Photo by Taylor Vick / Unsplash
Algorithmic Collusion Without Communication
When a market is restricted to a handful of large firms, each firm has enough market share to affect the revenue of others. This situation is known as an oligopoly market, and results in each firm having to behave strategically. A market with many small firms, by contrast, sees firms behave independently because no single firm is large enough to impact the revenue of its rivals. These monopolistically competitive markets allow firms to make decisions without considering the behavior of competitors.
In monopolistically competitive markets, a single firm’s decision on output or price does not affect rivals because its market share is too small. Theoretically, even if one firm drastically reduced its prices, there would be enough total customers in the market to allow other firms’ revenues to continue as normal. Oligopolies, however, have firms large enough to substantially affect the number of customers enjoyed by the others. If one firm drastically lowers its prices in an oligopoly, it is large enough to draw significant numbers of customers away from all of its rivals.
Oligopolies and Collusion
If one oligopolist raises its prices, its rivals are large enough to absorb most (or all) of its fleeing customers. This keeps oligopoly prices lower than one might expect - most consumers know of similar substitutes. However, if multiple oligopolies agree to raise prices, a practice known as collusion, many consumers will be stuck with having to pay the higher costs. Collusion is illegal, but can be somewhat difficult to prove depending on the methods of oligopoly price-adjusting.
Cartel and Output Restriction
Studies of oligopolies and collusion typically focus on attempts to create cartels, or oligopolies where all firms work together to set a reduced market output that results in higher prices (supply shifts left, causing market price to rise). Fortunately for consumers, this is usually less successful than planned - cartel members often cheat on the collusion agreement by overproducing to make more profit. As a result, many cartels have less market power than intended.
Price Fixing
A second method of collusion involves simply charging higher prices. After all, if all firms in the market charge a higher price, they will all make more profit! The process is illegal, as are agreements to restrict output to raise prices “naturally”. Price fixing may be easier to prove than output restrictions because an objective monetary amount will have to be agreed upon by participating firms, creating a paper trail. Restricting output can be done without an objective value. Setting a “fixed price” among firms makes them more vulnerable to whistleblowers who may reveal the plot to regulators.
“Unintentional” Collusion? AI Modeling
Until now, collusion has been planned by corporate executives, managers, and major shareholders in pursuit of greater profits. As firms increasingly use AI modeling to set optimal output and prices, however, the possibility of AI collusion emerges. Could AI modeling programs at large firms in oligopoly markets begin to collude to maximize profits, with no human direction or interference? This scenario is possible under game theory conditions.
Game Theory and AI
Firms will set their AI models to maximize profit. Inevitably, these models will discover that when they and competitors all set higher prices, they all make more profit. Thus, barring legalities, the AI models will begin seeking to set collusive high prices. These models may also develop “punishments” to dissuade others from dropping their own prices to undercut the prevailing high market prices. This punishment would likely be a form of price war, where other firms drastically lower their prices to undercut the undercutter.
To avoid the pain of a price war, all AI models may tacitly agree to pursue higher prices than market-clearing equilibrium, thus forcing consumers to pay higher prices than would normally occur. Reinforced by running thousands of simulations, all AI models may independently decide that defaulting to higher prices is the optimal choice, as it eliminates the risk of triggering a price war and sends a signal that other oligopolists can safely raise their own prices. Ultimately, each oligopolist’s AI model independently settles on a constant “high price” mode, making it the Nash equilibrium strategy of all firms in the market.
Limiting AI Collusion
Despite AI models coming to “collude” independently, based on economic strategy alone, consumers and non-colluding firms will still be harmed. Regulators will have to adjust their policies to include the likelihood of AI models behaving identically to human colluders. This may include requiring firms in oligopoly markets to incorporate coding that limits AI’s tendency to raise prices. In a more extreme maneuver, regulators like the Federal Trade Commission (FTC) might ban firms from using AI modeling to set dynamic prices altogether.
Difficulties of Regulating AI
Unfortunately for regulators, the rapid proliferation of AI makes it difficult to pinpoint whether a firm is using AI to determine its prices and output. After all, firms don’t need to have their own servers running AI software - individuals can utilize AI subscription software on cloud-based services from the comfort of their own homes. Thus, regulators may struggle to prove that an oligopolist is using AI at all; corporate officers would almost certainly not subscribe to the AI modeling software under their own names!
Societal Welfare Implications
AI collusion would harm consumers by subjecting them to higher prices. Because all firms in the market are colluding through their AI models, customers are unable to find substitutes for any given brand at a lower price. In addition to colluding to set higher prices, AI models will also likely harm consumers by eliminating consumer surplus and charging consumers the highest prices they are willing to pay. The combination of collusive pricing and price discrimination will doubly harm consumers.
The ability of AI-using firms to charge consumers the maximum possible price at all times could lead to market collapse in the long run as consumers “give up” and stop seeking to maximize their productivity. After all, why seek to earn a higher income if all the extra money will be quickly drained by price discrimination? Consumers will also shop less as they see it impossible to get a “deal” on anything, viewing shopping as always being taken advantage of by powerful companies.
