Rollover risk and stress test credibility

This paper studies information disclosure when financial supervisors cannot commit to communicate truthfully. A regulator performs a stress test and chooses whether to disclose bank-specific or aggregate results. Results can be biased at a cost (the higher this cost, the more credible the regulator). Manipulating aggregate information may avoid bank failures, but only if credibility is high enough. Supervisors with little credibility cannot prevent systemic runs by misreporting aggregate information and must release bank specific reports (truthful or not), triggering partial runs. The results have implications for institutional design: ex ante, a social planner would choose an interior level of credibility. (C) 2021 Elsevier Inc. All rights reserved.