Selective Enforcement and Multi-Party Antitrust Infringements: How to Handle "Unilateral Agreements"?

Y EARBOOK


I. Introduction
The difference between Article 101 and 102 TFEU may seem straightforward.The former is about collective practices, the latter about unilateral actions.Article 101 TFEU investigations typically concern groups of undertakings, while Article 102 TFEU investigations concern single undertakings. 1Yet, at some point of the enforcement history of Polish antitrust law, a doubt has been cast over whether "unilateral agreements" might be a thing.This would be the case of an Article 101 TFEU investigation that is directed at a single undertaking.Recently, a case of this kind has also attracted attention in the Czech Republic, where the Czech National Competition Authority (hereinafter: NCA) decided to only fine an organiser of a distribution system but then saw its decision being overturned in judicial review for not identifying clearly enough the members of collusion. 2he term "unilateral agreements" was originally us ed by Jurkowska-Gomułka, and then more extensively by Kolasiński, as a criticism of the approach adopted by the Polish NCA in relation to vertical agreements. 3he approach has been that in vertical cases (mostly Re tail Price Maintenance, RPM) the Polish NCA often initiates proceedings only against the undertakings responsible for setting up distribution systems, that is, it pursues a policy of selective enforcement.In consequence, decisions of UOKiK (the Polish Competition and Consumer Protection Office, hereinafter: UOKiK) are only addressed to the "organisers" of such systems and fines are imposed only on them.Conversely, other undertakings (typically retailers) remain unpunished and there is no finding of an infringement with regard to them.Against this backdrop, Jurkowska-Gomułka and Kolasiński argued that "unilateral agreements" are becoming a target of antitrust enforcement -something not envisaged under antitrust law, and thus an enforcement error. 4Instead, all collusion members should be prosecuted and infringement decisions should indicate them clearly.
YEARBOOK OF ANTITRUST AND REGULATORY STUDIES approach. 5The issue is thus relevant also in a broader European context.This is even more so taking into account that in digital markets agreements may involve hundreds, if not thousands, undertakings -a restriction of the power of antitrust authorities to "selectively" enforce Article 101 TFEU may undermine the effectiveness of competition rules for years to come.
There are some challenges in discussing this topic in a broader European context.This is because the controversy occurred at the national level, meaning that many arguments are of a more national nature, mostly coming from national procedural frameworks.There is thus a risk of either going too deeply into the peculiarities of national laws, or leaving certain arguments that are relevant in a specific national context unanswered, due to their nonuniversal character which does not justify a thorough analysis in a supranational context.
The article looks for a possibly balanced approach.It focuses on two general issues.First, what might be the reaction of the Court of Justice of the European Union (hereinafter: CJEU) if someday the approach adopted by the European Commission is challenged in an appeal?Second, what might be the reaction of the CJEU, if it is asked for a preliminary ruling by a national court?The article starts with an outline of the policy adopted by the European Commission and EU Member States.Subsequently, counterarguments to this policy are presented.With this background, the validity of these claims is discussed.This discussion does not aim for an exhaustive rebuttal of the counterarguments discussed earlier -such a rebuttal would need to address each specific argument made on the national level, while the goal of the article is to think about these issues from a broader European perspective.Still, based on this discussion, the article concludes that courts should be wary of curtailing antitrust authorities' powers when it comes to how they establish infringements of Article 101 TFEU.Antitrust authorities themselves, however, should be more cautious in their enforcement practice and more aware of the consequences of the policies they adopt.

II. The current way 1. General outline
The number of undertakings i nvolved in cartel cases is typically limited.While not impossible, it is harder to set up and maintain a cartel that involves many market players (mostly due to coordination issues, communication, balancing interests, monitoring, cheating etc.).
Conversely, in vertical cases, there is typically a supplier that has a lot of power in deciding how its distribution network works. 6The supplier may also easily keep a bird's eye view on actions taken by the buyers.On the other hand, buyers' impact on the distribution system is often limited.In essence: "eliminate" the supplier, and any anticompetitive agreement that emerged thanks to the supplier will likely collapse.This can be seen as a form of "selective enforcement". 7he European Commission used to target suppliers only in a mo re distant past, around the time of the notification system.8However, this is also an approach that has been used more recently when the European Commission made its comeback to RPM cases.9If the European Commission follows this way of enforcement, i ts decisions include findings that there 6 In this paragraph and throughout this article, the terms "supplier" and "buyer" (both included in e.g.Article 3 of Regulation 2022/720) are used.However, insofar RPMs are concerned, typically an agreement will take place between wholesalers and retailers.

