Relaxing whilst doing Competition Law is not an Oxymoron

Substantive Legal Tests and Standard of Proof: Rules Lost in Translation? (by Andriani Kalintiri)

with 10 comments

Evidence Standards in EU Competition Enforcement

The CK Telecoms (Three/O2) judgment has revived the interest in issues relating to the standard of proof in EU merger control (and competition law at large). It adds valuable insights on the question. As I read it for the first time, I was, yet again, reminded that there is some tendency by commentators to conflate two separate questions: (i) the standard of proof and (ii) the substantive legal test (and more precisely the applicable threshold of effects – capability, likelihood, certainty).

Our friend Andriani Kalintiri (King’s College London) knows these issues inside out, which is why I immediately thought of inviting her to present them in a post. The judgment, by the way, is very much in line with her own take on the standard of proof.

I leave you with her analysis, which hopefully will complement mine. If hungry for more, take a look at her superb book, from which I have learnt a great deal (on evidence, presumptions and judicial review). It is particularly timely now that discussions on presumptions and the burden of proof are all the rage. Thanks again, Andriani!

There is only one way to start this post. First, with a heartfelt ‘thank you’ to Pablo and Alfonso for the kind invitation! And second, with a confession: that the ‘misconception’ this note hopes to illuminate and caution against – i.e. the tendency to conflate substantive legal tests and standards of proof – has been the cause of many headaches during my research! Indeed, in the EU Courts’ antitrust and merger judgments, one often comes across various references to the ‘plausible’, ‘potential’, ‘probable’, ‘likely’ or ‘actual’ effects that a practice or arrangement must give rise to, in order to be prohibited. Are these references, I was wondering, indications of the applicable ‘standard of proof’ or of something else, and why?


To answer this, let me begin with a simple yet important reminder: competition enforcement takes place under conditions of ignorance and uncertainty (as does law enforcement in general). In an ideal world, courts, authorities, and businesses would possess perfect information and would always make correct decisions about the proper meaning of the antitrust and merger rules and their application. In real life, however, information is incomplete and resources, time and our cognitive capacity are limited. Yet, decisions must still be made, and the rules must be enforced.

Among other mechanisms, legal tests and standards of proof are what allows ‘the show to go on’. On the one hand, the legal tests developed by the EU Courts specify the scope of the vaguely worded antitrust and merger rules by setting out the conditions that must be satisfied for a practice to be prohibited – for instance, where the conduct has caused harm or where there is a stronger or lesser chance of such harm occurring. On the other hand, the standard of proof indicates the threshold, falling short of certainty, that the party with the burden of persuasion must surpass for the evidence to be accepted as proof of an allegation.


­In practice, however, these two thresholds are commonly conflated – for instance, it is sometimes said that different ‘standards of proof’ apply to different types of unilateral conduct. Based on what I have said, a key – albeit not the only – explanation for this tendency may be already obvious: that when we think about competition legal tests and standards of proof, we consciously or subconsciously employ probabilistic language – i.e. we use words such as ‘plausible’, ‘potential’, ‘probable’, ‘likely’ and so on, in relation to both of them. For instance, we may say that a practice will be prohibited when it is ‘probable’ or ‘likely’ to foreclose competition. Or that an allegation is proved when it is ‘more likely than not’.

Despite any illusion of unity though, there are important differences between the two. For one, they serve distinct functions in that they answer different questions. The key question competition legal tests are ultimately preoccupied with is: ‘what level of harm is required for a conduct to be deemed unlawful?’ On the other hand, standards of proof address a different issue – i.e. ‘what level of evidence is sufficient for an allegation to be accepted as true in the eyes of the law?’

Most importantly, the considerations that inform their design – i.e. what the substantive legal test and the standard of proof should be – are different, too. As a starting point, both aim to minimise false convictions and false acquittals since either error entails costs. While, however, these costs are not entirely disassociated (at least in the longer term), they are not identical either – and nor are the factors that underpin their balancing. This discussion is particularly complex, but in rough terms the following may be noted.

