Crisis Modeler: A Tool for Exploring Crisis Predictions

This paper presents CrisisModeler as a tool for exploring financial crisis predictions. Despite wide interest in crisis prediction, little attention has been given to generalizable modeling solutions, real-time implementations, thorough comparisons among methods and interactive interfaces to explore models and their output. CrisisModeler combines many approaches used in predicting financial crises within a fullyfledged framework for modeling and evaluation, and provides an implementation of a general-purpose tool with a web-based interactive interface to explore model output. We present the underlying framework behind CrisisModeler and illustrate the use of it with a case study on European banks, including a horse race of methods and investigations of different specifications. The case study illustrates the versatility and suitability of the tool for supporting exploration and communication of models for crisis prediction.