Algorithms are increasingly taking over pricing decisions for companies. They are ideally suited to deal with the wealth of data on competitors and customers and can respond immediately to changes in the market environment.
A major problem with pricing algorithms, however, is that without explicit collusion, they can prevent competition among firms, leading to inflated consumer prices. Unlike explicit price collusion among humans, self-learning reinforcement algorithms (such as Q Learning) leave no evidence for competition authorities to exploit.
This project analyzes pricing algorithms in the presence of "hybrid" interaction between human actors and algorithms. So far, higher price levels have only been achieved when algorithms compete exclusively against algorithms. However, the impact of algorithmic pricing has not been well studied when algorithmic and human sellers collide.
Our research shows that algorithms can be anti-competitive. In markets with two firms that both use algorithms, prices are always higher than in human markets. Even in markets with three firms, algorithms can contribute to higher price levels. The algorithms develop strategies that promote cooperative behavior with humans in hybrid markets. Our results highlight the dangers that pricing algorithms can pose compared to markets with human decision makers.