Is the "smoke-filled room" necessary? An experimental study of the effect of communication networks on collusion

We investigate how well firms can collude with partial communication in relation to full communication. We find that firms vary their communication strategies with network structures. In the networks that have either an isolated firm or full communication, more players send "pure promises" suggesting everyone select the collusive actions unconditionally. In the network with leadership, more players send a "promise and threat" which includes a reward for collusion and a punishment for deviating. Because of the inability to communicate in the network with the isolated firm and the high frequency of deviation and punishment in the network with leadership, the full communication network achieves significantly higher payoff than partial communication.