Thread by @CompetitionProf

What happened yesterday in @EUCourtPress Intel v Commission, T-286/09 RENV? Hot Takes. For antitrust geeks only 1/n
bloomberg.com/news/articles/…

The General Court (GC) was asked for 2nd time to review legality of an EUCommission setting a > €1 Bi fined against chip maker Intel for abuse of dominance. 2/n

When GC looked the 1st time, it gave the EC a satisfecit. The case was appealed to the upper EU Court (CJEU), which appeared more uncertain 3/n

The CJEU suggested a diff analysis should have been applied to test the robustness of the decision. It resent the case to GC, for a new review under a diff test. More on this later 4/n

Now, the significance of the GC 2nd judgment of yesterday must be approached in legal and econ context. Some words on the history of the case 5/n

The EC Intel decision was 1 of these trendsetting cases, as correctly pointed out by @aoifewhite101. 1st time fine > 1 Bi €. More important 1st EC decision to sanction abuse of dominance in application of new policy of “effects-based” - read economics-turbocharged- analysis 6/n

Those old enough will recall a day when the law of article 102 law was a monument of formalism, rife with presumptions, quick look analysis, and even per se rules in some cases 7/n.

In those times, the EC used its discretion to pick cases carefully, and discipline the excesses of the law. But this came at the cost of huge legal uncertainty 8/n

Besides, there were few judicial annulments (and interest for facts in competition cases). The EC enjoyed a lot of leeway to police dominant firm conduct (regardless of what today’s DMA advocates claim) 9/n

In end 2000s, the EC initiated brave policy effort seeking to raise legal predictability and economic literacy in article 102 law. The direction of travel was clear => focus law enforcement on effects, not on abstract structural conjectures about dominant firms’ conduct 10/n

What came out of this initiative is a Guidance paper on exclusionary abuse. The paper embodies economics-minded tests, standards, and measurement indicators for establishing unlawful conduct under Article 102 TFEU 11/n

The Courts, meanwhile started to gradually move their case law in the same sense, that is twd more effects analysis. The CJEU Grand Chamber ruling in Post Danmark 1 stands out 12/n

But some judgments were still rehashing old formalistic drivel for the evaluation of abusive conduct, like Post Danmark 2 or TeliaSonera 13/n

Guidance paper also lost sometraction within the EC. With eyes on ‘winning’ cases as 1stvariable in preference function, Legal Service of the EC fought tooth and nails against the econ approach. No support inside the house. No mention in pleadings. Guidance Paper ghosted 14/n

Case teams in COMP subject to performance constraints were also finding the more economic approach a hard burden in practice, incompatible w/ demands for swift enforcement and the legitimate pursuit of a successful career track 15/n

At same time application of Article 102 TFEU in practice remained awfully discretionary, unpredictable and rigid. Oligopoly firms involved in price cutting campaigns - a prima facie boon for consumers - struggled to know what to do and what not 16/n

Now back to the case. Some facts. Intel, dominant in computer chips, had given rebates to PC and laptop manufacturers conditional on them meeting purchasing targets, volumes, etc. EC said A-BU-SE. Blood in the water => AMDs’ big n2 to Intel 17/n

In other words, the case was abt legality of financial incentives nudging customers to single source chips requirements 18/n

With this, lawyers will recall that main contribution of CJEU 2017 judgment was a clarification of past case law in regard of “exclusivity options”, which the Court deemed not caught fully by the quasi per se rule applicable to exclusivity obligations 19/n

Some clear principles stood out from CJEU => domcos can lawfully exclude rival by superior efficiency; EC can rely on case law presuming abuse, but if domco argues econ in defense it must engage; econ analysis focused on 5 data points is required when domco raises defence 20/n

In the case in dispute, the EC had, in addition to applying the old case law presumption agst rebates, reviewed Intel’s exclusivity options through the lenses of a methodology called the AEC test for “as efficient competitor” 21/n

What is the AEC test? It’s not that complex. The intuitive idea is this => a rival fighting to grab the fraction of orders that is not covered by the dominant position must bring fresh money to customers to compensate lost financial incentives 22/n

Or even simpler: I like my barber but I like my wallet more. I will change barber, if new barber in town compensates the free cut incumbent barber promised after 10 visits. Now, if new barber must freely serve all clients to take business away from incumbent it’s very costly 22/n

