Prudential Regulation Authority Rulebook

The following framework should be used to assess wholesale LGD models in the circumstances set out in paragraph 13.15 of this supervisory statement:

1. (a) For unsecured recoveries, if a firm has fewer than 20 relevant default observations of recoveries in a specific country for an individual type of exposure, then the maximum recovery a firm can assume must be equivalent to that which would give a 45% LGD for senior unsecured exposures, 75% for subordinated exposures and 11.25% for covered bonds.
2. (b) If a firm is taking account of non-financial collateral which is not eligible under the foundation approach, where they do not have 20 or more relevant data points of recovery values for that type of collateral or do not have a reliable time series of market price data for the collateral in a specific country, then the LGD for the exposure to which the collateral is applied must be floored at 45%.
3. (c) If a firm is taking account of non-financial collateral, which is eligible under the foundation approach, where they do not have 20 or more relevant data points of recovery values for that type of collateral or do not have a reliable time series of market price data for that collateral in a specific country, then the LGD for the exposure to which the collateral is applied must be floored at 35%.

Firms should note the following when applying the framework to LGD models:

1. (a) The 20 or more relevant data points can include internal or external data, however the PRA expects firms to ensure that each data point is independent, representative and an accurate record of the recovery for that exposure or collateral type in that specific country.
2. (b) We would anticipate firms being able to use market price data within the framework where they have fewer than 20 defaults only in exceptional circumstances. As a minimum, firms would need to demonstrate that the market price data being used is representative of their collateral and that it is over a long enough time period to ensure that an appropriate downturn and forced sale haircut can be estimated.
3. (c) The framework does not affect the use of financial collateral.
4. (d) The framework does not affect the use of unfunded credit protection.
5. (e) Where a model takes account of multiple collateral types, if this only includes collateral that is eligible under the foundation approach then LGDs must be floored at 35%, and if any collateral type is not eligible under the foundation approach then LGDs must be floored at 45%.
6. (f) The effect of this framework is to floor bank and non-bank financial institution (NBFI) exposures at foundation values unless sufficient country-specific recovery data is available. This floor should be applied where the exposures are to types of banks and NBFIs that are not sufficiently represented in the available historic data (eg if the historic recovery data only relates to small banks then the floor will affect large banks).
7. (g) When applying the framework the PRA expects firms to assess whether the 11.25% LGD floor for covered bonds is sufficient given the quality of the underlying assets.

Firms should select the most appropriate of the following three options when using the framework to assess wholesale EAD models in the circumstances set out in paragraph 14.8 of this supervisory statement:

1. (a) Rank-order the off balance sheet product types (separately for lending and trade finance) according to their drawdown risk. The EAD parameter for a product with 20 or more default observations can then be applied to low-default products with a lower drawdown risk; or
2. (b) For product types where the firm has the defaults needed to estimate the EAD for committed credit lines (or an estimate derived from the option above) but fewer than 20 defaults for uncommitted credit lines, use 50% of the committed credit line conversion factor as an estimate of the uncommitted credit line conversion factor; or
3. (c) Apply the foundation parameters.

Firms should note the following when applying the framework to EAD models:

1. (a) Firms may select more than one option when applying the framework providing that they can demonstrate that their chosen combination is appropriate, reflecting their particular mix of products and risks, and is not selected in order to minimise their capital requirements.
2. (b) As we believe that the EAD experienced by firms is dependent on their own credit management processes we would expect only internal data to be used to estimate EAD. However, where firms can convincingly demonstrate to the PRA’s satisfaction that the credit processes are consistent across countries then we would accept that data sourced from these countries could be combined to estimate the EAD for each product (ie the 20 default data points do not have to be country specific for the purposes of estimating EAD).
3. (c) Firms using the option in paragraph 22.3(a) above should be able to demonstrate that a sufficiently robust approach has been taken to rank-ordering their product types by drawdown risk. This approach must be fully documented and assessed by an independent reviewer.