Group Dynamics in Experimental Studies - The Bertrand Paradox Revisited

Different information provision in experimental markets can drastically change subjects' behavior. Considering the repeated Bertrand duopoly game of Dufwenberg and Gneezy (2000), we find that population feedback about the prices in other markets outside a subjects' own current market causes group dynamics that prevent prices from convergence to Nash equilibrium. Limited information comprising only the decisions of a subject's own opponent, in contrast, leads to competitive behavior. When we extend the number of periods from 10 to 25 in the full information treatment we observe a very robust cyclical up and down movement of prices. We can explain tacit coordination in our experiment with an extended learning direction model and leadership by example.