CMA research confirms pricing algorithms can facilitate collusion

• United Kingdom

• Competition, EU and Trade

11-10-2018

On 8 October 2018, the UK Competition and Markets Authority (“CMA”) published an economic working paper on the use of algorithms (“Working Paper”) by businesses operating on online markets and how they can be used to achieve a tacitly coordinated pricing outcome in breach of competition law.

In its investigation, the CMA found evidence of widespread use of algorithms to set prices, particularly on online platforms, where “many sellers” use them. As well as the simple pricing rules provided by the platforms themselves, the CMA discovered that some third-party firms sell more sophisticated algorithms to retailers or take on the role of price determination using computer models on behalf of their clients. A summary of the CMA’s findings, as set out below, will be useful for businesses to ensure that their use of algorithms complies with competition law.

What are pricing algorithms?

The CMA’s Working Paper defines an algorithm as any “computational procedure that takes some value, or set of values, as input and produces some value, or set of values as output.” A pricing algorithm uses one price as the input and uses a computational procedure to determine another price as the output. This definition captures algorithms used for price monitoring, price recommendation, price setting and price ranking.

Pricing algorithms can be used by a business to observe the prices of its competitors almost instantly, to detect any deviation from regular prices and to implement a price response in reaction.

How can algorithms raise competition concerns?

Algorithms can lead to consumer benefits including reducing transaction costs for businesses and giving consumers more information on which to base their decisions. Algorithms can, however, also give rise to competition concerns where they facilitate collusion between competitors.

By providing a business with updates on prices used by their competition, algorithmic pricing can be effective in facilitating price collusion in markets that are already susceptible to coordination. This is particularly relevant where the businesses’ offerings are homogenous, or in an online context where price monitoring and responses can happen quickly.

A pricing algorithm which leads firms to adopt very simple, transparent and predictable pricing behaviour, such as price matching or price cycles, can indicate market coordination between competitors. Another key indicator is the prevalence of similar pricing algorithms on the same market.

The Working Paper explains that pricing algorithms can lead to “tacit coordination” resulting in coordinated outcomes even when each business uses the pricing algorithm to make unilateral pricing decisions. For tacit coordination to take place, the algorithm must be willing to sacrifice short-term profits in favour of longer-term, more profitable outcomes.

On the other hand, where the function of an algorithm is very short-term and designed to maximise profit on each sale with no regard for the impact on future profits, the CMA advises that the algorithm is less likely to lead to coordination.

The CMA can audit the objective function of the algorithm to determine whether it is capable of tacit coordination.

Using an algorithm to achieve a tacitly coordinated outcome

The Working Paper highlights three ways in which an algorithm can result in a tacitly coordinated outcome.

• Hub and spoke: This can occur when a third party provides an algorithm to multiple businesses, gaining access to data and an understanding of several suppliers’ pricing policies. A concern could arise if this gives the platform or ‘hub’ the ability and incentive to increase prices above the competitive level in order to maximise collective profits.

• Predictable agent: Competition concerns arise when algorithms react to events in a predictable way that allows competitors to work out what is going to happen, which increases the likelihood of achieving a tacitly coordinated outcome.

• Autonomous machine: Some algorithms become so complex and sophisticated that, when given an objective to maximise products, the algorithm can learn by itself and reach a tacitly coordinated outcome. This can happen without the intention by its human owners to collude and there is a limited possibility of discovery by regulators such as the CMA.

From these three theories of harm, the CMA considers that the hub-and-spoke concern is likely to give rise to the most immediate risk, as it only requires businesses to adopt the same algorithmic pricing model. The other two theories depend on pricing algorithms to become sufficiently advanced and wide-spread.

In terms of next steps, the Working Paper does not envisage any immediate enforcement action or specific market studies. Instead, the CMA considers that there could be value in conducting further economic research into pricing algorithms. The Working Paper outlines four possible areas of research.

• Auditing algorithms: A business can use an auditing algorithm to determine whether the increase of profits following the use of a pricing algorithm is the results of attracting new customers, increasing sales to existing customers or engaging in collusion.

• Algorithms that should be presumed to be anti-competitive: The CMA wishes to find out whether there are certain kinds of “decision rules” used by algorithms which have no plausible purpose other than to facilitate an anti-competitive outcome.

• Secret offers and masking: Customers may be able to use countermeasures such as other algorithms or “shopbots” to undermine collusion. It would be useful to investigate to what extent customers can build up and exercise buyer power through joint purchasing.

• Replicating studies using UK data: The CMA hopes to replicate existing studies and apply these to UK data in order to conduct a more in-depth and conclusive examination of online personalised pricing and search discrimination in the UK.

Comment

The CMA’s Working Paper follows studies undertaken by other competition authorities on pricing algorithms and the risks they pose from a competition law perspective. Indeed, last year, the Organisation for Economic Co-operation and Development (“OECD”) published a paper on Algorithms and Collusion. Furthermore, EU Commissioner Margrethe Vestager stated that “one of the biggest things we need to deal with is the risk that automated systems could lead to more effective cartels”¹.

Businesses that use pricing algorithms should ensure that they understand how their algorithms operate in practice and that they are designed to comply with competition law, otherwise they could face severe penalties. As Commissioner Vestager stated, businesses “need to know that when they decide to use an automated system, they will be held responsible for what it does. So they had better know how that system works.”

1. Speech on Algorithms and Competition Law (16 March 2017) available here.