Georgetown Law
Technology Review

When algorithms fix prices instead of CEOs, does antitrust law have an answer? Scholars have largely lamented the so-called liability gap, dismissing algorithmic tacit collusion as lacking an ‘agreement’ between competitors and thereby evading the watchful eye of the antitrust laws. This Article challenges that premise, developing a legal theory on which algorithmic tacit collusion falls within the reach of the antitrust laws.

The assumptions that justify permitting traditional tacit collusion—where companies coordinate prices without an express agreement, simply by watching and matching each other’s behavior—do not apply to collusion by algorithms. While tacit collusion is in principle characterized by the same incentives and anticompetitive outcomes as an express agreement, it is lawful because no conceivable antitrust rule can prohibit it without generating error costs—that is, without inadvertently deterring otherwise economically efficient behavior. First, the problem of proof: evidence of parallel conduct alone is ambiguous, supplying no clear evidence of restraint of trade—a shift from competitive to monopoly prices—rather than an independently rational response to changing market conditions. Second, the problem of remedy: the only conceivable remedies—judicial price regulation or limits on public price disclosure—are more costly than beneficial.

Pricing algorithms help courts circumvent both problems. Because algorithms follow predictable computational rules, the strategic logic underlying firm behavior is observable through technical or economic tests, solving the problem of proof. Moreover, courts can instruct firms to avoid anticompetitive pricing algorithms without imposing price controls, suppressing public price information, or chilling the use of procompetitive pricing algorithms, solving the problem of remedy. Algorithmic tacit collusion therefore falls within the doctrinal category of tacit agreement: coordination that falls short of an express agreement but is facilitated by some demonstrably inefficient and practicably remediable mechanism.

This Article further refutes the claim that an unlawful agreement requires some formal element—collective decision-making, nonpublic information, or shared intent—that various forms of algorithmic tacit collusion lack. None of these features are necessary or sufficient conditions of an unlawful agreement; treating them as such contradicts the welfare-oriented principles of the antitrust laws. This internal inconsistency between the economic rationale of the agreement doctrine and its formalistic application to algorithmic tacit collusion is termed here the algorithmic antitrust paradox.