IZA DP No. 12347: An Experimental Study on the Effects of Communication, Credibility, and Clustering in Network Games
Revised version forthcoming in: Review of Economics and Statistics (accepted August 2021)
The effectiveness of social interaction depends strongly on an ability to coordinate actions efficiently. In large networks, such coordination may be very difficult to achieve and may depend on the communication technology and the network structure. We examine how pre-play communication and clustering within networks affect coordination in a challenging experimental game on eight-person networks. Free-form chat is enormously effective in achieving the nonequilibrium efficient outcome in our game, but restricted communication (where subjects can only indicate their intended action) is almost entirely ineffective. We can rationalize this result with a novel model about the credibility of cheap-talk messages. This credibility is much larger with freeform message communication than with restricted communication. We are the first to model this credibility and show, both theoretically and experimentally, an interaction effect of network structure and communication technologies. We also provide a model of message diffusion, which indeed predicts that diffusion will be more rapid without clustering and is consistent with our data.