Instantaneous Defection

Algorithms may make collusion easier, by reducing the cost of monitoring and reacting to the behavior of other participants in an oligopoly. But even the fastest of algorithms can’t guarantee collusion, contrary to what Stucke and Ezrahi seem (p. 62-65) to suggest. After all, the classic model of the failure of collusion, the prisoner’s dilemma, is a simultaneous move game! And the best algorithms can do for oligopolists is to allow them to act simultaneously.

Suppose all the oligopolists in a market have super-fast algorithms that allow them to monitor the decisions of competitors and adjust accordingly. Then as soon as one firm adjusts for another firm’s decision, that other firm will adjust to the first firm’s adjustment. As adjustment times fall, the firms become unable to adjust to each others’ decisions, because they end up effectively all making their decisions at the same time.

That is precisely the setup of the basic prisoner’s dilemma game, in which no player observes the decisions of the other players in advance. The game shows that under this condition, rational firms will all choose to betray the oligopoly and charge a low price, because each firm will know that it is in the best interest of each of the other firms individually to betray the group as well.

The prisoner’s dilemma tells us that even in a world of infinitely fast algorithms, tacit collusion can fail.

Stucke and Ezrahi argue (pp. 56-64) that algorithms would facilitate oligopoly in gas station pricing because algorithms would allow each station to meet a price-cutter’s low prices before consumers have had a chance to drive over to the price-cutter to take advantage of those low prices. As a result, consumers would not flock to the price-cutter, but would buy at the low prices offered by their local gas stations instead, eliminating the reward for cutting and ensuring that all members of the oligopoly charge a high price.

What makes algorithms effective as an aid to oligopoly in this story is the inability of consumers instantaneously to take advantage of a price cut. But here’s the thing: the same technology that makes quick price changes possible also makes quick purchase decisions possible. By offering consumers the ability to use an app to commit to a low price as soon as it is offered, or to have gas delivered to their car, a price-cutter can lock in volume generated by betraying the oligopoly.

That puts us back in the prisoner’s dilemma.

I don’t mean to say that algorithms never facilitate oligopoly pricing — they may well do so today for gas stations, because the technology that would allow consumers to fight back is still in its infancy. My point is only that there’s no necessary relationship between the speed of algorithms and harm to competition.