YEARBOOK OF ANTITRUST AND REGULATORY STUDIES
were agreements (or agreement) between the supplier and buyers, with the buyers being indicated by cross-references to the discussion of the facts of the case.These agreements are found to be part of the same (single and continuous) infringement.The operative part of a decision, in turn, indicates that the supplier infringed Article 101 TFEU by restricting the ability of buyers to set their market policies independently.
In the European Commission's enforcement practice, the reasons for such an approach are not provided, and there are no court judgments known to this author that discuss this issue in detail.10Generally, it seems justified to assume that expediency is the main reason for opening proceedings against just the organiser of a distribution system.
At the national level, the Polish case is an interesting example where the standpoints of the NCA, the judiciary, and of antitrust literature were considered.Generally, the approach adopted in Poland is similar to that used by the European Commission. 11However, in some instances, the Polish NCA opens proceedings against all undertakings -typically, when the number of undertakings involved is smaller and the infringement itself is more akin to a cartel. 12he legal reasoning behind this approach has been more thoroughly discussed in Poland than at the EU level.The Polish Competition Act includes a provision saying that the party to antitrust proceedings is the undertaking against whom proceedings were opened. 13This is in contrast to general Polish administrative rules, which specify that the term "party to proceedings" should be understood as everyone whose legal interest or obligation is the subject of proceedings. 14Given the language of the Polish Competition Act, the Polish NCA and the judiciary concluded that: (a) a decision can be addressed to just one undertaking and the operative part of the decision does not have to list all members of collusion; (b) the authority enjoys discretion in deciding against whom proceedings will be opened (and may thus choose just one undertaking, e.g. the organiser). 15his approach received a mixed reception by lower instance courts. 16Yet, so far it has received unequivocal support from the Polish Supreme Court.According to the Supreme Court, as long as a decision indicates concrete facts and evidence of collusion, it cannot be said that the supplier could not defend itself. 17n Baby Direkt, on the other hand, the Czech NCA addressed a decision to just the organiser of a distribution system and, at the same time, it did not include the names of the relevant retailers in the operative part of its decision. 18his, and the fact that the Czech NCA did not discuss communications with each specific retailer, led the Czech Court of 1 st instance to conclude that it cannot properly review the case, as proving the concurrence of wills was crucial.
A more targeted review aimed at identifying examples of how NCAs handle investigations suggests that the European Commission, the Polish NCA, and the Czech NCA are not the only ones that follow the approach outlined above.For example, in its recent Samsung case, the Dutch NCA addressed its decision and imposed a fine only on Samsung. 19In Super Bock Bebidas (which was subject to a preliminary judgment by the CJEU, albeit concerns a different issue), the Portuguese NCA followed a similar route. 20The German NCA also appears to see this approach as feasible, e.g. this is how it handled its Booking.com investigation, which was part of a broader effort of the NCAs to tackle MFN clauses in the hotel sector in 2015-2016. 21

Stakes and practical relevance of the doctrine
To illustrate th e practical relevance of this approach, it is useful to look at actual cases.Poland offers an interesting example, as it used to prosecute all members of anticompetitive vertical agreements, but over time changed its policy to a more nuanced one.In consequence, some of the older cases followed by the Polish NCA show actual instances of a competition authority prosecuting large numbers of undertakings.
For instance, in case Tikkurila I, the Polish NCA found an infringement that involved 86 undertakings. 22In case Poltrade, it prosecuted 141 undertakings. 23oth cases were in fact "small" in the sense that they were handled by one of the regional offices of the NCA (which in Poland are generally tasked with YEARBOOK OF ANTITRUST AND REGULATORY STUDIES pursuing smaller cases) and did not involve large entities.On a technical level, in each of these cases all undertakings had to be served with relevant legal documents, could access the file, make their views known, and have their liability proven in unequivocal terms.With regard to each of the undertakings, the decisions indicated when undertakings' liability ended and so forth.
While these investigations were national, one can easily imagine cases of similar sort at the EU level.The example of agreements between hotels and booking sites, which was mentioned earlier, illustrates this.While the hotel agreements were dealt with by the NCAs, it is likely that the European Commission could have also been well-placed to conduct an investigation.Had it done so, it would have likely needed to deal with thousands of hotels around the European Union, unless it had decided (in accordance with its current practice) to just focus on the booking sites.
The practical issue is, in fact, not only limited to classic vertical agreements.As there are more and more business models that revolve around digital services and platforms, it is not unfathomable to see, in the future, agreements of a more horizontal character that involve large numbers of undertakings.Uber's case may provide an example, as there was a time when it was theorised whether Uber's pricing algorithm might result in price-fixing. 24Had it been so, thousands of car drivers would need to be considered as colluding undertakings.

III. The right way?
The approach outlined above is not without controversy.The primary one is that an anticompetitive practice that results from actions taken by at least two undertakings, is then found to be illegal within a procedure that involves just one undertaking, and leads to an attribution of liability and fine to just this single undertaking. 25As mentioned earlier, it is argued that this is in a way a "unilateral" agreement -something akin to an abuse of a dominant position, yet still an agreement. 26Rights of defence are also indicated as a concern. 27ince not all members of an alleged anticompetitive agreement become parties to a procedure, competition authorities also feel less pressure to prove the agreement with regard to each member of collusion -thus, the question of the level of detail of decisions, as well as their operative parts, is raised. 28efore moving to a more detailed discussion of these arguments, a clarification seems justified though, as it appears there is confusion as regards what actually takes place within the type of proceedings in question.While the phrase "unilateral agreement" is catchy, it is a confusing one.This is because competition authorities do, in fact, establish the existence of agreements in this type of cases.It does not logically follow from the fact that liability was attributed to one undertaking that there was only one party to an agreement.Therefore, the question is rather whether liability can be attributed to just one undertaking, if an agreement involved more undertakings (and an agreement must involve at least two undertakings) -and if so, what are the consequences of doing so.
The counterarguments to the approach at hand can be grouped in a following way: (a) arguments concerning substantive rules (Article 101 TFEU and its national equivalents); (b) formal arguments concerning procedural frameworks; (c) rights of defence; (d) procedural autonomy (this is of relevance only in relation to the NCAs); (e) private enforcement; (f) leniency (this might be of relevance in Member States that use broad leniency programmes that include e.g.RPMs -this is the case in Poland, but also e.g. in Romania).

