Emerging Practices for Capital Adequacy © Copyright 2003, CCRO. All rights reserved. 32 Figure 17: Elements in Generating Portfolio Unexpected Loss • Portfolio unexpected loss – This is the potential volatility of actual losses relative to the expected level of loss. • Loss correlations and portfolio concentrations12 - These measure the likelihood of multiple simultaneous defaults. Idiosyncratic default risk results from the random chance that any single counterparty will go into default over the one-year time horizon because of conditions specific to that counterparty. Systematic default risk results from the chance of experiencing many defaults due to prevailing economic conditions. A common practice is to infer correlations among defaults from equity prices. Each obligor is assigned to an industry and geographical sector. Correlations are then inferred from co-movements of the common risk factors. • Loss volatility – This approach is used by some credit risk models to measure the variation in the combination of the exposure and loss given default parameters. • Default volatility – Even when good estimates of the expected default probability (EDP) of a counterparty are used, uncertainty exists about whether or not it will default. Default volatility captures the fact that default either does occur or does not and reflects the fact that actual default rates could differ from expectations. • Ratings migration – Some credit risk models use this to measure mark-to-market change with regard to the change of credit quality from one notch to another within a given time period. • Distribution of portfolio credit losses – By considering the interaction of exposures, loss given defaults (LGDs), default probabilities, and correlations and concentrations across all counterparties, we arrive at the distribution of credit losses at the portfolio level. Accounting for volatility not just in the exposure but also in the default probabilities and LGDs results in a richer distribution of credit losses. 12 In practice, it is very difficult to measure these correlations accurately or precisely. D istribution of Exposures by C ounterparty Loss C orrelation, Loss Volatility, Portfolio C oncentration, and R ating M igration Portfolio U nexpected Loss Econom ic C apital Expected D efault Probability by C ounterparty Loss G iven D efault by C ounterparty Variance to Expected Loss, Target C onfidence Level Portfolio Expected Loss D istribution of Portfolio C redit Losses