In this study, I model the predictors and manifestation of bank stress during the financial crisis using a Multiple Indicator Multiple Cause model. Unlike most early warning models that predict failure probabilities, this article describes a framework for predicting a broader notion of bank stress that need not rely on regulatory decisions. As such, this method can be easily applied to large institutions, and avoids the complications associated with modelling a regulatory decision such as failure or a CAMELS downgrade. Using bank reliance on Term Auction Facility funds and the out-of-sample incidence of failures and acquisitions, I demonstrate that the measure of bank stress generated here accords with other notions of bank-level distress. Finally, this method catalogues predictors of distress during the financial crisis. Thus, this article can help assess the validity of several recent regulatory proposals. I find that those banks entering the crisis with more Tier 1 capital, more liquid balance sheets, and relatively stable liabilities subsequently came under less stress. These findings support the Basel III recommended increases in banks' capital adequacy, liquidity and stable funding.