During the global financial crisis, a large number of banks worldwide either failed or received financial aid thus inflicting substantial losses on the system. We contribute to the early warning literature by constructing a dynamic competing risks hazard model that explores the joint determination of the probability of a distressed bank to face a licence withdrawal or to be bailed out. The underlying patterns of distress are analysed based on a broad range of bank-level and environmental factors. We find that institutions with inadequate capital, illiquid and risky assets, poor management, low levels of earnings and high sensitivity to market conditions have a higher probability to go bankrupt. Bailed out banks, on the other hand, face both capital and liquidity shortages, experience low earnings, and are highly exposed to market products; however, neither the managerial expertise, nor the quality of assets is relevant to the odds of bailout. We further document that large and complex banks are less likely to fail and more likely to be bailed out and also that authorities are more prone to provide support to a distressed bank, which is well-connected with politicians and political parties and less prone to let it go bankrupt. Importantly, our model outperforms the commonly used logit model in terms of forecasting power in all the in- and out-of-sample tests we conduct.