13

Loss Given Default in IRB approaches

13.A1

When applying the CRR requirements relating to the estimation of loss given default, the PRA expects firms to comply with the EBA’s Guidelines on PD estimation, LGD estimation and the treatment of defaulted assets (EBA/GL/2017/16) and the EBA’s Guidelines for the estimation of LGD appropriate for an economic downturn (‘Downturn LGD estimation’) (EBA/GL/2019/03). The PRA only expects firms to comply with paragraph 135 of the EBA’s Guidelines on PD estimation, LGD estimation and the treatment of defaulted assets to the extent that this would be consistent with paragraph 13.5A of this SS.

Negative LGDs

13.1

The PRA expects firms to ensure that no LGD estimate is less than zero.

Low LGDs

13.2

The PRA does not expect firms to be using zero LGD estimates in cases other than where they had cash collateral supporting the exposures.

13.3

The PRA expects firms to justify any low LGD estimates using analysis on volatility of sources of recovery, notably on collateral, and cures (as outlined below). This includes:

  1. (a) recognising that the impact of collateral volatility on low LGDs is asymmetric as surpluses over amounts owed need to be returned to borrowers and that this effect may be more pronounced when estimating downturn rather than normal period LGDs; and
  2. (b) recognising the costs and discount rate associated with realisations and the requirements of CRR Article 181(1)(e).

13.4

In order to ensure that the impact of collateral volatility is taken into account, the PRA expects firms’ LGD framework to include non-zero LGD floors which are not solely related to administration costs. For residential mortgages, the PRA expects firms to ensure that the LGD estimate for each exposure is no less than 5%.

(CRR Article 179(1)(f))

Treatment of cures

13.5

Where firms wish to include cures in their LGD estimates, the PRA expects them to do so on a cautious basis with reference to both their current experience and how this is expected to change in downturn conditions. In particular, this involves being able to articulate clearly both the precise course of events that will allow such cures to take place and any consequences of such actions for other elements of their risk quantification. For example:

  1. (a) where cures are driven by the firm’s own policies, we would expect firms to consider whether this is likely to result in longer realisation periods and larger forced sale discounts for those exposures that do not cure, and higher default rates on the book as a whole, relative to those that might be expected to result from a less accommodating attitude. To the extent feasible, the PRA expects cure assumptions in a downturn to be supported by relevant historical data.
  2. (b) the PRA expects firms to be aware of and properly account for the link between cures and subsequent defaults. In particular, an earlier cure definition is, other things being equal, likely to result in a higher level of subsequent defaults

(CRR Article 5(2))

13.5A

For the purpose of applying paragraph 135 of the EBA’s Guidelines on PD estimation, LGD estimation and defaulted assets (EBA/GL/2017/16), the PRA expects:

  1. (a) the artificial cash flow should reflect:
    1. (i) principal: total outstanding amount of the full loan at the moment of cure, but only the amount of missed payments (i.e. actual past due payments) accrued up to the moment of cure should be discounted;
    2. (ii) interest: amount accrued between the moment of default and the moment of cure;
    3. (iii) fees: amount accrued between the moment of default and the moment of cure;
    4. (iv) additional observed recoveries: total amount received up to the moment of cure;
    5. (v) additional drawings: firms should follow the requirements of CRR Articles 182(1)(c), 181(2)(b) and 182(3), and paragraphs 139 – 142 of the GL on PD & LGD. Additional drawings included in the artificial cash flow should be treated in the same way as the principal; and
    6. (vi) costs: amount accrued between the moment of default and the moment of cure.
  2. (b) in applying 13.5A(a), the “moment of cure” is defined as the moment when no triggers of default continue to apply at the start of the final probation period[7].
  3. (c) the artificial cash flow should be discounted over the actual period of default only (i.e. between the moment of default and the moment of cure) and, therefore, should not be discounted over any additional time period after the moment of cure, such as the final probation period.

Footnotes

  • 7. Paragraph 71(a) of the EBA Guidelines on the definition of default require that a cured exposure show no triggers of default for a minimum of three months in order to be considered a cure (‘probation period’).

Incomplete workouts

13.6

In order to ensure that estimates of LGDs take into account the most up to date experience, we would expect firms to take account of data in respect of relevant incomplete workouts (ie defaulted exposures for which the recovery process is still in progress, with the result that the final realised losses in respect of those exposures are not yet certain).

(CRR Article 179(1)(c))

LGD – sovereign floor

13.7

To ensure that sovereign LGD models are sufficiently conservative in view of the estimation error that may arise from the lack of data on losses to sovereigns, the PRA expects firms to apply a 45% LGD floor to each unsecured exposure in the sovereign asset class.

