Monetary loss surveillance for credit models.

Abstract

There is a vast collection of statistical methodologies devoted to measure the customer’s credit risk.Well-knownstatistical techniques are logistic regression, genetic algorithms, and support vector machines, among others. However, there is a lack of statistical tools for monitoring monetary losses implied by a given credit model in operation. This article introduces a sequential procedure to favor such monitoring. Our method favors early detection of increased expected monetary losses. Analytical expressions are derived for the calculation of the statistical power performance of the proposed method. An application for a credit portfolio of a German bank is offered.