Competition Infringements Carried Out Through Pricing Algorithms

INTRODUCTION

In recent years, one of the most significant developments observed in the e-commerce, hospitality, transportation, advertising, and retail sectors has been the increasing role of algorithms, rather than humans, in the dynamic determination of prices. The widespread use of pricing algorithms, their impact on competitive market equilibria, and the legal and ethical implications of this development have become critical issues that warrant close examination, and have increasingly become the subject of scrutiny by competition authorities around the world.

A. Pricing Algorithms and Their Operating Principles

Algorithms are defined as methods, procedures, or techniques used to perform a task or solve a problem, and through the sequential instructions they embody, algorithms enable the execution of a task or the resolution of a problem. Algorithmic pricing, in turn, may be defined as the automatic determination of prices by computer software.

Broadly, there are two types of pricing algorithms: (i) rule-based algorithms and (ii) self-learning algorithms. In this context, rule-based algorithms determine the sales price in accordance with the instructions embedded within the algorithm. For example, commands such as "match the lowest/highest price available online," "match x% of the lowest competitor price," or "increase the price by y% once inventory reaches level x" may be coded into the algorithm, and the price of the relevant product or service is determined by the algorithm in line with these instructions. In essence, since the price is set based on the commands programmed into the algorithm, the algorithm merely performs a calculation consistent with those instructions; the price or price range itself is determined by the individuals who design and input the commands.

In the case of self-learning algorithms, no specific instructions are programmed into the algorithm; instead, the algorithm experiments in order to identify the optimal sales price and autonomously learns market conditions and the strategies required to achieve long-term profit maximization.

B. Competition Infringements Arising from the Use of Pricing Algorithms

The use of pricing algorithms by undertakings does not, in and of itself, give rise to any competition concerns. However, although the use of pricing algorithms may in certain cases produce positive effects that enhance consumer welfare, it may also, as a result of collusive arrangements between undertakings, give rise to effects that eliminate or restrict competition in the market.

1. Competition Infringements Implemented Through Anti-Competitive Agreements

One of the most common situations in which pricing algorithms give rise to competition-infringing effects arises where undertakings come together and, in breach of Article 4 of Law No. 4054 on the Protection of Competition, covertly determine sales prices, engage in market sharing, or agree on the use of pricing algorithms in order to increase sales prices, and subsequently use such algorithms as an instrument for implementing the agreement.

Competition infringements arising from pricing algorithms have not yet been subject to review by the Competition Board. However, in this regard, the case of U.S. v. David Topkins adjudicated in the United States examined a situation in which Topkins and his competitors agreed to increase the prices of certain posters sold on Amazon and used an algorithm, designed to collect competitors' pricing data and set prices above the competitive level, to implement this anti-competitive agreement. Nevertheless, as Topkins admitted to the charges during the proceedings, the case was closed without the initiation of a formal investigation.

Similarly, in the Trod/GB decision issued by the CMA, the use of an algorithm by two retail companies operating on the Amazon United Kingdom online sales platform, whereby the companies agreed not to undercut each other's prices unless a third-party seller offered a lower price, was assessed as a competition infringement, as the algorithm set prices in line with the parties' agreement. Accordingly, the CMA imposed an administrative fine on Trod.

2. Hub-and-Spoke Type Infringements

Through the use of pricing algorithms, hub-and-spoke type infringements may also arise, which can be defined as the establishment and maintenance of an anti-competitive equilibrium through the exchange of information via a common supplier or customer, without competitors directly coming together. In this context, hub-and-spoke infringements may occur where competing undertakings determine their pricing strategies by using the same algorithm, or where undertakings create an anti-competitive outcome by exchanging information through, or using information transmitted via, software positioned as a "hub," rather than through a common supplier or customer.

This issue was examined in the Eturas case, which was referred to the CJEU as a request for a preliminary ruling, in the context of the imposition of a common upper limit on discounts applied to services offered through an online reservation platform, the communication of such limit to agencies via a message, and the implementation of technical restrictions within the reservation system preventing agencies from offering discounts below the установлен upper limit. In its decision, the CJEU held that, unless the agencies that received the message publicly and clearly distanced themselves from the conduct expressed therein, informed the competent authorities, or provided evidence capable of rebutting the presumption that they acted in accordance with the message, it must be accepted that they participated in a concerted practice through the system.

It is therefore evident that, even where undertakings independently use a platform's algorithm, if they are aware that competing undertakings are using the same algorithm and that such algorithm fixes prices, an infringement similar to a classic hub-and-spoke arrangement will arise, and the undertakings concerned will be deemed to bear liability.

3. Infringements Implemented Through Resale Price Maintenance

Competition infringements may also arise where undertakings determine resale prices through the use of pricing algorithms. In this context, resale price maintenance may take the form of fixing a resale price or setting a minimum resale price. Pricing algorithms may assist supplier undertakings in detecting any deviations from the fixed or minimum price determined, or may directly transform a recommended resale price into a fixed resale price.

In this regard, in the CMA Casio investigation; the CMA held that Casio had infringed competition law by using software that monitored the resale prices of its distributors in the digital piano market and prevented resellers from offering discounts, and accordingly imposed an administrative fine on Casio.

4. Competition Infringements and Liability Arising from Self-Learning Algorithms

The use of self-learning algorithms also requires examination from a competition law perspective. In other types of infringements, rule-based algorithms are used and, since the relevant rules are in fact determined by the undertakings themselves, undertakings may be held directly liable where the prices set by the algorithm infringe competition law. However, as noted above, in the case of self-learning algorithms, the algorithm is programmed to maximize profit and independently determines the price that yields the highest profit as a result of its own experiments. Accordingly, although restrictive principles prohibiting unlawful conduct such as price fixing or market sharing may be incorporated, the algorithm remains free to engage in self-learning and experimentation, and undertakings do not exercise a decisive influence over pricing. For this reason, where prices determined by self-learning algorithms constitute a concerted practice or produce anti-competitive effects in the market, the question of who bears responsibility remains contentious. Although the Competition Board has not yet adopted a decision in this regard, there are views arguing that undertakings may be held liable, regardless of whether the algorithms are self-learning, where they fail to take the necessary measures to prevent algorithms from giving rise to anti-competitive coordination.

Nevertheless, there are also views arguing that where algorithms instructed to maximize profits directly engage in cartel-like conduct (such as price fixing, market sharing, or supply restriction), undertakings should be held liable even in the absence of a direct agreement between them. This is because it would be contradictory to argue that undertakings, while being held responsible for the actions of only a few among thousands of their employees, could at the same time be considered not responsible for the algorithms they deploy.

CONCLUSION

While pricing algorithms provide undertakings with significant efficiency gains and may enhance consumer welfare, they may also produce outcomes that restrict or eliminate competition. As pricing algorithms and the competition infringements they give rise to have not yet been examined by the Competition Board, it remains unclear which practices constitute infringements under Law No. 4054 and who should bear responsibility for such infringements. It is therefore of critical importance for undertakings to adopt the necessary safeguards to avoid anti-competitive conduct in the use of algorithms and to ensure that these technologies are deployed within ethical and legal frameworks. In this regard, it may be argued that the Competition Board should proactively examine this issue and develop preventive measures in order to safeguard market equilibrium.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.