As far as ‘erroneous’ legal tests are concerned, their primary costs are chilling procompetitive behaviour and encouraging anticompetitive conduct (including in both instances, the respective harm to competition and consumers). Their balancing in designing the ‘optimal’ legal test – and threshold of effects – largely depends on the nature of the conduct in question in the light of current knowledge about it, and the cost of enforcement as well as business compliance. This explains why the substantive legal test varies practice to practice – cartels, for instance, are not treated in the same way as refusals to supply, for which the required threshold of effects is much higher, given the implications for firms’ incentives to innovate.

By contrast, the primary costs of an ‘erroneous’ standard of proof are the undue interference with the freedom and rights of the defendant and of harm to society. Their balancing in designing the ‘optimal’ standard of proof largely depends on fairness considerations linked to the seriousness of the consequences involved for the person concerned and the specific features of the enforcement model, including any inequality of arms. This explains why the standard of proof is ‘static’ – the same standard of proof (i.e. ‘firm conviction’) applies to cartels and refusals to supply, and rightly so, since the consequences for the undertakings involved and the nature of the enforcement proceedings are the same.


This discussion is not purely academic for several reasons, but I will confine myself to two.

First, conflating substantive legal tests and standards of proof may lead to confused discussions about what these are or should be. Two examples illustrate this risk. On the one hand, recent reports on digital competition policy suggested, among others, modifications in the burden and standard of proof; more accurately though, these proposals arguably concern the substantive legal test, and should be assessed as such.[1] On the other hand, in merger control it is often argued that the applicable standard of proof is – and should be – the ‘balance of probabilities’ due to the absence of fines and the prospective nature of the analysis. As I have explained elsewhere though, the case law suggests – and rightly so, in my opinion – a much higher standard of proof. The recent judgment in CK Telecoms seems to confirm this: as the General Court noted, ‘the standard of proof (…) is therefore stricter than that under which a significant impediment to effective competition is “more likely than not”, on the basis of a “balance of probabilities” (…)’, although it is ‘it is less strict than a standard of proof based on “being beyond all reasonable doubt”’ (para 118).

Second, substantive legal tests and standards of proof are equally important to the correctness and legitimacy of enforcement. Indeed, ‘correct’ legal tests may not compensate for shortcomings in the design of the standard of proof. Think, for instance, of cartels – the fact that they are rightly subject to a ‘by object’ prohibition could not ‘save the day’, if the standard of proof were too low – say, equivalent to the balance of probabilities (or similar threshold of belief) given the operation of the presumption of innocence. The opposite is also true: ‘high’ standards of proof may not rectify inappropriately formulated substantive legal tests. Think, for instance, of Google Shopping Service: even if the Commission has successfully established a ‘firm conviction’ that the elements of the substantive legal test – as it understands it – are met in the light of the evidence, this will not mean much, if the latter is eventually found to be incorrect.


Ultimately, it is important to appreciate that substantive legal tests and standards of proof are different rules. Of course, the stricter the substantive legal test, the more difficult it will be to discharge the standard of proof – and vice versa, the stricter the standard of proof, the more difficult it will be to prove in specific cases the constituent elements of the substantive legal test. However, they remain distinct yet equally important for an authority’s decision to be not only lawful but also legitimate. In this regard, the discussion is not about semantics – rather, words such as ‘plausible’, ‘potential’, ‘possible’, ‘probable’, ‘likely’ and so on, may lead to very different perceptions of the applicable threshold of effects or standard of proof, depending on what they actually refer to, and we should thus use them more consciously and accurately to avoid confusion and misunderstandings.

And a final remark, for the avoidance of any doubt: I have no interest to disclose – other than a deep academic fascination with the topic!

[1] For example, in the Report on Competition Policy for the Digital Era, it is noted (p 4) that ‘(…) competition law should try to translate general insights about error costs into legal tests. The specific characteristics of many digital markets have arguably changed the balance of error cost and implementation costs, such that some modifications of the established tests, including allocation of the burden of proof and definition of the standard of proof, may be called for.’

Written by Pablo Ibanez Colomo

8 June 2020 at 5:27 pm

Posted in Uncategorized

10 Responses

Subscribe to comments with RSS.

  1. Excellent post. Illuminated the distinction very clearly and agree with the author that it is a distinction that is often not treated as such.


