This paper is designed to identify predictors of eventual bank failure as well as factors that could enhance the survival ability of agricultural and non-agricultural banks. This study utilizes a split-population survival model that addresses two shortcomings of the basic duration model. First, this model departs from the restrictive assumption of the traditional model that all bank observations in the sample would eventually fail. Second, this model provides for a clear distinction between the determinants of the probability of failure and factors influencing the timing of the failure. The results of this study suggest that failure to allow for a split population among sample banks represents an important misspecification with serious implications in identifying the determinants of the timing of bank failure, more than just the probability of failure.