Traditional approaches use balance sheet assessment to forecast bank rescue costs and often underestimate real costs. Thus, governments commit to bailouts without credible forecasts of costs, which can lead to serious fiscal issues. We propose a new approach based on distributional information from past banking crises than can correct for problems faced by traditional approaches. We also derive a basic but very informative model of debt sustainability incorporating various macroeconomic factors. Used in conjunction with our forecasting approach, it sets an upper limit to the ability of governments to shoulder banking bailouts. Our results show that rescuing banks in the absence of hard information on the quality of their balance sheets involves high tail risks through potentially high fiscal costs.