We present and analyze results from a laboratory experiment in which groups of inexperienced subjects are repeatedly asked to make forecasts of the behavior of other members in their group and must then decide which of two types of player they want to be, "Type A" or "Type B." The Type A players are "predators" while the Type B players are producers. Players make these decisions using their forecasts and a payoff table that provides them with expected payoffs conditional on their forecasts. Those choosing to be Type B players (producers) must also decide the fraction of their unit time endowment to devote to "Action 1" - the defense of their production - with the remaining amount of their time going to "Action 2" - actual production. The model has a unique Nash equilibrium outcome where there are positive numbers of both predators and producers. There is also a utopian, "no predation" outcome, which is not stable under best response dynamics. We find that subjects quickly move away from the utopian outcome, but have difficulty coordinating on the unique Nash equilibrium outcome, because producers choose too low a level of defense, and, as a result, too many players choose to be predators than would be predicted in the unique Nash equilibrium. In a second treatment, we introduce a (benevolent) government "agent" - an additional subject - who has the power to set the fraction of time that all producers devote to defense. This government agent earns the same payoff earned by each producer (each Type B player). We find that the addition of the government agent leads to a substantial improvement in agents' forecasts and enables them to perfectly coordinate on a payoff superior Nash equilibrium where the fraction of time devoted to defense is high, but all players choose to be producers.

Co-author Minseong Kim