Evolutionary implementation of efficient networks

This paper considers an evolutionary implementation problem of efficient outcomes (Sandholm 2007) in the context of network formation. We assume that players interact with each other in the long run, facing stochastic mistakes. Under no constraints, resulting networks can be inefficient in general. Our main result shows that we can construct a Pigouvian-type taxation mechanism such that the resulting networks are efficient with probability one in the long run.