AI Pricing and the Case for Permissive Competition Law Enforcement

How will algorithmic pricing driven by AI (‘AI pricing’) affect antitrust enforcement?  AI pricing increases significantly the probability of coordinated, supercompetitive pricing, even in markets with many firms.  Technology will be able to predict rivals’ behaviour and will be able to react immediately to new information, including price cuts by a rival, in a manner that will stabilize parallel but independently-set supercompetitive prices.  While conscious parallelism has always been a threat to competition, the spread of AI pricing will mean that many unconcentrated markets will experience uncompetitive outcomes.  Agreements between competitors to rely on a particular AI pricing algorithm to coordinate pricing would be illegal, but firms will be able independently to adopt AI pricing technologies that will themselves recognize the value of cooperation over competition. 

In a forthcoming paper in the University of Chicago Law Review Online, I consider the impact of ascendant AI pricing technology on enforcement approaches to mergers and abuse of dominance.  I agree with commentators such as Michal Gal and Daniel Rubinfeld that AI pricing should in the short run lead to stricter merger review.  Mergers increase the risk that AI pricing—while relatively crude at present—will facilitate cooperative outcomes in markets previously thought to be invulnerable to cooperative pricing.  In the medium to longer term, however, AI pricing ought to lead to a more permissive approach to merger review. 

As AI pricing develops and becomes more and more effective in producing cooperative outcomes, a merger is less and less likely to affect pricing in a market: even if the market is unconcentrated, something approaching monopoly pricing in many markets will arise.  As a consequence, the market without a merger will be as uncompetitive from a pricing perspective as with a merger.  Given that mergers will have a weaker effect on pricing, highly developed AI pricing technology calls for permissive, not strict, mergers policy.

The policy logic of a permissive approach is clear, but does not necessarily sit well with current enforcement approaches.  The Merger Guidelines in the US, for example, provide that the agencies will be more likely to challenge a merger if there is evidence of successful tacit or explicit collusion in the market pre-merger.  This begs the question: if the market is uncompetitive even without the merger, why challenge the merger?

There are two reasons that support the status quo approach to uncompetitive, no-merger counterfactuals, both of which will lose force over time.  One is that while supercompetitive pricing may have been sustained without a merger, such cooperative understandings are vulnerable to breaking down, and a merger would reduce the chances of breakdown.  With sophisticated AI pricing, however, it is predictable that intelligent algorithms will avoid a breakdown of cooperative pricing.  The threat of immediate retaliation by rivals following a price cut, for example, would eliminate the incentive to cut price.  With minimal risk of cooperative understandings breaking down with or without the merger, there is little reason to block a merger hoping for an outbreak of competition.

Another reason for the current approach is that while firms may be able to sustain supercompetitive pricing absent the merger, the merger may allow the firms to set even higher prices.  Intelligent algorithms, however, will be better able to achieve monopoly pricing with or without the merger, both through rapid learning from experience and by relying on data on costs and demand to calculate profit-maximizing prices.  A merger is therefore less likely to affect price and ought not to be blocked out of price competition concerns.

To be sure, not all markets will be prone to cooperative outcomes with AI pricing.  Markets in which there is a large lag between sales will be vulnerable to price-cutting by firms hoping to gain sales at the expense of their rivals.  A massive defence procurement contract, for example, may arise only occasionally, and bidders will be tempted to undercut rivals knowing that any retaliation for doing so will be years away.  But in other contexts, AI pricing will render uncompetitive markets in which competitive outcomes would result today.

Rather than relying on mergers law, as AI technology improves, there may be a justification to shift to forms of price regulation.  As AI pricing technology develops, there is a greater chance that antitrust authorities could sensibly rely on AI to require competitive pricing – AI may help authorities determine the competitive price benchmark.  The current, persuasive argument that regulating conscious parallelism is impractical because it is tantamount to costly and inaccurate price regulation will potentially lose power over time.  Note, however, that mergers policy ought to be lax either way, either because the authorities can keep prices competitive in any event, or because prices will be uncompetitive in any event.

Similar logic also calls for a more permissive approach to abuse of dominance as AI pricing develops.  Consider the canonical case of an incumbent monopolist seeking to exclude an equally efficient potential entrant selling a similar product.  With powerful AI pricing technology, even if the entrant succeeds in entering the market, this will not have an impact on pricing: the firms will cooperate by relying on AI pricing and will set monopoly prices.  There is no efficiency reason to prevent such exclusion through abuse of dominance enforcement.  Indeed, with lax mergers policy in place for the reasons given above, it would be better to allow one firm to compete in the market—if a more cost-efficient entrant comes along, it will simply acquire the incumbent, leading to a gain in productive efficiency and no loss in allocative efficiency given the existence of monopoly prices with or without multiple firms.  (There is a limit to this logic, however: if the entrant is sufficiently more cost-efficient, it may be that monopoly pricing would not prevail in the event of entry since the entrant might prefer to undercut its rival’s price rather than cooperate, and in this case the authorities ought to intervene to prevent exclusion.)

To conclude, AI pricing will create cooperative outcomes where today there would be competition.  This ought to lead to less strict antitrust enforcement against mergers and abuse of dominance since pricing in many markets will be uncompetitive with or without antitrust intervention.  But this does not mean that antitrust becomes irrelevant.  In some markets, conventional pricing considerations will continue to apply.  Moreover, there is a strong and strengthening case for competition authorities to shift focus away from static efficiency considerations like monopoly pricing, and towards the critically important question of dynamic efficiency and innovation.  While the relationship between competition and pricing is more easily understood, it is arguable that the relationship between competition and innovation is of greater economic significance.  AI pricing provides further impetus for competition law enforcers to attend to innovation and dynamic considerations, rather than pricing and static considerations.

Edward M. Iacobucci is Professor and TSE Chair in Capital Markets, Faculty of Law, University of Toronto.

This post is part of the series ‘How AI Will Change the Law’. The other posts in the series are available here.