Substantive rules
The substantive argument can be phrased as a "wide substantive argument" or a "narrow substantive argument". 29he wide argument is following.Article 101 TFEU says that an agreement takes place between undertakings.It can be argued therefore that to find such an infringement, liability has to be attributed to all undertakings involved.This is because the legal question in Article 101 TFEU investigations is about the interaction between a group of entities (i.e. the fact that they agreed on something that constituted a common plan to restrict competition).The argument follows that since establishing that A agreed with B, that means that B agreed with A -this requires having both A and B in the same legal proceeding.Moreover, if there are A, B, C, and D involved, all four need to become the addressees of a decision -the authority cannot just opt for a "legal fiction" of an agreement between e.g. two of them, while ignoring the remaining parts of the infringement.This is because the objective and YEARBOOK OF ANTITRUST AND REGULATORY STUDIES material fact, which is under investigation, is the market interaction in its entirety.This market interaction as a whole gives rise to a legal consequence (an infringement of Article 101 TFEU) -establishing an infringement between two undertakings when, objectively, four undertakings were involved is not an option. 30urthermore, Article 101 TFEU does not provide any room to treat undertakings differently and e.g.initiate proceedings and address a decision only in relation to the organiser of an agreement -it speaks of all undertakings in equal terms. 31he narrow argument is that the competition authority is in fact allowed to select undertakings it will prosecute and to find an infringement just between them (without making reference to any of their contacts with other undertakings, i.e. ignoring them). 32Still, there will always be at least two undertakings that ar e addressees of a decision (and in such a case only an agreement between these two undertakings will be covered, while their actions with regard to possible other members of a single and continuous infringement will be ignored and no fines will be imposed in that regard).

Formal requirements
It might be that procedural frameworks prevent competition authorities from adopting decisions just in relation to a single undertaking if an infringement follows from a collective practice.It is argued that this is the case in e.g.Poland. 33Since in a broader European context specific rules included in national legislation are of less relevance, this subsection focuses on the ideas underlying this argument, rather than the peculiarities of national laws. 30In Poland, this argument is further supported by saying that it is unjustified to conclude that there was one agreement (between the wholesaler and retailers in general), if evidence indicates multiple agreements with specific retailers.However, this argument seems to be based on a misunderstanding.The context here is that the Polish NCA does not mention in the operative part of its decisions that a single and continuous infringement took place.It instead replicates the statutory wording of the Polish Competition Act and mentions that there was an "agreement".It is not uncommon under EU law that when a single and continuous infringement is established, the contribution of undertakings to this infringement may vary.This has been called by some as "asymmetric liability", see Kevin Coates, 'Defining a single and continuous infringement in cases with asymmetrical participation' (2016) <https://www.twentyfirstcenturycompetition.com/2016/05/sci-and-asymmetry>accessed 31 March 2023. 31Kolasiński (n 3), 20-22 (insofar he refers to the Polish equivalent of Article 101 TFEU). 32Kolasiński (n 27). 33Kolasiński (n 3), 23-36.
Overall, national procedural rules might, for example, require a specific level of precision as regards the finding of an infringement.This may follow directly from national legal acts or their judicial interpretation.In Poland, for example, there are no specific legal provisions envisaged by the Polish Competition Act that list the elements of an antitrust decision (e.g. a provision saying that all members of an anticompetitive agreement need to be listed by name, even if they are not parties to the procedure).At the same time, it is argued that since administrative courts require a high level of precision with regard to the operative parts of administrative decisions, so should antitrust courts, since the Polish NCA is an administrative body. 34The argument further goes that since: (a) one of the formal requirements of an administrative decision under general Polish administrative law is that it includes a list of the parties to given proceedings; (b) then all members to an anticompetitive agreement need to be party to proceedings because of the wording of the Polish equivalent of Article 101 TFEU (see the substantive rules argument discussed earlier)addressing a decision to just a single undertaking is thus an error. 35 similar argument can also be made in relation to EU law.It is argued that while the European Commission had adopted the approach of addressing (some) decisions to just single undertakings, this approach had not been approved by EU courts. 36At the same time, it is suggested that case law, such as Air Canada, indicates that EU courts might not support the European Commission's approach, if they are given a chance to review its policy. 37ir Canada was a case concerning an air freight cartel that operated on various routes within the EEA and from/to the EEA.In the grounds of its decision, the European Commission characterised the cartel as a single and continuous infringement.However, due to a complex nature of the case (the cartel was long-lasting, it concerned multiple routes, and the applicable law included the TFEU, the EEA agreement, and the EU-Switzerland agreement, with significant differences with regard to which specific provision the European Commission was empowered to enforce at a given point in time), the European Commission divided, in the operative part of the decision, the conduct into four articles.This gave an impression of four separate infringements and led to certain inconsistencies in comparison to the grounds of the decision.The decision was thus annulled by the General Court, which YEARBOOK OF ANTITRUST AND REGULATORY STUDIES pointed out that the operative part of the decision needs to be "particularly clear and precise". 38n consequence, an argument against the current approach of the European Commission can be that it leads to the adoption of decisions that are not clear enough -in other words, "different route, same conclusion" as under national law.

