Many regulations aim to promote coordination among creditors in bankruptcy by ex post restricting their ability to exit distressed firms. However, such restrictions may harm creditors’ ex ante incentives to stay invested, thereby worsening coordination outcomes. We build a dynamic coordination model to show how this force shapes creditor runs, bankruptcy filings, and regulation designs. Intriguingly, filing for bankruptcy early, thereby preserving more assets for latecomers, can prolong firm life. Furthermore, regulators’ clawbacks on prebankruptcy repayments can be superior to firms’ commitment to early bankruptcy filing. Our analysis generates implications for automatic stay, avoidable preference, bank failures, and seniority structure.

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.

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.