Learning to play Cournot duopoly strategies

The paper reports results from experiments designed to determine whether subjects can learn to play Cournot duopoly strategies and whether their out-of-equilibrium play is consistent with the predictions of learning models. The experiments include duopolies with constant and with decreasing marginal costs, and with theoretically stable and unstable equilibria. After a few periods, subjects do play stable interior equilibria but they do not play stable boundary equilibria nor unstable interior equilibria. Subjects' out-of-equilibrium play is inconsistent with the predictions of the learning models.