Rights of defence
The undertaking needs to understand the identity of the accusation levelled against it.It is argued, therefore, that when a competition authority opens proceedings just against the organiser of an agreement, and issues a decision (and earlier a Statement of Objections) that does not name each and every undertaking involved in the agreement, the undertaking cannot effectively defend itself. 39This is because it is unaware of the identity of whom it allegedly colluded with (or rather, for what it is charged, since it may very well know with whom it colluded). 40

Procedural autonomy
Insofar as national proceedings are concerned, it can be argued that even if it is possible for the European Commission to address a decision to just the organiser of the practice, it might still be impossible to do so by NCAs, if only their procedural frameworks prohibit doing so (i.e. the formal argument discussed earlier is incorrect in relation to the European Commission, but correct with regard to an NCA). 41This is, therefore, not a standalone argument, but rather a supporting one. 38Air Canada (n 37), para.35. 39Kolasiński (n 7), 51. 40In Baby Dirket, this was also found relevant by the court insofar as the infringement decision did not list all members of collusion, see para 24.See also para.27 where the court observes that an argument that the undertaking "knows very well" with whom it colluded is not a proper argument. 41Kolasiński, (n 3), 39.This is thus a "procedural autonomy" argument.Some argue that the concept of "procedural autonomy" is misleading, and suggest replacing it with the term "procedural competence", see Walter van Gerven, 'Of rights, remedies and procedures' ( 2000

Private enforcement
An argument against finding of an infringement by just one undertaking might be that this makes private enforcement harder. 42The reason for which private enforcement would be hindered is that the infringement decision would only be binding on courts with regard to one undertaking (rather than all of them) and it would be unclear whether a specific plaintiff would in fact have sustained damage (e.g.whether retailer A, with whom the plaintiff contracted, was in fact involved in collusion).

Leniency
The leniency argument is simple: if leniency is available in vertical cases, and a competition authority establishes a policy that it only investigates the organisers of distribution systems, there are no incentives for the distributors to file for leniency. 43The distributors might assume that whatever happens, no proceedings will be instigated against them.In consequence, there is no point to bother with leniency, which in turn lowers deterrence and detection.

IV. The middle way?
Taking into account the arguments discussed above, the default alternative would be to expect competition authorities to open proceedings against all parties to agreements.
Yet, there is also another possibility.For instance, the authority may open proceedings (and address a decision) with regard to only some undertakings involved in an agreement -selecting which distributors are chosen could be based on e.g.their volume of sales. 44To some extent, this was the approach adopted by the Polish NCA in e.g.Fischer, where its decision was addressed to the supplier and one of the distributors. 45EARBOOK OF ANTITRUST AND REGULATORY STUDIES Such an approach goes against the wider substantive argument discussed earlier, since it explicitly accepts that not all undertakings involved in an infringement are prosecuted.However, it eliminates some of the doubts concerning the proper way of proving the concurrence of wills.Also, more than one party to an agreement would be clearly listed in the resulting decision -all parties whose participation in an agreement is discussed in the decision would also be in a position to challenge it, since liability would be attributed also to them. 46

V. Arguments supporting the current approach
Taking into account the discussion above, there appear to be at least three possible approaches: (a) antitrust authorities should always prosecute all members of collusion; (b) antitrust authorities should always prosecute at least two undertakings (and when doing so, they should not attribute liability for any actions involving undertakings that are not party to the proceedings); (c) antitrust authorities may prosecute single undertakings and attribute to them liability for all their actions taken in connection with an anticompetitive agreement.In the last scenario, controversy remains on how the authority should establish an infringement, e.g.whether it should list all members of collusion one by one and prove the concurrence of wills, and/or whether it should indicate them all in the operative part of the decision.This section discusses reasons for which the last approach might gain the support of the CJEU, in case of an appeal from an antitrust decision or a preliminary ruling, despite the arguments outlined earlier.The discussion in this section is "reversed" in the sense that it starts with addressing the least convincing arguments covered earlier, and then moves to discuss more relevant ones.Section VI, in turn, covers practical guidelines.

Leniency
The leniency argument is not, in fact, of much relevance.First, because it is only applicable in jurisdictions where vertical leniency is available.Second, since leniency might serve as a reason not to follow a policy of selective enforcement, but it is not a proper legal argument for not doing so.
explained that the distributor had a large volume of sales.However, it should be pointed out that "coincidentally" the very same distributor was also a leniency applicant.It was granted immunity under the leniency system. 46The word "discussed" is of relevance here, since there is a difference between discussing someone's actions and attributing liability -see section V.5.
It is true that the incentives of distributors to file for leniency might become weaker if liability is only attributed to the organiser (in fact, in terms of economic models, they should be expected to become weaker).However, this is a policy question concerning costs and benefits.If a competition authority loses more on litigating against distributors than it can possibly gain from hypothetical leniency from distributors, it is reasonable to focus on organisers.This is ultimately an area of political accountability, not a legal issue to be decided by courts -there are no legal links between the issue of addressing decisions and leniency incentives.

Private enforcement
Like leniency, private enforcement may also serve as a reason for a specific policy choice, but not as a legal argument. 47There are two types of links between public enforcement and private enforcement: legal and practical.
Legal links have been defined in the private enforcement directive -they concern e.g.presumptions associated with infringement decisions, amicus curiae opinions with regard to the assessment of damages, and discovery rules.Nevertheless, there are no other obligations on how antitrust authorities should shape their proceedings.Furthermore, to construe the private enforcement directive in such a way that antitrust authorities are obligated to prosecute undertakings in a specific way (which maximises the chances of private plaintiffs in a specific case), would be tantamount of creating a universal right to have a case investigated and prosecuted so that it is easier to claim damages, as this is the goal of the private enforcement directive.Such an interpretation remains highly questionable in its own rights; still, if such an argument is considered, then one should also take into account the ECN+ Directive which provides the NCAs with much leeway on how they prioritise their cases.
Practical links, on the other hand, come down to the fact that, admittedly, addressing a decision to all undertakings might make it easier for those injured to recover damages.Yet, as with leniency, this is merely a policy issue.Ultimately, there is a cost on the part of competition authorities of both conducting investigations against multiple parties, and of litigating against (likely) a large number of them.As there are no free lunches, one can expect 47 I first argued this in: Jan Polański, 'O skuteczności zwalczania naruszeń wertykalnych.Komentarz do wyroku Sądu Najwyższego z 15 lutego 2019 r. w sprawie I NSK 10/18 (Anyro)' (2019) internetowy Kwartalnik Antymonopolowy i Regulacyjny 2019, nr 6(8), 107-116, criticising a ruling of the Polish 2 nd instance court for accepting this argument -this ruling was set aside in Anyro (n 17).

