Probability of Default (PD) models are a mathematical means of assessing how likely an obligor is to default on a credit obligation. The scorecard models are composed of quantitative and/or qualitative factors, which are considered indicators of how likely an obligor is to default. The idea is that when all factors are taken together in the model, they can be used to predict how likely an obligor is to default, based on historical default experience on a representative sample. The resulting PD scores can be used in capital and impairment calculations, so it is important (and it is a regulatory requirement) that the models are monitored regularly to ensure that the PD scores output are accurate. This talk will cover 5 different aspects of model monitoring, why these aspects are monitored, and the statistical tests that are used to do so.
Supported by