    8 June 2020 at 5:47 pm

  2. Well-done Andriani. I recommend Andriani’s book to all competition law scholars. By the way, I liked your reference to Google Shopping 🙂
    More seriously, this is indeed an area of total mess and confusion. So much that even the EU legislator has confused things. For example, Art 2 of Reg 1/2003 has confused legal & evidential burden of proof, when referring to Art 101(3). It probably had in mind the evidential burden of proof, yet it refers to the legal one… But that’s now history.


    9 June 2020 at 12:47 pm

  3. Thanks so much for the kind words, Makis and Tom! I am really glad you enjoyed the post.

    Andriani Kalintiri

    9 June 2020 at 9:15 pm

  4. Thank you for this very informative post. I really enjoyed reading it.

    I wonder if this distinction might be also relevant for a discussion going on in Germany at the moment as regards private enforcement. Here, the Federal Court has ruled that there is no prima facie evidence for an, cartel related, increase in price as regards a specific transaction. It, however, also said that there is instead something what the FCJ called ‘positive presumption’ for this.

    When the judgement was delivered, there was some confusion about the difference this distinction actually brings. But maybe it is comparable to the distinction you have described here. There is a burden of proof to provide enough evidence as required for a proper assessment by the court. The assessment itself is then, however, is influenced by the positive presumption.

    I understand that this issue is somewhat difference from the problem discussed in your post, as the identification of specific damages suffered is probably not so much one that is open for a “substantive legal test”. However, at least for the sake of illustration, it might be useful to draw something like an analogy here. Thus, thanks again for the input!


    10 June 2020 at 8:49 am

    • Thank you very much, Philipp! I assume you are probably referring to the rail cartel damages litigation? Without knowing the details of the case or of German law, it seems to me the distinction here might be between optional and mandatory factual inferences. As I understand it, the prima facie evidence rule comes very close to a presumption in the technical sense, inasmuch as it enables the claimant to provisionally satisfy the standard of proof with respect to an issue (i.e. existence of harm as a result of the cartel) by proving something else (for instance, that there was a cartel agreement and the cartelists continued conducting business, including with the claimant, after its conclusion). I have only read commentary about the judgment (I do not speak German, unfortunately) but the FCJ seems to have said that the empirical foundations of this presumption (or prima facie evidence rule in German terminology) are not strong enough to warrant a mandatory factual inference and it should thus be ‘downgraded’ to a (possibly strong yet) not automatic indication of harm that may still be drawn but only upon examination of the evidence and the specific circumstances at hand. That’s really interesting – many thanks for raising this!

      Andriani Kalintiri

      10 June 2020 at 9:50 pm

  5. Thank you for this, Andriani !
    After decentralisation of EU competition law (and many cases being dealt with by NCAs) there is, I think, another important consequence of this. The legal test is to be applied uniformly, also by national authorities, whereas the standard of proof is not harmonised (Eturas, 30-31) even if it will be influenced by the principle of effectiveness and the presumption of innocence, so variations will be ultimately limited. And yes, unfortunately certain “semi-official” documents (let alone academic articles) still conflate the two. I understand they can be seen as related (they are “hurdles” for conviction) and the fact that they are often presented in probabilistic terms may suggest that the distinction is just semantic, but it is not and one should expect conceptual rigour from academia and the enforcing authority.
    Yet, I have the impression that conflating the two notions has also been an “interested” way to change the legal test (if perceived as not setting out a “high enough” hurdle) and/or to give more leeway to the reviewing court: if the legal test is converted into an issue of evidence and proof, the legal test may in fact be changed (or rendered ineffective) by the backdoor and generally the Court may be invited/encouraged to apply a more stringent attitude, as this would be just an issue of evidence (so the “free evaluation” of evidence and the presumption of innocence apply). However, the principle of presumption of innocence plays a role in proving the facts/conduct which are material for the legally applicable test but plays “as such” no role, I think, when defining the legal test (which does not mean that legal tests can be set out which would make everybody guilty…). Whether the legal test is the right one, from a policy/economic perspective, or whether the law is sufficiently clear, are different issues, the latter engaging potentially other fundamental rights.


    10 June 2020 at 11:02 am

    • Many thanks, Juan! Both points you raised are very important.

