Let the worst one fail: a credible solution to the too-big-to-fail conundrum

We study time-consistent bank resolution mechanisms. The key constraint is that governments cannot avoid bailouts that are ex post efficient. Contrary to common wisdom, we show that the government may still avoid moral hazard and implement the first-best allocation by using the distribution of bailouts across banks to provide incentives. We analyze properties of credible tournament mechanisms that provide support to the best-performing banks and resolve the worst-performing ones. We extend our mechanism and show that it continues to perform well when banks are imperfect substitutes, when they are differentially interconnected as long as bailout funds can be earmarked, and when their risk-taking is driven by overoptimism instead of moral hazard.