An experiment was conducted to test whether discounted repeated play l
eads to greater cooperation and coordination than one-shot play in a p
ublic good environment with incomplete information. The experiment was
designed so that, theoretically, repeated play can sustain equilibria
with substantially higher group earnings than result in the one-shot
Bayesian Nash equilibrium. The design varied a number of environmental
parameters, including the size of the group, and the statistical dist
ribution of marginal rates of substitution between the public and priv
ate good. Marginal rates of substitution were private information but
the statistical distribution was common knowledge. The results indicat
e that repetition leads to greater cooperation, and that the magnitude
of these gains depends systematically both on the ability of players
to monitor each other's strategy and on the environmental parameters.