Punishment strategies in repeated games: Evidence from experimental markets

An experiment is designed to test the implications of the equilibria typically studied in the repeated game literature (i.e. those based on Nash reversion and optimal symmetric two-phase punishment strategies). One hundred pairs of subjects repeatedly set prices in a differentiated demand duopoly setting. Unlike existing experiments there was no fixed end date or time constraint (some pairs played the game for over six months). Subjects were occasionally asked to enter prices through one-period ahead pricing strategies. The equilibrium conditions predicted by theory are not satisfied in any of the one hundred markets. Less than 5% of elicited strategies have properties similar to those typically studied in the literature and almost 95% of strategies are less harsh (in their immediate response to undercutting) than implied by Nash reversion. Future earnings are highest for subjects adopting matching strategies, even after controlling for a subject’s past earnings.