Equilibrium selection in generalized Polya-urn coordination games with partial information on population shares

I investigate deterministic best-reply dynamics in generalized Polya-urn coordination games, where a Pareto dominant equilibrium coexists with a risk-dominant one and players have partial information on current population shares. After analytically deriving estimates for the basin of attraction of either equilibrium, I study numerically (and via simulation) equilibrium selection in the long run. Results indicate that risk dominance robustly prevails over Pareto efficiency as a selection principle. Furthermore, I find that the probability of selecting a Pareto-efficient outcome decreases the smaller the sample size drawn by entrants and increases the larger the initial pool of incumbents.