Credit risk measurement

Components of credit risk

Credit risk exists in every credit engagement. In measuring credit risk at a counterparty level we reflect three components – the “probability of default_ by the client or counterparty on its contractual obligations; our current exposure to the counterparty and its likely future development, from which we derive the “exposure at default_; and the likely recovery ratio on the defaulted obligations to give us the “loss given default_. These components are also important parameters in determining portfolio risk, not only for our internal credit risk measures but also for future regulatory capital calculations, since they are the basis of the Basel II Advanced Internal Rating Based approach, which we intend to adopt when it comes into force in 2008.

We assess the likelihood of default of individual counterparties using rating tools tailored to the various categories of counterparty. They have been developed internally and combine statistical analysis with credit officer judgment and are validated, where appropriate, by comparison with externally available data. Clients are segmented into 15 rating classes, two being reserved for cases of impairment or default. The UBS rating scale, which is shown above, reflects not only an ordinal ranking of our counterparties, but also the range of default probabilities defined for each rating class. This means that, in principle, clients migrate between rating classes as our assessment of their probability of default changes. The rating tools are kept under review and upgraded as necessary. We regularly validate the performance of our rating tools and their predictive power with regard to default events. Where statistical analysis suggests that the parameters of a model require adjustment, we reflect such changes in our external reporting once the calibration is confirmed and implemented across the portfolio concerned. In the interim we estimate the impact which the future model amendments might have on our internal credit risk measures and adjust them accordingly.

The ratings of the major rating agencies shown in the table above are mapped to our rating classes based on the long-term average default rates for each external grade. We use the external ratings where available to benchmark our internal credit risk assessment. Observed defaults per rating category vary year-on-year, especially over an economic cycle, and therefore this mapping does not imply that UBS expects this number of defaults in any given period. As we validate our own internal rating tools for their ability to predict defaults, we also monitor long-term average default rates associated with external rating classes. If our analysis suggests that the probability of default associated with external rating grades has substantially changed, we adjust their mapping to our internal rating scale. We reflect such changes in our external reporting once the calibration is confirmed.

Exposure at default is based on the amounts we expect to be owed at the time of default. For a loan this is the face value. For a commitment, we include any amount already drawn plus the further amount which may have been drawn by the time of default, should it occur. For repos and securities borrowing and lending transactions, we assess the net amount which could be owed to or by us following adverse market moves over the time it would take us to close out all transactions (“close out exposure_). Exposure on OTC derivative transactions is determined by modelling the potential evolution of the value of our portfolio of trades with each counterparty over its life (potential credit exposure), taking into account legally enforceable close-out netting agreements where applicable. From this model we can derive both an “expected future exposure_ profile and a “maximum likely exposure_ profile measured to a specified confidence level. The ability to call collateral and any collateral actually held are also taken into account.

Loss given default (LGD) or loss severity represents our expectation of the extent of loss on a claim should default occur. It is expressed as percentage loss per unit of exposure and typically varies by type of counterparty, type and seniority of claim, and availability of collateral or other credit mitigation.

In line with our own internal governance standards and the requirements of the new regulatory capital framework, we subject all models developed for credit risk measurement, including the components of such measures, to independent review by a specialist team in Corporate Center prior to implementation, and to ongoing validation once they are deployed.