Current technological developments in the field of artificial intelligence (AI) have added further complexity to the discussion of whether, in the absence of overt communications, mere tacit coordination between competitors should be outlawed. Whereas some commentators argue that the dangers posed by AI should tip the balance towards making tacit coordination illegal, there are others who are either not entirely persuaded of the plausibility of such dangers or who point out that a competition rule focusing on mere inter-firm interdependence is not administrable.
This paper, available here, reviews this debate with a view to establishing whether successful price coordination achieved by self-learning algorithms should be punishable under EU competition law, and whether the current regulatory framework is suitable.
Section 2 explains how AI relates to antitrust.
AI is expected to arise from certain types of software algorithms. An algorithm is merely a specified sequence of steps for producing a solution to a problem. Software is a composition of individual algorithms (written in a programming language) that solve specific problems. In a market context, an algorithm could take the form of a series of steps to arrive at a best price according to a profit maximisation objective. This requires the algorithm to reflect an accurate supply and demand model incorporating a variety of variables.
Artificial intelligence has automated this process via machine learning. In essence, AI powered software adjusts the model automatically according to the data with which it is fed – to a point where it does so faster, and processing more variables, than even the best statisticians are able to. For the purposes of this paper, the question is whether machine learning technology can aid in solving uncertainties that oligopolists need to overcome in order to tacitly establish a price above marginal cost.
Section 3 discusses EU law on tacit coordination.
If the use of algorithms leads to a coordination of prices between different undertakings, Art. 101 of the Treaty on the Functioning of the European Union (TFEU) might apply. Article 101 TFEU prohibits coordination between undertakings and spells out three alternative modes of coordination as forbidden, namely ‘‘agreements’’, ‘‘decisions by associations of undertakings’’ and ‘‘concerted practices’’. The existing provisions against price fixing capture several, but not all, scenarios where the use of algorithms leads to price coordination between competitors. There are, nonetheless, several instances where pricing algorithms lead to higher prices and harm consumers, but where the application of competition law is not possible or is disputed.
This is particularly the case as regards concerted practices. The prohibition of concerted practices addresses coordination between enterprises which falls short of being qualified as an agreement, but which nevertheless establishes practical coordination between the undertakings and knowingly substitutes practical cooperation between them for the risks of competition. A concerted practice requires some form of coordinating behaviour, subsequent conduct on the market, and a causal relationship between these two elements. Since the causal relationship between concertation and market implementation can be presumed upon evidence of concertation (e.g. by participating in an exchange of information), the burden of proof then shifts to the undertaking to show that it did not engage in anticompetitive conduct.
However, mere parallel behaviour or purely unilateral measures do not amount to anticompetitive conduct. For example, the rational and independent conduct of companies in oligopolistic markets can lead to parallel behaviour. This is often referred to as tacit collusion. Algorithmic collusion raises similar difficulties to those identified in discussions on the demarcation between forbidden tacit collusion on the one hand, and permissible independent parallel behaviour on the other.
Section 4 provides a concise overview of oligopoly theory.
In an oligopolistic market, the oligopolist may be in a position to maximise its profits by taking into account the reaction of competing oligopolists. Strategic behaviour in an oligopoly situation, and its effects on the market and on welfare, have been analysed by a large volume of academic economic literature.
Economic theory and empirical research have identified business strategies and market conditions where, even absent explicit collusion, price levels above competitive prices can be achieved in oligopolistic markets. For example, it has been shown that a ‘‘tit-for-tat strategy’’ where a player repeats the strategy of its competitor from the preceding round can lead to behaviour akin to collusion. Similar effects can be observed when one enterprise follows a leader with regard to price or quantity, or when a ‘‘trigger strategy’’ is implemented – i.e. when a punishment is imposed upon a player who deviates from a supra-competitive pricing strategy, with a view to creating an incentive for cooperative behaviour.
Further, economic theory has identified conditions and market characteristics which favour the emergence of supra-competitive price levels in oligopolistic markets – such as market transparency, low number of competitors, high number of consumers, high number of repeated interactions, stable demand and cost conditions, product homogeneity and information exchange. These conditions and market characteristics lead to a type of coordination which does not require implicit or explicit agreements. Further, for coordination to be durable, any deviation needs to be detectable by other market players and there needs to be a credible threat of retaliation that outweighs the benefits of cheating.
Section 5 discusses whether tacit coordination should be illegal.
The main challenge for competition law in oligopolistic markets consists of identifying the point at which coordination amounts to a forbidden concerted practice. Since tacit coordination – also referred to as tacit collusion and interdependent pricing – can lead to supra-competitive prices, there has been some controversy as to whether the law on concerted practices should be changed in order to address kind of conduct.
