Imperfect information, algorithmic price discrimination, and collusion*

Abstract

We analyze the ability of firms to sustain collusion in a setting in which horizontally differentiated firms can price discriminate based on private information. Firms receive private, noisy signals regarding customers’ preferences. We find that there is a non-monotonic relationship between signal quality and the sustainability of collusion. Starting from a low level, an increase in signal precision first facilitates collusion. There is, however, a threshold beyond which any further increase renders collusion less sustainable. Our analysis provides important insights for competition policy, particularly in light of firms’ growing reliance on increasingly sophisticated computer algorithms to analyze consumer data and to make pricing decisions. In contrast to previous findings, our results reveal that a ban on price discrimination can help to prevent collusive behavior as long as signals are sufficiently noisy.