On 6 November, the French and German competition authorities published their long-awaited joint Algorithms and Competition study (and executive summary). It follows their 2016 joint study into Big Data.
The study focuses on algorithms which, in the authorities' view, may potentially impact competition. In particular, it focuses on the risks of collusion flowing from the use of such tools. In doing so it considers: economic theory; practical scenarios where algorithms could result in collusion; and practical challenges that antitrust regulators face when investigating algorithms.
| what you need to know - key takeaways |
- The growth of pricing software used by businesses, particularly in online markets, comes at a risk that such tools increase the risks of firms colluding in way which breaches competition law.
- As the application of competition law to algorithmic pricing develops, businesses should ensure that automated and machine-learning pricing technologies are compliant with competition law.
- Compliance training for those adopting and programming such technology is just as important as training the rest of the business.
- Businesses which adopt algorithms to assist in commercial decision-making should also monitor the effect of that automated decision-making to ensure that it is competition law compliant.
|
The theory: how algorithms might result in stable collusion
The study finds that the actual impact of the use of algorithms on the stability of collusion in markets is a priori uncertain and depends on the specific characteristics of the relevant market. For example, whilst certain algorithms might have the potential ability to coordinate prices, stable coordination is likely to be undermined in markets where algorithms are used to engage in personalised pricing (thus reducing transparency - a key factor of stability for a collusion). A similar view was reached by a study by the UK Competition and Market Authority in October 2018.
As regards the role of algorithms in the emergence of collusion, the study concludes that economy theory provides few insights into which kinds of algorithms are more prone to facilitate the emergence of tacit collusion.
Practical scenarios: where collusion might arise
The study goes on to examine three practical scenarios in which algorithms may enhance collusion:
- Explicit direct collusion: Where algorithms are used to facilitate the implementation, monitoring or enforcement of collusion which has been expressly agreed between competitors. Examples include, the UK Posters and frames cartel, and the European Commission Asus, Denon and Marantz RPM case. The use of algorithms in these scenarios is likely to reinforce the negative effects of anticompetitive collusive conduct (which may lead to a larger fine).
- Algorithm-driven collusion involving a third party: Where competitors have no direct contact with each other, but coordinate through a third party (e.g. where the third party provides similar pricing software solutions or advises on the design and use of pricing algorithms). In such cases the competitors' liability may depend on the extent to which they are aware of the illegal coordination being facilitated by the third party. A similar scenario was considered by the ECJ in the E-TURAS case in 2016 (see the EU section of our Ashurst publication Competition policy in the digital era: a comparative guide).
- Collusion induced by the parallel use of individual algorithms: The report considers whether algorithms could result in illegal coordination in circumstances where multiple competitors use distinct pricing algorithms which result in convergence towards the same price, but without the competitors having coordinated on such use. Instead, the algorithms may interact between themselves. The study concludes that there is still much uncertainty as to how algorithms would actually "communicate" in real market conditions, blurring the distinction between illegal coordination and legitimate parallel behaviour. However, the study stresses that it is a company's responsibility to consider how to ensure their tools are compliant.
Practical challenges faced by regulators
The study concludes that classical investigation tools (e.g. requests for information, inspections, interviews) will remain the primary tools used by the competition authorities to understand the aim and functioning of any price-setting algorithms. However, more in-depth analysis of an algorithm itself may also be conducted (e.g. analysis of the source code or simulation of the algorithmic behaviours).
Conclusion
Overall, the study does not provide any brand new insights regarding competition law risks associated with the use of algorithms. However, it is helpful that both regulators have outlined some factual scenarios where concerns might arise, providing some guidance as to where businesses might focus their compliance efforts.
As the use and power algorithms and linked technology develops, so will the regulatory focus and reassessment of the associated risk. For example, the study does not conduct a detailed assessment of when personalised pricing might risk infringing competition law (e.g. on grounds of discriminatory conduct). However, the authorities have expressed a need for "careful analysis" in that regard.
For an earlier summary of how these, and some other key jurisdictions around the world, are tackling the interface between an increasingly digitised economy and competition law, see our Comparative guide to competition policy in the digital era.
