BANK DEFAULT PREDICTION MODELS

摘要: This study examines and compares the predictive performance of multiple default prediction models assesses ability these to correctly predict credit rating transitions. The set that is examined comprise probit model, logit hazards model neural networks model. focus this on US banks time period starts at 1987 ends 2008. out-of-sample results show performances are not very divergent all perform adequate in defaults. Credit transitions, notwithstanding fact various financial ratios prove contain valuable information regarding a bank’s condition, demonstrated be more difficult predict, since deteriorated relative Argumentation for result can found presence subjectivity process specification two events differs some extent. * MA-student Erasmus University Rotterdam, School Economics, Department Finance, student intern Ernst & Young, Financial Services Organization, Risk Management. ERASMUS UNIVERSITY ROTTERDAM Economics Finance SUPERVISOR Prof. Dr. W. F. C. Verschoor ERNST YOUNG Organization Management Drs. J. Menken