It may even be that an equally or more efficient barber will not be able to cover costs. Example has imperfections but you see the point I am sure 23/n

Now in Intel the EC had actually done more than AEC. It had also applied a presumption of illegality stemming from old case law. Result? The decision was weird beast. V long and implying that AEC was just an aesthetic exercise that was not material to the finding of abuse 24/n

At any rate, the EC had found abuse and fined Intel 25/n

Ok over to the GC judgment now => what do we learn from the judges? 26/n

One. The Commission cannot dispense w/ an effects analysis in rebates cases, even in those when the presumption of illegality set in 1979 Hofmann La Roche case appears applicable. Para 124. There’s no rebate that is, by very nature, abusive. Para 145. 27/n

Two. If an antitrust defendant raises a defence of lack of effects - which it can be assumed will happen in 99% of cases - the EC *must* review effects in light of five filters. Para 125. 28/n

Three. The EC must NOT run an AEC test in any and every rebate case (open Q: just in some?, which ones?) but if it runs one, the EC cannot disregard its prediction by falling back on a legal presumption of abuse. Para 126.29/n

But there’s more. If the EC does an AEC test as a complement, it is accountable for it. And this even if had found the test was not strictly required to find abuse. Para 148 30/n

How to understand this? The Court says => If the EC ran the test, one can assume it wanted to complement its findings w/ further facts, and reach a higher level of confidence on existence of abuse or absence thereof. 31/n

Implication? A finding of abuse w/o AEC signals a case based on a relative more probabilistic theory of foreclosure. 32/n

My take? Probabilistic abuse theories can fly as a matter of law (GC confirms). No AEC needed in all cases. But such cases might not deserve the kind of high sanctions and fines (and remedies) that one can see in some cases. Proportionality matters. 33/n

Next. The other main feature of the case worthy of mention is that the GC gives a striking demonstration of thorough judicial control, scrutinizing in depth the facts and the economics of the EC decision. Clearly, we’re not in deference (Chevron, US friends) territory here. 34/n

So we see the Court disagree w/ the 7.1% number advanced by the EC in regard of the disputable part of Dells orders open to competition. 35/n

The Court challenges the periods over which the EC based its analysis in regard of HP’s rebates 36/n

And the Court disagrees that some portions of the incentives schemes given to NEC were conditional on exclusivity. 37/n

This is based on a thorough review of the documents in the file, plus reports of the parties’ economic advisers (Shapiro and Salop advising Intel - TBC) and I guess the CET for COMP. 38/n

The Court also faults the EC for its application of the AEC test => the test asks if a hypothetical as efficient competitor could match the rebates of the dominant firm, not whether this competitor exist. Para 439. 39/n

The fact that no such competitor exists today is not predictive of foreclosure under the AEC test 40/n

Then, regardless of AEC analysis, the Court looks at allegations that the EC had not looked at the 5 filters deemed relevant by the upper Court for the evaluation of rebates 41/n

The filters are cumulative. The EC did not correctly review 1/ the market coverage of the rebates; and 2/ their duration and amount 42/n

This is a problem bc the 5 filters approach is the default standard w/ or w/o an AEC analysis, as long as the domco had raised a defence 43/n

Plus, the EC has adopted that approach in its own Guidance Paper. In other words, the Guidance paper should produce some binding effect on the EC. Para 499. 44/n

Take away: even if the AEC and Guidance paper did not exist, the EC would remain subject to a demanding 5 filter standard in its analysis of rebates. And the Court hints that the 5 filters standard can require a quantitative analysis 45/n

Said differently, the 5 filters standard is not a quick look or quasi per se rule that allows one to shortcut the burdensome AEC test. It’s a methodology that looks at different data points, that might be more or less appropriate depending on the particular case 46/n

For example if cost data is hard to collect or disaggregate, it might be more appropriate to do the 5 factors analysis. 47/n

Now I close. Full annulment. First time since 1979. All fine annulled. EC’s track record bf Court remains solid. But sure implications for pipeline of 102 cases => Qualcomm, Android, etc 48/n

Judgment will embolden DMA apologists, who will take it as one more piece of evidence that the system is broken 49/n

I recommend they go read the judgment. We see here the kind of approximations and mistakes that a well-meaning admin of skilled experts can be driven to inadvertently overlook when it enforces a law based on wooden rules that do not pay due attention of facts, just form END/

#antitrust