Agreeing to Implement

A principal faces a set of agents who will play a game. What can the principal ensure the agents play? We develop an implementation concept that bridges between best-case Nash and worst-case rationalizability. Before the game is played, the principal suggests a product set of strategies. Agreeable agents follow the suggestion if it is self-enforcing. Other agents understand this fact, and all agents are rational with common belief of rationality. The principal evaluates each implementable suggestion according to its worst case. When all agents are agreeable, the principal implements her best Nash equilibrium. When all are disagreeable, our concept reduces to worst-case rationalizability. We characterize the optimal suggestion in supermodular environments and apply the implementation concept to network games, team production, mechanism design, and settings in which agreeability is uncertain.

(with Marina Halac and Elliot Lipnowski)

Doron Ravid's homepage: https://sites.google.com/site/ravidd