      With respect to the first one, I could not agree more – this is indeed a further crucial reason why the distinction between substantive legal tests and the standard of proof is not just about semantics: the former are part of substantive EU law that NCAs and national courts are bound to apply, while the latter has not been harmonised and is for the Member States to determine, although in practice the scope of national procedural autonomy may be (significantly) limited by the principle of effectiveness and the operation of the Charter of Fundamental Rights.

      As for the second point, I should say there are different views on the presumption of innocence – some commentators (not in the competition law world, as far as I am aware!) have suggested that it should play a role in determining whether/what conduct should be prohibited in the first place (i.e. at the ‘criminalisation’ stage). However, I also agree that the case law of the EU Courts seems to endorse a narrower ‘proceduralist’ conception under which the presumption of innocence may only be violated when a person is convicted without proof of each constituent element of the offence and does not dictate what these should be, which is indeed a different matter.

      Thank you very much again for kindly sharing your thoughts!

      Andriani Kalintiri

      10 June 2020 at 9:31 pm

  6. While I fully agree that explaining the distinctions among the standard of proof and the different legal tests is indeed a very important task. They are clearly not the same (cannot be conflated) and serve for different goals. However, the discussion in your article is somewhat contradictory if I try to translate your excellent assessment into practical understanding. You alerady made it clear the the standard of proof is rather a “static”

    Almos Papp

    27 June 2020 at 7:02 pm

  7. or some reason I am unable to post my reply — so I copy the rest of my comment:

    You already made it clear the standard of proof – as opposed to the substantive legal tests – is rather a “static”, i.e., the same standard of proof applies to cartels and abusive cases. It is also “beyond any reasonable doubt” that different legal tests apply to different behaviors (cartels or other antitrust infringements). These latter tests are intended to demonstrate, as you correctly referred to, different levels of harms to qualify a given conduct unlawful. The substantive tests revolve in many cases around the “effects” that must be “plausible”, “potential” “likely”, “probable”, etc. And this is the point where, in practical terms, I am puzzled. First of all, if the evidentiary burden is the same in all cases (correctly), i.e., the evidence needs to be “beyond any reasonable doubt” and is to be sufficient to warrant a “firm conviction,” how can we apply it to prove legal tests that require the demonstration of less strict probability (i.e., “plausible,” “potential” “likely,” “probable”)? In other words: how we align the application of a strict standard of proof rule to prove a softer and more elusive legal test?
    Thanks for the excellent work, looking forward to hearing your thoughts of this,

    Almos Papp

    27 June 2020 at 7:09 pm

    • Thank you very much, Almos (if I may)!

      If I have understood it correctly, the question you raised touches on the separate issue of how the standard of proof can be discharged – rather than what it is/should be. This exercise though is not ‘static’ at all – on the contrary, it is inevitably case-specific and arguably more of an art than a science. How easy or difficult meeting the standard of proof will be depends on several factors. What needs to be proved (i.e. the elements of the legal test) is one of them, but not the only: the amount and quality of the available evidence, normality considerations based on common sense, past experience and economic premises (i.e. generalisations about the competitive effects of a practice based on economics), and the existence of any presumptions, play an important role, too. For example, in the case of cartels, establishing a ‘firm conviction’ about the potential negative effects of a specific cartel arrangement will not be as difficult, among others, thanks to our experience with conduct of this kind and the economic premise that cartel behaviour lacks any procompetitive explanation and is harmful. That said, the ‘puzzlement’ one might feel might also stem from two further issues: (a) the word ‘effect’ may be understood as outcome, i.e. something that has already happened, whereas in competition law the various thresholds of effects arguably indicate the required level of probability of harm (as we know, for a practice to be prohibited, it is not necessary that competition and consumers have been actually already harmed); and (b) the word ‘probability’ itself has different meanings and is used in different contexts (e.g. ‘subjective probabilities’ as degrees of belief, ‘objective probabilities’ as the chances or odds that an event will occur, as part of risk assessment etc.).

      I hope this makes sense and that I did not create more questions than answers! Many thanks again for engaging with the post.

      Andriani Kalintiri

      29 June 2020 at 7:20 pm

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: