof Whish sees this as a positive example to learn from. But in my construction this example would not make sense. Although, of course, the formal right of the parties to have a go under Article 101(3) is always there, one would not expect to clear something that is an object restriction through efficiencies – and certainly not several times. If one does identify substantial efficiencies (several times), one would rather reach the conclusion that the original categorisation as object restriction likely was mistaken, analyse the effects of the restriction, and reach the conclusion that the net effect is positive and the restriction therefore not anticompetitive. However, as Prof Whish is Prof Whish and I’m an amateur compared to him, my tentative conclusion from this discussion is that my construction may not reflect the real world.

But if so, what is the alternative? I hope that one day I will understand this better.

-It has been reported this week that the European Commission is getting ready to take a step that could have profound implications on the internet as we know it. We did not know whether to comment on this development or not given my conflict of interest, but the Commission’s view would appear to put at risk what this blog stands for as well as part of my work during the past two years. Yes, you guessed right, legal memes are allegedly at risk; for more, see here.

-Yesterday was a historically surreal day. But contrary to what has been reported, Trump and Kim Jong-Un had interacted before (on Twitter, regarding antitrust, and Commissioner Vestager was involved). Click here to see the screenshots (from slide 3 onwards).

-In what many see as a blow to Trump, the AT&T/Time Warner merger was unconditionally cleared yesterday after the DOJ’s suit was dismissed. The full text of Judge Leon’s Opinion is available here. Its drafting seems to be quite specific and not a challenge to the growing concern about vertical mergers. The Judge himself has stated that “the temptation by some to view this decision as being something more than a resolution of this specific case should be resisted by one and all”. This development will trigger many comments, but the most succinct, persuasive and full blown attack against vertical mergers we have read so far comes from 30 Rock and is available here  (make sure to click) 😉

-Moving on to more serious news, the Luxemburgish competition authority has exempted an algorithm price fixing mechanism for taxis. The Decision available here (in French) notes that the joint use of a multi-sided platform that fixes prices constitutes a horizontal agreement, but concludes that the agreement shall be exempted given the efficiencies generated by the agreement (which include lower prices for consumers thanks to the algorithm) and the absence of any viable alternative to attain them. Would the Commission and other NCAs agree with this view? We don’t know because the case concerns only Luxemburgish law (otherwise an exemption would not have been possible pursuant to the Tele 2 Polska case law). According to the decision, the taxi sector is subject to national regulation and does not impact trade between Member States (which arguably doesn’t fit squarely with what the CJEU held in paras. 65-70 of Eventech). Since Member States can’t exempt agreements under EU Law, they may be led to adopt this sort of jurisdictional interpretations, which may in turn not be ideal for legal certainty and for the internal market.

Wouldn’t it be nice if the European Commission decided to adopt 101(3) decisions too? Quizz question: what was the last 101(3) exemption granted by the Commission?  It’s already been 7 years (!) since we wrote about The Slow Death of Article 101(3), and it’s not that the landscape has improved.

Other stuff on our plate has prevented us from commenting on many of the competition law developments that took place in the past few weeks. But we know you are hungry for commentary. Without a doubt, the meatiest recent development is the class action accusing McDonald’s of anticompetitive tying.  Don’t assume this is just going to be another post with puns; there really is food for thought here.

Background

The plaintiff challenges McDonald’s distribution of the Quarter Pounder line of burgers. The allegation is that McDonald’s only markets the Quarter Pounder and Double Quarter Pounder with cheese (in the past it also sold defaults without cheese) and only as part of a value meal. The plaintiff argues that this practice forces consumers to “pay for two slices of cheese, which they do not want, order, or receive”.  They estimate that non-cheese-eaters are charged 30 to 90 cents for the cheese they don’t want even if they have to ask for the cheese to be removed (or have to remove it themselves). This would enable McDonald’s to charge for cheese it does not deliver.  The plaintiffs have done their research and noted that the original 1975 trademark does not refer to cheese (only to a frozen beef patty, a sesame-seed bun, one tablespoon of diced fresh onion, mustard, ketchup and two Heinz pickle slices). The full class action complaint is available here.

McDonald’s has replied that “the advertised Quarter Pounder burger comes with cheese. We try to accommodate our customers’ requests by allowing them to customize their orders, such as a Quarter Pounder with no cheese.”