There are a number of consensual matters underlying this debate. First, social harm does not differ depending on whether the joint profit-maximising price is achieved by overt communications or a tacit mutual understanding. Second, the perceived prevalence of pure interdependent behaviour is circumscribed to a rare set of circumstances that include highly concentrated industries, homogeneous goods, symmetric cost structures across firms, and price transparency.
The debate concerns primarily the limits of the rule to punish oligopoly pricing. Views range from advocating that only those arrangements where communications between competitors have taken place should be prohibited, to a broad prohibition rule that encompasses all kinds of coordinated pricing, including pure interdependence.
Arguments for requiring communication between competitors before finding a competition infringement include: (i) internal consistency of the legal system – since monopoly pricing is not outlawed, then neither should mere interdependent pricing; (ii) administrability – an injunction on tacit price coordination would in practice establish public utility-type price regulation by competition authorities for all types of industries; (iii) the risk of over-enforcement – a rule which focuses on mere interdependent behaviour would yield too many false positives.
Against this, it has been argued that: (i) consumer harm is the same with or without communication between competitors – game theory does not postulate that steady-state price depends on whether firms have been able to communicate orally or in writing; (ii) oligopoly can be more damaging than monopoly – the incentives of oligopolistic firms to invest in product or cost-reducing innovations are weaker than for monopolists, because oligopoly profits depend on refraining from competing to a certain extent; (iii) a rule against oligopolistic interdependence can be administrable – common antitrust wisdom dictates that direct evidence of communications between executives should not be required because the fact that they are illegal provides the incentive to hide them. Instead, enforcers should be able to make inferences from factors such as supra-competitive prices and market structure.
Section 5 also discusses the implications of AI-powered predictive pricing to discussions concerning whether tacit collusion should be illegal.
From a consumer welfare perspective, predictive pricing can have advantages. First, it reduces search costs: if a firm is able to make an accurate prediction on what the customer wants, customers may benefit from this. Second, predictive pricing may allow firms to serve more customer segments, expanding the number of consumers who can purchase a good or service. Third, AI can be more effective in differentiating passive from active consumers, and may therefore allow firms to compete more (effectively). However, there are also a number of competitive risks associated with the use of AI in pricing decisions. An important one is that algorithmic pricing may significantly increase the number of situations in which tacit collusion will occur. Price lags will tend to disappear, since pricing software can react instantly to changes from competitors. Price interdependence will also be stabilised by eliminating irrational reactions from firms, such as arbitrary management decisions. By reducing uncertainty, price coordination will become easier to achieve even in the absence of communication between firms.
If one were to develop a rule against oligopolistic pricing, how should one go about it? It has been suggested that interdependent pricing can be identified through sticky prices. However, if algorithms can reduce information and coordination costs (by quickly predicting what the other competitors’ response to a price change will be), then competitive and interdependent pricing patterns will tend to look the same. In addition, faster price changes may make price signals harder to detect. An alternative would be to devise an algorithmic pricing incubator that the public authorities can use to make simulations and discern whether firms are pricing competitively.
A different matter is whether, should one adopt a rule prohibiting (certain forms of?) interdependent pricing, it will still be rational to automate pricing software with AI tools in order to realise their efficiency benefits. This is a matter that goes not only to the appropriate level of intervention given the potential benefits to consumers of using AI pricing tools, but also to what tools competition enforcers should deploy against interdependent pricing. For example, fines and damages claims may pose challenges akin to those raised by excessive pricing as regards the identification of supra-competitive prices, i.e. freezing competition by deterring pro-competitive conduct.
Reflecting this, some authors have proposed alternative remedies to address tacit coordination by algorithms, e.g. by imposing structural or behavioural remedies, by reducing the frequency of price changes in order to make price cuts more profitable, by reducing price transparency for the algorithms but not for consumers, or by testing whether market structure influences the ability to collude in merger review. Other authors suggest dialogue with stakeholders and competition advocacy. Still others suggest that the market can take care of the problem more efficiently than public interventions, e.g. by promoting the use of AI tools by consumers. Lastly, it has been suggested that authorities could order firms to program their algorithms so as to play competitive instead of cooperative games.
Determining the relative merits of different competition policy responses is a work in progress, which the authors are currently pursuing.
I think this is a very interesting paper, even if I have to complain about its editing, which is quite shoddy.
Despite the popularity of the OECD’s work on algorithmic collusion, it has struck me that this enthusiasm was related to the practical challenges of enforcing traditional competition rules. Lurking further from view lies the much more challenging issue of whether our rules are fit for purpose in a world of algorithmic decision-making powered by big data, as was discussed in the OECD’s excellent background paper*. The present paper provides a good introduction to the matter, and contains a useful overview of the literature on the topic – before going on to make a number of speculative attempts at identifying potential problems and solutions raised by autonomous algorithmic pricing.
* I played no role in writing it.