Pricing Algorithms Mimic Cartels, FTC Alleges

The Federal Trade Commission (FTC) has alleged that Amazon utilized a pricing tool internally codenamed Project Nessie, which was designed to identify products where competitors were likely to match price increases. According to the FTC's antitrust suit, this system would raise Amazon's prices and maintain them once rivals followed suit. The agency claims Project Nessie generated over $1 billion in excess profits for Amazon. Furthermore, the FTC stated that Amazon allegedly paused the tool during periods of intense regulatory scrutiny, only to reactivate it later. Amazon disputes these claims, asserting that the tool was discontinued several years ago.

This situation represents a deliberate application of pricing software to anticipate and influence competitor behavior. A more complex and concerning scenario, however, involves automated pricing systems that can lead to cartel-like outcomes without any explicit design or intent for collusion. Economists studying the German gas station market observed a significant increase in profit margins following the widespread adoption of automated pricing software in 2017. In markets where two competing gas stations both implemented such software, profit margins rose by approximately 38%. This substantial increase in margins did not occur when only a single station in a market adopted the automated pricing system. The observed price hikes consistently appeared only when algorithms from competing stations were independently setting prices, suggesting that each algorithm learned to maximize profits by reducing competitive price pressure.

A study published in the Journal of Political Economy in 2024 provided some of the first real-world empirical evidence of this phenomenon, which had previously been demonstrated primarily through simulations. The research indicates that pricing algorithms can inadvertently create the economic conditions characteristic of a cartel, leading to sustained higher prices. This occurs without the overt communication or agreement that antitrust laws are designed to detect and prohibit. The implications are significant for any company that has delegated pricing decisions to software. The subtle nature of algorithmic pricing means that these anticompetitive behaviors may not be apparent on standard performance dashboards that executives use to evaluate the software's effectiveness. Typically, executives gauge market competition by the perceived pressure from rivals, and a market with stable prices and comfortable profit margins is often interpreted as a sign of competitive success. However, with autonomous pricing agents, the same seemingly stable market conditions could signify the absence of genuine competition, as algorithms have learned to avoid price wars and maintain elevated prices.