The merits

This is a US case, but imagine you wanted to run a case like this in the EU. Could it fly? What would you do to meet the requirements set by the case law? Is this really a crazy case as it may seem to many? Let’s see:

First, you would argue that the tying and the tied products are separate products. The two were once sold separately by McDonalds; cheese can be added or not; the trademark does not cite cheese; there are independent producers of cheese. At the same time it’s true that the forked versions of the Quarter Pounder do incorporate two slices of cheese (see e.g. here) [yes, there are many different types of McDonald’s forks]. Overall, one could get away with saying that a Quarter Pounder and cheese are different products. Box ticked.

Second, you could argue that McDonald’s is dominant in the tying product. You would only need to argue that McDonald’s franchisees don’t really have the option of buying burgers from any other source. So McDonald’s is dominant in the market for “franchisable burguers to be sold at McDonald’s restaurants”. The plaintiffs refer to McDonald’s market power in the “fast food quarter pound hamburger market” (para. 75). Box ticked.

Third, customers have no option of obtaining the tying product without the tied. That’s the point the plaintiff is making. One is certainly not forced to eat the cheese and to customize the burger further (you can pretty much put anything in there and McDonald’s will let you) but it is true that the cheese comes pre-installed as a default. The plaintiffs in this case also invoke the “uniqueness” and “desirability” of the tying products.

Fourth, you would need to show that the tying is capable of restricting competition. One can argue that McDonald’s is hindering consumers choice. You can even define a market for cheese to be used in McDonald’s Quarter Pounders and conclude that rival ingredients are being excluded (even if there is no technical obstacle for users to add other ingredients to be added to the Quarter Pounder; McDonald’s even facilitates that: see e.g. here).

Piece of cake. The plaintiffs wouldn’t even have to argue that they are overpaying, as they do in this class action. I guess they could make the same allegation even if McDonald’s provided the burger and the cheese for free.

[As this is a US damages claim, the plaintiffs also need to prove that they suffered injury as a result of their purchase. That may be easier to show, but I doubt the injury I have in mind could be attributed to the alleged tying…]

The lessons

One can argue anything in a competition case (see here for some example on market definition), but most of you will –hopefully- agree that the case makes little sense. But let’s try to slice and reason that intuition:

Franchise business model necessarily implies a package of interrelated items. Whilst tying is assumed to be restrictive (because it is presumed that it serves no procompetitive goal), franchise agreements are presumed procompetitive and justify even what would otherwise be seen as hardcore restrictions (non-compete clauses, outright exclusivity; see e.g para. 191 of the Vertical Guidelines). How do we then square this out? It all depends on the issue of severability. This was made clear by US Courts in a case involving… alleged McDonald’s tying (Principe v McDonalds; a case concerning licensees obligation to operate their franchises in premises leased from the franchisor). What is the actual licensed/franchised tying product? Is it just trademark or is it rather a business format?

In that case McDonald’s had argued that “the appellants are asking the court to invalidate the way McDonald’s does business and to require it to adopt the licensing procedures of its less successful competitors” (para. 23). The Fourth Circuit agreed. It observed the question depends on whether the items were “integral components of the business method being franchised” (para. 32).

After analyzing the specific contributions of the obligation at issue to McDonald’s business method (paras. 34-37, which take into consideration protecting the “system’s goodwill”, ensuring “consistent quality”. “broadening the applicant base and opening the door to persons who otherwise could not afford a McDonald’s franchise” as well as the extent of McDonald’s financial investment), it concluded the following:

“All of these factors contribute significantly to the overall success of the McDonald’s system. The formula that produced system wide success, the formula that promises to make each new McDonald’s store successful, that formula is what McDonald’s sells its franchisees. To characterize the franchise as an unnecessary aggregation of separate products tied to the McDonald’s name is to miss the point entirely”.

Severability in Principe v McDonalds was analyzed at the “separate products” step, whilst in in the EU it would be part of the assessment of the economic and legal context (this is true of any restraint that may be part of a wider context, not only franchising; see e.g. Cartes Bancaires 73-79). But the substantive reasoning is pretty much the same. Ignoring the relevant context to any restraint leads to missing the point entirely. Much like in Principe v McDonald’s, EU Courts have also observed that restraints are deemed to fall outside the scope of the competition rules when the such analysis shows that they are related to a main operation that “could not be implemented or could only be implemented under more uncertain conditions, at substantially higher cost, over an appreciably longer period or with considerably less probability of success” (T-112/99, Métropole, para. 111).

This explains why competition experts would find this new class action complaint hard to swallow.