YEARBOOK OF ANTITRUST AND REGULATORY STUDIES
that resources devoted to these tasks would need to be re-located from other tasks. 48The way these issues are balanced by an antitrust authority remains a question of policy -and the role of courts is not policy-making.

Procedural autonomy
Member States enjoy procedural autonomy with regard to the application of Article 101 TFEU.However, this autonomy is not without limits 49 -otherwise, NCAs could easily be prevented from investigating infringements of EU law by overly burdensome national procedural rules.Without limits in relation to procedural autonomy, each Member State could then define such rules in a "Roma locuta, causa finita" manner, having the last say over how antitrust infringements, including Article 101 TFEU violations, should be investigated.Supposing that, EU courts would support an approach where the European Commission may address a decision to just one undertaking (and as it will be discussed further on, there seem to be indications that the CJEU might do so), a national procedural rule that requires an NCA to conduct an investigation and address a decision to all members of collusion might thus be found to go against the duty of sincere cooperation, enshrined in Article 4(3) TEU.
To require an NCA to conduct proceedings against e.g.all Uber drivers, all hotels, restaurants, or fitness clubs in a country would be to force it to perform a gargantuan task, likely depriving EU law of its effectiveness. 50Cases such as T-Mobile, or more recently Whiteland, show that to preserve the effectiveness of EU law, the CJEU is willing to define standards of conduct even in relation to seemingly procedural issues. 51dmittedly, this issue can be easily brought to an extreme.While it can be clearly shown that the effectiveness of EU law would be significantly impeded when the number of undertakings is large, this is more ambiguous when the number of undertakings is smaller.Taking into account that effectiveness is a general principle of the EU and that the CJEU would need to apply it (if ever) in the context of a preliminary ruling, it might be that the CJEU would give a response adjusted to a specific case context, i.e. one that might be different when asked about a national provision and 10 undertakings, and a different one when asked about the very same provision and 1000 undertakings.It would likely also be a response that would require a national court to exercise its own scrutiny and make a factual decision regarding the risk of restricting the effectiveness of EU law in a specific case.
However, taking into account the aforementioned, it would be prudent for national courts and legislators to opt for the current model as the one that causes least problems when defining the mandate of competition authorities.There seems to be no good solution that could be adopted by national courts or legislators to discriminate between cases concerning 10, 100, or 1000 undertakings and define clear-cut rules.

Rights of defence
It is interesting that in Baby Direkt, the Czech court leaned towards the conclusion that the lack of a precise identification of members of collusion was a fundamental problem.By contrast, the Polish Supreme Court in Anyro arrived at a completely opposite conclusion, observing that the undertaking could exercise its rights of defence, as the Polish NCA named specific pieces of evidence showing that the undertaking operated an RPM within its distribution system.Is it then possible to defend oneself if other members of a multi-party infringement are not listed by their names?
When it comes to theory and models, it does not seem impossible to defend oneself, even if other undertakings are not indicated as parties to the proceedings, are not addressees of the relevant decision, are not mentioned by name in its operative part, or are not named at all.An example might help understanding this -the example is abstract and extreme, but it flashes out the relevant legal question. 52Since the example serves as a "model" (in a similar way as perfect competition and monopoly can be used as models in economics), it concerns a horizontal infringement, i.e. a type of conduct which is more straightforward than vertical restraints.
Let's imagine a tight oligopolistic market with four undertakings (A, B, C, D).One day, three of them meet and agree to raise prices."A" is one of the participants of the meeting and one of its employees writes a memo about this meeting.The memo says that an agreement was reached with "our YEARBOOK OF ANTITRUST AND REGULATORY STUDIES two competitors", but that "the remaining competitor did not attend, yet can be expected to increase its price to follow others" (remember that this is a tight oligopoly and it is not unusual to intelligently adapt oneself to the actions of others in such circumstances).And indeed: despite the fact that price increases were not implemented on the same day, they slowly rose to the agreed level. 53The competition authority conducts an investigation and finds the memo.However, it is unable to identify (name) other members of the collusion.Having the memo, should the authority prosecute the one and only perpetrator it knows, or close the case?54Is it possible for "A" to defend itself?
In my view: (a) it is not impossible to defend oneself in such circumstances; (b) there are no reasons to drop the case against the undertaking "A".This also shows that the legally relevant question is not whether the party to the proceedings was informed of the authority's beliefs with regard to the identity of other cartel members.The relevant issue was that it was given the opportunity to explain why e.g. the mentioned evidence was unreliableand this is something that needs to be analysed case-by-case; it is not a deceive argument against the model of selective enforcement itself.
To remain objective, a possible weakness of this parallel is that it simply replaces other collusion members with otherwise specific anonymous figures on a "nomen nescio" basis, while antitrust decisions that are typically subject to controversy refer broadly to "some" retailers.Sometimes such a general characterisation might still allow effective defence (e.g. when evidence is clearly presented), but it might be that in specific circumstances it will be questionable.Since this issue is nuanced, it might require more caution on the part of antitrust authorities, which will be further discussed in section VI.
A different way of looking at the rights of defence argument is from the point of view of undertakings which are not prosecuted.They do not face any liability (at least not within the proceedings that were opened), but their actions are discussed in the context of some other undertaking's liability.A concern can be voiced that their "liability" is established without giving them an opportunity to defend themselves.Since this issue is connected with more recent case law developments, which are relevant also from the point of view of the formal argument, this issue is covered in more detail in the next subsection.

