Two cement producers decide how much to ship into the same region, two airlines set fares on the same route, and in a third industry one firm builds capacity before its rival responds. The profit of each depends on the choices of all, and the form that competition takes decides how close the market comes to the outcome under monopoly or under perfect competition.
The game tests the predictions of the three basic models of oligopoly, and what repetition does to them. It shows how the number of firms moves a market between collusion and competition, how competition in prices differs from competition in quantities, and what a firm gains by moving first. With fixed markets over several rounds it also shows whether firms manage to collude without talking.
Under Cournot competition each firm chooses a quantity, taking the quantities of the others as given. With linear demand and constant cost, each of a number of firms produces the difference between the demand intercept and the cost, divided by the slope and by the number of firms plus one. With the default settings a duopolist produces 30 and the price is 40, where a cartel would produce 45 in total at a price of 55. Under Bertrand competition with identical products the firm with the lowest price takes the market, so two firms are enough to drive the price to cost, whereas prices stay above cost when products are differentiated.
In the Stackelberg model the leader commits to a quantity before the follower chooses. The leader produces the monopoly quantity, 45 with the default settings, the follower produces half as much, and the leader earns more than under Cournot. The contrast between the Cournot and Bertrand predictions is narrowed by Kreps and Scheinkman (1983), who show that firms which first choose capacities and then compete in prices reach the Cournot outcome.
The number of firms has the effect that the title of Huck, Normann and Oechssler (2004) announces. In their repeated Cournot markets, duopolies showed some collusion, markets with three firms were close to the equilibrium, and markets with four or five firms were never collusive and settled at or above the equilibrium quantity. Earlier experiments on such markets are reported in Fouraker and Siegel (1963) and Holt (1985).
Information matters as well, since Huck, Normann and Oechssler (1999) found that telling firms the quantities and profits of their rivals made markets more competitive, since some firms imitated whoever had earned the most, while information about demand and cost made markets less competitive. In price competition, Dufwenberg and Gneezy (2000) found prices above cost with two firms and close to cost with three or four. In sequential markets, Huck, Müller and Normann (2001) found higher total output than under Cournot, with leaders producing less and followers more than the model predicts, which the authors attribute to followers' dislike of ending up behind.