Nota di Lavoro, Fondazione Eni Enrico Mattei 107.2005
We present experiments on repeated non-cooperative network formation games, based on Bala and Goyal (2000). We treat the one-way and the two-ways flow models, each for high and low link costs. The models show both multiple equilibria and coordination problems. We conduct experiments under various conditions which control for salient labeling and learning dynamics. Contrary to previous experiments, we find that coordination on non-empty Strict Nash equilibria is not an easy task for subjects to achieve, even in the mono-directional model where the Strict Nash equilibria is a wheel. We find that salience significantly helps coordination, but only when subjects are pre-instructed to think of the wheel network as a reasonable way to play the networking game. Evidence on learning behavior provides support for subjects choosing strategies consistent with various learning rules, which include as the main ones Reinforcement and Fictitious Play.
Experiments Networks Behavioral game theory Salience Learning dynamics