Formal argument
In April 2021, the CJEU delivered a ruling on an issue seemingly very different than the one discussed in this article, namely the validity of a so-called "hybrid settlement" procedure in Pometon (the Steel Abrasives case). 55However, interestingly, the situation in this type of cases might be relevant from the point of view of selective enforcement.
Hybrid settlement procedures are used by the European Commission in the context of cartel cases, when some parties decide to settle and some do not.In consequence, the European Commission adopts a settlement decision (which is a type of an infringement decision) against some parties, and then (typically after some time) delivers another decision (or decisions) against the non-settling party (or parties).This caused concerns when it comes to the rights of defence of the non-settling party, in particular the presumption of innocence.However, on the theoretical level, the situation here is also similar to issuing a decision with regard to just a single undertaking -this is because an infringement is found, yet not all undertakings who actually took part in it are made the addressees of the decision.
In Pometon, the CJEU found that hybrid procedures are possible, but that caution should be exercised when it comes to describing the conduct of a nonsettling party.This is because the non-settling party might be found liable in a future decision.Indicating in the settlement decision that the company did infringe competition rules, would put in question whether the presumption of innocence was respected, and thus jeopardise the integrity of a possible future infringement decision.What needs to be emphasised is that the CJEU did not consider whether any of the decisions would be invalid due to a lack of precision with regard to the members of collusion.It was merely the possibility of "saying one word too many" in an earlier decision which could put at risk the decision that comes later. 56ometon can be taken as an indication that if the CJEU is confronted with an appeal or preliminary request, it might not follow a formal argument of the kind discussed earlier and based on e.g.Air Canada; apparently, even in cartel cases, settlement decisions that do not indicate all members of collusion as co-infringers are not an issue for the CJEU.Conversely, the CJEU appears to support the position that in some circumstances (i.e.hybrid procedures), indicating all members of a cartel as co-infringers should be avoided so that liability is not prejudged.

YEARBOOK OF ANTITRUST AND REGULATORY STUDIES
This is in fact not uncommon also in other types of cases, even outside antitrust.For example, in Pometon, the EU courts referred to similar issues in criminal cases.This concerned e.g.Karaman, which was heard by the ECHR. 57n Karaman, a criminal investigation took place in which co-conspirators were tried in separate procedures.This did not infringe fundamental rights, as long as the prosecuting authorities exercised caution with regard to the presumption of innocence.
This also seems to be the case in the US.American antitrust investigations make extensive use of plea bargaining.Settlements with specific parties do not have to be reached at the same moment -they can be staggered and lead to a "snowball effect". 58This indirectly means that a guilty plea of one party can be accepted before handling the liability of other parties -it is also possible to have a jury trial with regard to a cartel member that did not settle, even if others did.
While the outcome of Pometon can be used as a supporting argument for the current approach to addressing decisions, admittedly there are some differences between this case and prosecuting just one member of collusion.Still, upon closer inspection, none of those differences seems to undermine the Pometon parallel.
First, it can be argued that in Steel Abrasives, the European Commission first issued a settlement decision, wherein a single and continuous infringement was found, for which the Commission attributed liability to four undertakings, in other words, the decision did not include Pometon (the 2014 settlement decision). 59Then, it adopted an infringement decision (the 2016 infringement decision), which ultimately led to Pometon. 60This may give an impression that the first decision covered a part of the infringement, and that the second decision covered another part of the same infringement.
However, the language of the 2014 settlement decision indicates that it concerned all facts giving rise to a single and continuous infringement as a whole, but its legal conclusions were only relevant insofar as the liability of the four settling parties was concerned.In other words, the four settling parties' conduct with regard to Pometon was relevant, but without Pometon being an addressee of the decision.For example, in paragraph 18, the 2014 settlement decision lists all five undertakings (including Pometon) and calls them "parties", as opposed to "settling parties" -a term which it introduces in paragraph 22.Then, on multiple occasions, the European Commission refers to "the parties" when discussing conduct relevant for the decision, and says e.g."The geographic scope of the conduct, as regards all five parties, was EEA-wide during the entire period concerned by this Decision".Then, in the legal assessment it mentions e.g. that "With their contacts, the parties pursued a single anti-competitive object and a single economic aim, namely the distortion of the normal movement of prices in relation to steel abrasives".Still, the European Commission does not use the word "parties" in the context of attributing liability or when it directly states that some conduct was an infringement.Thus, the 2016 infringement decision "did the legal work" with regard to Pometon, and attributed liability to this undertaking also; nonetheless the conduct of the other four parties with respect to Pometon had been already covered by the 2014 decision. 61his, in fact, became a major point of contention when Pometon appealed the General Court's judgment. 62Pometon argued that the European Commission did prejudge its liability by including in the contested decision the aforementioned references.The Advocate-General also leaned towards a conclusion that the European Commission did not act with full impartiality. 63till, this approach was not followed by the CJEU, which accepted that, in 2014, there was a binding decision that covered certain actions of Pometon, but attributed liability to just the other four undertakings -Pometon not being any of them.Addressing a decision to just a single undertaking leads to a similar result.
Another argument against the hybrid procedure parallel may come from the language of Pometon itself.The ECHR, General Court, and the CJEU use the following wording: "in complex criminal proceedings involving several persons who cannot be tried together (…)".The word "cannot", used originally by the ECHR, can be taken as an objective and unavoidable obstacle that YEARBOOK OF ANTITRUST AND REGULATORY STUDIES prevents adopting a joint (single) decision against all parties at the same time -this is not the case of ordinary investigations concerning e.g.RPMs.
Still, the actions of the CJEU suggest that this is not how it construes the requirements that open the path towards the issuance of a decision that is not addressed to all members of collusion.This is because a decision to settle taken only by some parties, is not an objective obstacle (the European Commission is in a position to refuse a settlement submission and issue a full infringement decision against all parties, even those willing to settle).Furthermore, this argument in no way affects the relevant legal fact, i.e. that the CJEU appears to be comfortable with seeing decisions that are not addressed to all partiessettlement decisions issued in hybrid procedures become binding and effective, even despite the fact that the liability of some members of collusion is not decided at the time when such settlement decisions are adopted.
To conclude, there appears to be no indication that EU courts see any formal requirements to address a decision to all undertakings involved in an agreement.It is true that discussing the actions of an undertaking, which is not an addressee of a decision, might increase the risk of a further decision being invalidated due to a violation of the presumption of innocence.Nonetheless, this is not automatic, and: (a) such later decision may still hold, provided that caution was exercised; (b) if there is no further decision, no liability is attributed, and hence the issue does not arise at all; (c) this has no effect whatsoever on the earlier decision (since the presumption of innocence cannot be used as an argument by the addressee of the original decision, just by the addressees of a possible future decision).

