Conceptual Framework for Economic Capital Models and Required Inputs

Michel Araten

Credit-risky assets have the potential to result in losses for lenders. Risks of loss need to be assessed so they can be traded off against promised revenues. Credit risk assessments designed to estimate the risks of loss for individual credit transactions are expressed along three dimensions of risk: the likelihood of default, the loss in the event of default, and the exposure at the time of default.

To determine default likelihood, financial-statement and market-based information is evaluated using various forms of models. Models do not have to be quantitative. They can range all the way from expert judgement and rules-based methods to econometric approaches and neural networks. They are oriented towards either directly producing an estimate of default probability or indirectly estimating risk by assigning a rating grade. Rating classifications are designed to distinguish the relative risk of different obligors in an ordinal manner. However, credit capital models require an absolute probability of default (PD) to be assessed for each rating.

To assess loss in the event of default, commonly referred to as loss-given default (LGD), rules-based or more formal models have been