Testing for Monotonicity in Credit Risk Modelling

Abstract

Credit risk management, which deals with mitigating losses from lending activities, is crucial for financial institutions. Hence, credit risk modelling can be employed to reduce potential losses and avoid financial crises. There are sometimes monotonic relationships in credit risk models, which can simplify forms of models, reduce computational time, or be necessary to fulfill restrictions observed in reality. After reviewing commonly used credit risk models, several monotonicity testing methods are established and adapted to the situation for binary output of default indicators. Furthermore, we present a new test using the weighted sum of differences as the test statistic, with the weights optimized through combinations of their moments. Finally, we compare the performance of these tests on simulated data regarding the accuracy and power of the tests.