International Competition Network’s scoping paper on “Big Data and Cartels”: Big Data and algorithms as a new “threat” and a new “tool” for cartel enforcement

Earlier this year, we discussed the potential implications of the COVID-19 pandemic on cartels and collaboration between competitors (see our e-bulletins here and here). Regulators have been open to allowing certain types of cooperation between competitors to address the impact of the crisis.  They have nonetheless been careful to stress that this is not general permission for "crisis cartels" to take place (see the statement issued by the European Competition Network here).  Indeed, regulators remain vigilant and have been focussing on new approaches to monitor sectors of the economy and uncover anticompetitive agreements without relying exclusively or primarily on leniency applications.  Amongst others, regulators have been concerned about collusion taking place through the use of sophisticated algorithmic tools (see an HSF paper on this topic here).

These issues have been increasingly at the forefront of regulators' thinking and this was recently underlined by the publication of a paper by the International Competition Network (“ICN”) which we briefly discuss in this blog post.

On 28 June 2020, the ICN published a scoping paper on “Big Data and Cartels” (the “Paper”), which aims at fuelling a discussion amongst ICN members about the implications of Big Data and algorithms for cartel enforcement. The Paper suggests that this discussion could lead to an update to the ICN’s Anti-Cartel Enforcement Manual.

Over the past few years, there has been increasing debate about the new digital dimension of cartel enforcement.  The Paper does not reach any definitive conclusions on this topic but identifies the key issues and practical questions that need to be considered to appropriately assess the interplay between digitalisation and cartel enforcement. The Paper looks at these issues from two different perspectives: the first part of the Paper analyses Big Data and algorithms as a “vehicle for collusion”, whilst the second part looks at them as a “tool for cartel detection”.

1. Big Data and algorithms as a vehicle for collusion

In the first section of the Paper, the ICN observes that algorithms can affect “two structural factors of collusion”, namely the frequency of interaction and market transparency, which could in turn facilitate (explicit and tacit) collusion given that competitors can use algorithms to “monitor each other’s prices and programme immediate reactions to any changes”.  The Paper further notes that Big Data and algorithms could be used in the context of a wider anticompetitive practice when this is facilitated or implemented “through means of automated systems”, thus increasing the chances of coordination, monitoring and punishing.  Such collusion, which supports or facilitates “typical” anticompetitive practices, may be harder to identify and may give rise to more complex cases.

The Paper observes that competition authorities may face various challenges where Big Data and algorithms are used as a vehicle for collusion.  In particular:

• Difficulties in establishing a “concurrence of wills” between competitors using digital tools to set and adjust prices.  According to the Paper, certain digital forms of interaction between competitors might come close to some form of conscious cooperation.  However, in the absence of any explicit contacts, this behaviour might fall short of demonstrating a “concurrence of wills” or “meeting of minds” between competitors (which is an essential element for establishing collusion).  Several questions arise in this regard, including whether the monitoring and adaptation of capacities provided by Big Data and algorithms could be taken into account for demonstrating the existence of a “concurrence of wills”, and whether the level of market transparency allowed by these technologies could change the methods of analysis of documentary evidence (e.g. by allowing authorities to infer the existence of an agreement from unclear or cryptic exchanges).
• Evidentiary standard to prove that firms have not acted independently. Establishing the intent to collude through the use of algorithms could also prove difficult in the context of Big Data.  According to the Paper, one of the key questions that would need to be assessed in this regard is whether mere “consciousness of the collusive outcome” is sufficient for the conduct at issue to fall within the provisions on anticompetitive agreements.
• Algorithms can achieve fast price- matching even in markets where traditional price fixing is unlikely to succeed.  In this vein, the Paper sets out a number of questions that would need to be considered.  For instance, should market structures be considered as potential evidence to demonstrate the manifestation of will (e.g. when the degree of transparency of the markets is known to all market players)?  Furthermore, should competition authorities systematically adopt an effects-based approach or should certain types of conduct be addressed through a “by object”/per se approach in the context of the “highly technical nature of Big data-related practices”?
• There are questions over the control of, and liability for, algorithms.  According to the Paper, a number of considerations must be taken into account in this regard – e.g. when shall an algorithm be deemed to remain under the company’s “control”?  Also, could the developer and the beneficiary of the algorithm be found jointly and severally liable for anticompetitive uses of the algorithm?  The Paper notes that this option could incentivise both the developer and the beneficiary to ensure that the use of the algorithm is compliant with competition rules.
• Monitoring capabilities and duration/continuity of practice. The Paper notes that in most jurisdictions case law requires “the demonstration of regular contacts” so as to establish the duration and continuity of the relevant practice.  In view of the monitoring capabilities offered by digital technology, the Paper raises the question of whether interaction between the parties is necessary from time to time, or whether the duration of the relevant practice is determined by the fact that the parties continue using Big Data or algorithms which align prices.
• Facilitators.  The Paper notes that competition authorities could also assess the situation in which an IT company knowingly provides a group of competitors with algorithms that allow illegal coordination between them.  In this vein, the key question is under which conditions the IT company would be viewed as a facilitator of the infringement.
• Tacit collusion. The Paper notes that the parallel use of identical or similar algorithms may result in the application of identical prices by competitors “in particular when algorithms include predictive features”.  Given that tacit collusion generally falls outside the scope of competition enforcement in most jurisdictions, the key question is under which conditions competition authorities may pursue such conduct.  According to the Paper, the intentional implementation of algorithms by competitors in order to tacitly collude should in any event be viewed as illegal.