Substantive argument
Much of the discussion that was relevant in relation to the formal argument is also applicable to the substantive argument.When it comes to the formal argument, the emphasis was on whether it is possible to issue a decision that does not attribute liability to all members of collusion.As regards the substantive argument, the emphasis is on whether it is even possible to find an infringement in such a case (this was called the wide substantive argument).This includes whether a concurrence of wills can be found without all parties being the addressees of a decision (this is also relevant to what was earlier called the narrow substantive argument).
Pometon suggests that the CJEU does not see an issue with this.As it was explained earlier, in Pometon the European Commission has found a "wide" cartel -it had not referred to Pometon as a liable undertaking in the legal assessment within the 2014 settlement decision, but it had referred to Pometon's actions in the discussion of case facts.
Furthermore, given the unwillingness of EU courts to compromise the effectiveness of EU law, it seems unlikely that they would settle in situations such as those discussed in section II.2 and V.4, which would make it prohibitively difficult or impossible to find an infringement.
There is also a different problem with the substantive argument, in particular the "wide" one.If it is concluded that under substantive law all members of collusion need to be made into the addressees of a decision, then the decision is invalid if the authority fails to prove someone's liability.This leads to a somewhat absurd situation in which if the authority fails to establish someone's liability, or does not pursue a case in relation to someone (e.g.due to lack of evidence), it is in the interest of other undertakings to attempt striking down the decision by saying that there were more parties to the agreement.This could marginally increase their liability, but it would also make it possible to drag out the investigation for years -and doing so may both work as a defence strategy and a life-saver for board members, whose interests are not always fully aligned with those of shareholders.This might not be a huge issue in cartel cases, but in RPM cases, that may involve hundreds of undertakings, it is not unlikely.This would make Article 101 TFEU highly ineffective.
The "narrow" substantive argument, that requires finding an infringement by at least two members of collusion, is also questionable from the point of view of the effectiveness of Article 101 TFEU.First, it requires leaving some parts of a single and continues infringement unaddressed, unless all undertakings involved are made into addressees of a decision.Second, this would have an even stronger impact on the "by effect" part of Article 101 TFEU.Under the narrow substantive argument, the proposal is that the competition authority can simply pick the largest undertakings and prove an agreement between them.The (alleged) substantive requirements would be met, since an agreement between a group of addressees of a decision would be clearly proven, and, at the same time, the authority would alleviate itself from the need of prosecuting all members of collusion.Still, in "by effect" infringements, picking just the largest undertakings might still be insufficient to prove an infringement -in cases such as hotel booking this would likely require addressing a decision to all hotels.
Ultimately, there is also nothing in Article 101 TFEU itself that says that all members of collusion need to be the addressees of a decision, and that this is somehow part of substantive law.The "there is nothing saying that…" issue is not an official element of CJEU's method of legal analysis, yet one could say that there is a tendency within the EU judicature to opt for the effectiveness of EU law.Thus, the EU courts opted for effectiveness when e.g. they were asked YEARBOOK OF ANTITRUST AND REGULATORY STUDIES to decide on cartel facilitation, continued inspections, and also in Pometon itself, with regard to hybrid settlement procedures. 64he fact that the requirement of addressing a decision to all members of collusion might not be part of Article 101 TFEU on the substantive level is of high relevance.It means that any attempt to introduce such an obligation in relation to the NCAs would need to rely on merely national procedural law, since there is no autonomy when it comes to the interpretation of what follows from Article 101 TFEU.Thus, in a worst case scenario (from the point of view of the effectiveness of competition rules), an NCA could simply enforce Article 101 TFEU (if possible).

