We study a canonical collective action game with incomplete information. Individuals attempt to coordinate to achieve a shared goal, while also facing a temptation to free-ride. More similar information can help them coordinate, but it can also exacerbate free-riding. Our main result shows that more similar information facilitates (impedes) achieving the common goal when it is sufficiently challenging (easy). We apply this insight to show why less powerful authoritarian governments may face larger protests if they restrict press freedom, when committee diversity is beneficial in costly voting, and when a more diverse community contributes more to public good provision. (JEL C71, D71, D72, D81, D82, D83, H41)

In a regime change game, privately informed agents sequentially decide whether to attack without observing others’ previous actions. To dissuade them from attacking, a principal adopts a dynamic information disclosure policy, frequent viability tests. A viability test publicly discloses whether the regime has survived the previous attacks. When such tests are sufficiently frequent, in the unique cutoff equilibrium, agents never attack if the regime passes the latest test, regardless of their private signals. We apply this theory to demonstrate that a borrower can eliminate panic-based runs by sufficiently diffusing the rollover choices across different maturity dates. (JEL C72, D82, G21)

We consider bargaining environments in which public opinion provides leverage by making compromises costly. Two parties make initial demands, before knowing the public opinion. If deadlocked, they can bargain again after public opinion forms, but suffer reputation costs if they compromise, i.e., scale back their demands. We show that in equilibrium, parties may choose to make incompatible demands initially and gamble over public opinion even though one or both parties must bear a cost later. We characterize when deadlocks arise, and how this affects the welfare of the public in a representative two-party democracy compared to a direct democracy. (JEL C78, D72)

We study optimal adversarial information design in a dynamic regime change game. Agents decide when to attack, if at all. We assume (1) delay incurs a continuous cost and (2) agents doubt the correctness of their actions. The game may end in a “disaster” due to weak fundamentals or panic—agents attacking despite sound fundamentals. We propose a “timely disaster alert” that promptly warns about impending disasters, making waiting for and following the alert the unique rationalizable strategy, thereby eliminating panic. We relate this optimal policy to early-warning systems such as bank stress tests and debt sustainability analysis.