2. Big Data and algorithms can be used as a useful tool for cartel detection

In the second (and much shorter) section of the Paper, the ICN identifies challenges to digital cartels detection and investigation, and looks at different ways in which Big Data and algorithms can be used by competition authorities as a tool to detect cartels.  In particular:

• Gathering and analysing large data sets during inspections.  Once data is gathered, competition authorities need significant IT capacities to efficiently store, index and search the relevant data, while the seizure of large data sets might also create issues regarding the treatment of legal privilege or protection of personal data.  These exercises require significant resources on the part of competition authorities.  In this regard, the Paper suggests that ICN members could share good practices regarding data management techniques.
• Localisation of digital information.  The Paper notes that many authorities adopt an “access approach”, meaning that data that is accessible to the company needs to be rendered accessible to the investigators and the failure to do so could result in fines and other sanctions.  However, other authorities adopt a “location approach”, meaning that they can only search data located within the premises or in specific places provided under the search warrant (this approach may not allow certain data, e.g. data stored on clouds, to be seized).  The Paper asks in this regard whether an “access approach” should be promoted and how such an approach could be consistent with legal systems based on the “location approach”.
• Cartel screening.  The Paper notes that data screening tools based on algorithms are used in some jurisdictions (in particular in relation to public procurement) and can help in the fight against bid-rigging.  In this regard, the Paper suggests that competition authorities exchange views on their existing data screening mechanisms to make their systems more efficient.

In addition, the Paper notes that competition officials could successfully cooperate with other agencies such as public procurement authorities and anti-corruption or anti-fraud bodies regarding the transfer and use of data gathered during the investigations of those bodies.  The Paper further highlights the importance of disseminating a “competition culture” to those who design and use tenders (e.g. by means of training and recommendations) as an effective means of mitigating the risks of bid-rigging.  Finally, the Paper notes that competition authorities could discuss with public authorities that are in charge of managing electronic procurement what types of data could be useful for the purposes of bid-rigging investigations.

The Paper is likely to focus regulators' efforts on uncovering cartels that take place through the use of Big Data/algorithms, whilst regulators will continue to make increasing use of such instruments themselves in the effort to detect and analyse cartels.  Companies should pay particular attention when deciding to use algorithmic tools that are aimed at monitoring the activity of their competitors. They should also be aware of the increasing risk of cartel detection as well as the risks with regard to COVID-19 related cooperation efforts straying into impermissible collusion.  Compliance manuals/training as well as audits in key areas of activity remain a helpful tool to reduce such risks.

Contacts