Calibration (banking)
In banking, calibration is the assignment of a default probability of a credit risk to a rating level . The calibration only affects the two internal rating procedures, namely the simple and the advanced IRB approach . The standardized approach to credit risk (CRSA) is not affected by this.
General
If credit institutions choose one of the two internal rating procedures , scaling must be created in a rating , which - similar to a school grading system - provide for a gradation from “no counterparty risk ” to “high counterparty risk”. These scales are called rating levels and are associated with a probability of default (PD). This probability of default must be measured empirically by each credit institution using a default history. The actual default rate as a proportion of the total credit volume, in particular the credit defaults, is now transferred to future development using trend extrapolation . The lowest probability of default is assigned to the rating level with no counterparty default risk. This also applies to all further levels.
Procedure
In the opinion of the Bundesbank, a rating system is considered to be well calibrated if the estimated probabilities of default do not deviate at all or only slightly from the actual default rates. Calibration involves both the “correct” amount of equity to be kept by credit institutions and the relative weighting of the individual risks, in the credit risk area also the slope of the curve of the risk weights . The determination of a representative “average portfolio” plays a central role in calibrating the risk weights in the IRB approach. This average portfolio should, on the one hand, reflect the weighting of the various risk asset classes in the IRB approach and, on the other hand, within an exposure class, the distribution of risk assets across the various rating classes.
Calibration in the broader sense
The calibration of a rating system in the broader sense also includes the assignment of additional risk parameters, in particular the loss rate (abbreviation LGD from English loss given default ) and the default loan amount (abbreviation EaD from English exposure at default ) as part of the calibration of risk weights. These represent - as well as the probability of default - hypothetical variables, since it will at the time of creditworthiness classification of the borrower can only be estimated. In particular, they are dependent on the intrinsic value of any loan collateral or the loan drawn up until default. In contrast to the probability of default, however, these parameters only have to be estimated by a credit institution itself in the advanced IRB approach, while in the simple IRB approach they are specified by the supervisory authority.
The capital adequacy requirement for credit risk should be calibrated to an average of 6.4% (8% minus 1.6% for operational risk ).
Calibration and market prices
Credit default swaps insured against losses from an occurrence loan default, as the guarantor (English protection seller ) in the case of a credit event to the secured party (English protection buyer ) the secured loan amount to pay. For this purpose, the protection buyer must pay the protection seller a premium (credit surcharge, collective term spread ). This credit premium is a market price that depends on the expected probability of default of the respective credit risk (e.g. a country). A credit spread of 200 basis points indicates to market participants that the market is pricing in an annual probability of default of around 3%. Assumptions have to be made about the recovery rate - this is the amount that a creditor can expect as a percentage repayment of his original loan amount in the event of a debt rescheduling or debt relief . In the example, the revenue rate can be 40%, which is a common market convention in many CDS contracts.
Individual evidence
- ↑ Monthly Report of the Deutsche Bundesbank April 2001, p. 29 f. (PDF file; 319 kB)
- ↑ Monthly Report of the Deutsche Bundesbank September 2003, p. 64 f.
- ↑ Monthly Report of the Deutsche Bundesbank June 2006, p. 35 ff.