Algorithms and Collusion: Competition Law Challenges of Pricing Algorithms

The purpose of this paper is to find possible measures to tackle algorithmic collusion, caused especially by pricing algorithms, from the perspective of competition law. This paper does not intend to analyse all possible measures that may exist, but rather, focus on a few measures that are seen reasonable. Before proceeding further to analyse these possible measures, the concepts of algorithms and collusion are examined. The second chapter consists of analysis examining the meaning of algorithms as well as certain benefits derived from them. The third chapter analyses the concept of collusion from the perspective of EU Competition law. Essentially, competition law of the EU will function as an appropriate benchmark for further analysis. The fourth chapter is there to examine the concept of algorithmic collusion through four scenarios: the Messenger, the Hub-and-Spoke, the Predictable Agent, and the Digital Eye. The first two scenarios are dealing with situations where algorithms are used as helpful tools to make explicit collusion possible. In other words, there is an element of agreement or meeting of the minds between the colluding undertakings. In the last two scenarios, there is no element of agreement, but undertakings are seen to tacitly collude with the help of algorithms. Greater focus will be on the Predictable Agent and the Digital Eye since these scenarios are not as easily tackled with current competition rules as the first two scenarios. In the fifth chapter, it is finally time to analyse certain measures to address algorithmic collusion. First of all, when it comes to fully autonomous pricing algorithms, these algorithms may constantly learn from past and current data to always optimise the best price for their masters. This optimisation may take a form of stable supra-competitive pricing harmful for the consumers. Now, in order to tackle these concerns, we may expand the interpretation of agreement or concerted practices to forbid the usage of harmful pricing algorithms that cause anti-competitive effects to the relevant market. Alternatively, the possibility of prohibiting tacit collusion with the help of clearly formulated criteria is examined. Both “by object” or “by effect” approach could be applicable. However, the effects-based approach would offer a more reasonable balance between the opposing interests of competition authorities and the undertakings. Overall, it is up to each jurisdiction to decide their own approach on how to tackle algorithmic collusion based on their political will for instance. After all, there is no one single measure that would be definitively correct in tackling this challenge