This paper proposes a general model that combines the Mixture Hazard Model with the Stochastic Frontier Model for the purposes of investigating the main determinants of the failures and performances of a panel of U.S. commercial banks during the financial crisis that began in 2007. The combined model provides measures of the probability and time to failure conditional on a bank’s performance and vice versa. Both continuous-time and discrete-time specifications of the model are considered in the paper. The estimation is carried out via the expectation-maximization algorithm due to incomplete information regarding the identity of at-risk banks. In- and out-of-sample predictive accuracy of the proposed models is investigated in order to assess their potential to serve as early warning tools.
JEL classification codes: C33, C41, C51, D24, G01, G21.