Pricing Algorithms May Be Creating Unintended Cartel-Like Outcomes
A recent study highlights how automated pricing software can lead to cartel-like outcomes without any collusion or human design. In Germany, when two gas stations in the same market adopted such algorithms, their profit margins increased by about 38%, suggesting that independent systems learned not to undercut each other for higher profits. This phenomenon is difficult to detect because it doesn't involve communication or data sharing between competitors. The Federal Trade Commission's lawsuit against Amazon centers on a tool called Project Nessie, which allegedly raised prices in anticipation of competitor reactions and generated over $1 billion in excess profit before being paused during scrutiny. Amazon denies these claims but admits the system was discontinued years ago. The issue is particularly concerning because algorithms can optimize for margin without executives realizing they are suppressing competition. Regulators face challenges addressing this problem, as it lacks traditional collusion indicators like meetings or data exchanges. Three scenarios illustrate how competition may disappear: independent learning by algorithms (the 'ghost'), unilateral price-setting based on predicted rival behavior ('mirror'), and shared algorithmic guidance from a common provider ('hub'). The first scenario is the most difficult to detect and legally address.
Tickers: $AMZN
Topics: Amazon artificial intelligence cartels FTC lawsuit pricing algorithms