VI. Speed limits
The discussion above suggests that out of the three options mentioned at the outset of section V, the option of requiring antitrust authorities to always address a decision to all members of collusion, and the option of requiring them to always open a case against at least two undertakings, might not gain the support of the CJEU.Does it then mean that antitrust authorities face no risks in addressing a decision to just a single undertaking?There seem to be indeed certain limits to this approach.
One of clear technical limits follows from Pometon.Supposing that an authority would e.g.issue a decision with regard to just the organiser of an RPM system, any other decisions adopted with regard to distributors would run a serious risk of violating the presumption of innocence.This is more of a theoretical concern, in the sense that the authorities specifically target organisers to achieve more efficiency -they typically do not issue further infringement decisions after the first one is adopted (contrary to what happens in hybrid procedures).Nonetheless, it would still be good practice to ensure that decisions addressed to just a single undertaking mention only the most relevant facts in relation to non-addressees, and, in particular, that such decisions do not impose "in passing" liability on them.
Furthermore, since the policy of not opening cases against any other undertakings after addressing a decision to the organiser is not stated anywhere, there might be added value, in terms of transparency and good governance, to clearly formulate such a policy.Obviously, however, this is not a requirement of any form.
A different issue is whether competition authorities should generally increase transparency in relation to who will become an addressee of a decision, e.g.just the organiser of a vertical agreement if it concerns an entire distribution system that was initiated by the supplier; or a selected group of undertakings in some other circumstances; or to all undertakings involved in the agreement in yet another type of situation. 65While this might increase transparency, it is not necessarily a good policy choice.This is mostly because delineating such scenarios might be difficult in abstract terms, and what can be expected is that whenever some other undertaking, rather than the organiser, is made subject to an investigation, arguments will follow that the authority misapplied its soft law, which in turn means that its decision should be annulled.Such transparency might also have a stronger negative effect on deterrence.
The more contentious issue is whether an antitrust authority should always provide a clear and precise list of members of collusion, even when a decision is addressed to just a single undertaking -as indicated earlier this was of much relevance in Baby Direkt.This issue appears to be case-specific.On the one hand, antitrust authorities should produce "sufficiently precise and consistent evidence to support the firm conviction that the alleged infringement took place".66However, to construe this as an indication that, in each and every case, all undertakings have to be identified, might go too far.This is because the standard of proof, at least under EU law, is that the authority should provide a coherent body of evidence indicating that there was an infringement.Evidence is assessed in a holistic way.67Thus, in some circumstances, case facts might suggest that merely a general description of, for example, a distribution system will indicate that all retailers were involved in the practice -in such a case listing them does not seem indispensable.Still, in a case where there was a tighter group of retailers, with some of them clearly not participating in e.g.meetings -a different approach might be needed: either explaining why in spite of such non-attendance they were still involved, or a clear indication that they were not involved.
A suggested approach therefore should be that merely because a case is vertical, the analysis conducted by antitrust authorities should not become overly superficial and automatic.However, as pointed out by the Czech court in Baby Direkt, this does not necessarily mean the same level of precision as in relation to cartels, e.g.reservations such as "at least" can be used more YEARBOOK OF ANTITRUST AND REGULATORY STUDIES easily. 68Also, if it is shown that a distribution system generally worked in a specific way, then there is more scope for using presumptions that already exist under competition law, e.g.concerning failing to distance oneself from communications, or of a causal link. 69

VII. Conclusion
Vertical agreements have different dynamics than horizontal ones.They often involve far more undertakings, yet are also far more reliant on the actions of undertakings that organise them, i.e. suppliers, who act as ringleaders.Targeting leaders has always been an efficient and effective strategy: this includes crime-fighting, warfare, and antitrust enforcement.
The European Commission's comeback to RPM investigations in 2018 was exceptionally smooth, since each of the decided cases was closed in a cooperative way and without litigation, even despite the fact that the EU cartel settlement procedure does not apply to RPM.Nevertheless, such a successful outcome is not guaranteed in the future.Furthermore, the actions of the European Commission provided additional incentives to NCAs to re-adjust their priorities and also investigate more vertical cases.However, the "rules of engagement" in relation to vertical investigations seem under-developed in comparison to horizontal infringements.Hence, without unambiguous standards on how to conduct proceedings in such cases, it is possible that at some point the CJEU will be faced with either an appeal or a preliminary request concerning this issue.
There are at least three models of enforcement that can be used: (a) expecting antitrust authorities to address decisions to all members of collusion; (b) picking more than one undertaking and imposing liability only for the part of infringement that took place between the parties of proceedings; (c) allowing decisions to be addressed to single undertakings, with liability being imposed for all anticompetitive actions.The first two approaches have been already advocated in literature and e.g. in Poland have been considered and (so far) rejected by the Supreme Court.Yet, as the recent Czech Baby Direkt case shows the issue of how to shape proceedings is still lively debated and one can imagine that arguments that were unsuccessful in one jurisdiction might become successful in another one. 68Baby Direkt (n 2), para.24. 69See e.g.case T-342/18 Nichicon Corporation EU:T:2021:635, para.383.See also e.g.Guess (n 9), para.97-98 on tacitly agreeing to vertical restraints.When it comes to the causal link, see e.g.T-Mobile (n 52).
This article suggests that courts should not opt for the first and second model outlined above, and that there seem to be unexpected parallels between the case of targeting just organisers of anticompetitive agreements, on the one hand, and hybrid cartel settlements, on the other.
When it comes to the current approach, however, it would be useful if antitrust authorities exercised more caution in drafting their decisions, as it seems that ambiguous drafting of decisions might be of itself a source of attempts to introduce a requirement of addressing decisions to all members of collusive agreements.
The discussion provided in the article also suggests that despite ongoing soft harmonisation that is possible through the European Competition Network, and harmonisation through such instruments as the ECN+ Directive, the EU enforcement system still runs serious risks of arriving at divergent results that can come from developments in national courts.While the system of preliminary rulings will likely still serve as an important counter-measure in that regard, the upcoming revision of Regulation 1/2003 might offer important opportunities to bringing more uniformity to e.g. the shape of decisions issued by the NCAs with regard to infringements of Article 101 TFEU.This could happen by e.g.introducing a more detailed regulation in what is today Article 5 of Regulation 1/2003, which currently merely says that infringement decision can be issued, but does not list necessary elements of such decisions.