Joint German-French antitrust enforcers’ study on Algorithms and Competition

On the 6th of November the German competition authority (Bundeskartellamt (Federal Cartel Office (FCO)) and the French competition authority (Autorité de la concurrence) have published their joint study on "Algorithms and Competition", which deals with the opportunities, but also the challenges and risks of the use of algorithms in the digital economy. The study comes just before the envisaged 10th amendment to the German Competition Act, the Act Against Restraints of Competition (Gesetz gegen Wettbewerbsbeschränkungen (GWB)), which will implement other substantial new features on digital aspects of competition law.
 

Past and future competition law initiatives in the field of digital business

Already in 2016 and 2017, the authorities published the joint working paper "Competition Law and Data" and the article “Competition Law and Big Data: The Enforcers’ View”. Together with the FCO´s contribution to "Big Data and Competition" (available only in German) published in October 2017, the competition authorities are showing once again pioneering spirit with regard to the further development of European and national competition law in the field of digital business.

The German legislator took up new challenges for competition law arising from the digitization process of the economy already in 2017 by amending the German Competition Act. Back then, a new value based threshold was introduced to enable the FCO to review acquisitions in particular of start-ups in the digital economy. Moreover, specific rules for multi-sided markets and networks were enacted taking into account the increased significance of search engines and comparison websites.
 

Different effects of algorithms on competition

The study published now goes a step further. It focusses primarily on the use of pricing algorithms, which entail many advantages for undertakings, as the enforcers admit. Such algorithms can help a company to save costs or automatically adjust the price of a product to daily conditions such as purchase prices of raw material, capacity and the demand situation. This ultimately could even be beneficial to end consumers. However, the study also outlines the associated risks to competition which can arise when such algorithms are used to adjust a company's own prices to those of competitors.

The authorities come to the conclusion that, generally speaking, it is not prohibited to adjust one's own pricing intelligently to the existing or anticipated conduct of competitors. Yet, the enforcers stress that the outcome of pricing algorithms might also lead to collusion and negative price effects for consumers which is not easy to detect and prove.
 

Accountability for the behavior of algorithms

The study also deals with the question of whether a company can be held liable for the use of – possibly self-learning – algorithms under the existing provisions of Article 101 TFEU (ban on cartels) or the national provisions respectively.

In essence, the joint paper describes three scenarios:

1. “Algorithms as supporters or facilitators of ”traditional” anticompetitive practices”
   
   In this scenario the algorithm is only used as a technical tool to implement the anticompetitive behavior of a prior agreement. Therefore, the study concludes that the use of the algorithm “does not raise specific competition law issues”.
    
2. “Algorithm-driven collusion between competitors involving a third party”
   
   In the second scenario the algorithm is used by a third party to provide information as an intermediary for competitors (“hub and spoke”). With reference to European case law such as VM Remonts and Eturas the study stresses that it will depend on the specific case if such conduct is prohibited.
    
3. “Collusion induced by the parallel use of individual algorithms”
   
   In the last scenario there is no prior or ongoing communication between competitors, but the pricing algorithms might communicate in a way to facilitate an alignment of market behavior. In particular self-learning algorithms might indicate a certain change of a relevant parameter of competition to other competitors which then could lead to parallel behavior that cannot be differentiated from an intended coordination. The paper concludes that it remains an open questions whether an alignment of pricing algorithms could arise “by chance”, which makes it difficult for enforcers. Yet, the authorities outline the need to attribute algorithmic behavior to companies.

Conclusion

The study shows that various legal questions with regard to the use of algorithms and their effect on competition are still unanswered. In particular when complex self-learning algorithms are involved, the authorities have not found “one-size-fits-it-all” answers.

However, it would be welcomed if competition authorities could in fact give more legal certainty by defining conditions which would need to be fulfilled in order that algorithmic behavior falls under the cartel ban and can (or cannot) be attributed to companies in the digital economy making use of it. Suggestions have been made in the literature to limit excessive accountability and the study explicitly refers to these voices (study p. 57 with reference to e.g. Janka/Uhsler, Antitrust 4.0, European Competition Law Review 2018, pp. 112 et seq. (121)). Yet, the authorities – not much surprisingly – suggest a very broad approach, argue that escaping liability when using an algorithm should only be possible under “exceptionally atypical circumstances” and emphasizing the need for “compliance by design”.

Thus, aspects of competition law in the context of the use of algorithms will have to be further developed and clarified in future (court) cases. 

If you have any questions please feel free to contact

	Dr. Sebastian Felix Janka, LL.M. (Stellenbosch) Partner Lawyer Munich T +49 89 23714 10915
	Roland Schiller Associate Lawyer Munich T +49 89 23714 20967