(CRR Articles 144(1) and 179(1)(a))

Downturn LGD

13.7A

As required by the UK Technical Standards on the specification of the nature, severity and duration of an economic downturn[8], firms that identify an economic downturn must examine economic indicators over a historical time-span that provides values that are representative of the likely range of variability in the future, and that this period must have a duration of at least 20 years. The PRA expects that firms should select a historical time-span which enables the identification of economic indicator values that represent sufficiently severe downturn conditions. If the values do not represent sufficiently severe downturn conditions, firms should extend their historical time-span beyond the minimum 20 year period.

Footnotes

  • 8. As implemented by the Technical Standards (Economic Downturn) Instrument 2021.

13.7B

Section 4.3 of the EBA’s Guidelines for the estimation of LGD appropriate for an economic downturn (EBA/GL/2019/03) sets out a hierarchy of three types of approaches for calibrating downturn LGD for each considered downturn period. The PRA has the following expectations for the application of these approaches:

  1. (i) the PRA considers that both a component-based modelling approach and a direct estimate modelling approach are permitted for use under Section 5 ‘downturn LGD estimation based on observed impact’, and that a component-based approach is permitted for use in Section 6 ‘downturn LGD estimation based on estimated impact’. However, the PRA considers that it is unlikely that firms will be able to produce robust direct estimate downturn LGD models for residential mortgages. Therefore, the PRA expects firms to use a component-based approach for these exposures; and
  2. (ii) the PRA expects that it is unlikely that firms will be able to demonstrate that they cannot use one of the approaches in (a). The PRA therefore does not expect firms to use the approach to estimating downturn LGD set out in Section 7: ‘downturn LGD estimation where observed or estimated impact is not available’.

13.7C

The PRA expects firms using a component based approach to modelling downturn LGD to ensure that all components reflect a downturn and that each component reflects the same downturn. Firms should take into account any time lags between the downturn period and the potential impact on the firm’s loss data. Therefore, while model components should reflect the same downturn, a time lag may be necessary so that the peak value within the same downturn is used for each model component. Firms should ensure the time lags are not so long that they result in LGD estimates that are reflective of an upturn or improved economic conditions.

13.7D

Firms should adjust their downturn LGD estimates, where necessary, in order to ensure that ‘downturn LGD estimates are not unduly sensitive to changes in [the] economic cycle’.[9] Firms should consider whether any adjustments they already make to address cyclicality in wholesale LGD estimates, for example through conservative approaches to haircutting collateral values, satisfy this requirement.

(CRR Article 181(1)(b))

Footnotes

  • 9. Paragraph 17 of the EBA’s Guidelines for the estimation of LGD appropriate for an economic downturn (‘Downturn LGD estimation’) (EBA/GL/2019/03).

LGD — UK retail mortgage property sales reference point

13.8

The PRA believes that an average reduction in property sales prices of 40% from their peak price, prior to the market downturn, forms an appropriate reference point when assessing downturn LGD for UK mortgage portfolios and expects a firm’s rating systems to assume a reduction consistent with this. This reduction captures both a fall in the value of the property due to market value decline as well as a distressed forced sale discount. The PRA expects the assumption for the fall in the value of the property due to house price deflation not to be lower than 25%.

13.9

Where firms adjust assumed house price values within their LGD models to take account of current market conditions (for example with reference to appropriate house price indices) we recognise that realised falls in market values may be captured automatically. Firms adopting such approaches may remove observed house price falls from their downturn house price adjustment so as not to double count. The PRA expects all firms wishing to apply such an approach to seek the consent of the PRA and to be able to demonstrate that the following criteria are met:

  1. (a) the adjustment applied to the market value decline element of a firm’s LGD model is explicitly derived from the decrease in indexed property prices (ie the process is formulaic, not judgemental);
  2. (b) the output from the adjusted model has been assessed against the 40% peak-to-trough property sales prices decrease reference point (after inclusion of a forced sale discount);
  3. (c) a minimum 5% market value decline applies at all times in the LGD model; and
  4. (d) the firm has set a level for reassessment of the property market price decline from its peak. For example, if a firm had initially assumed a peak-to-trough market decline of 25%, then it will have set a level of market value decline where this assumption will be reassessed.

(CRR Article 181(1)(b))

Downturn LGDs

[The section was deleted in its entirety in May 2020]

Discounting cash flows

13.11

In order to ensure that their LGD estimates incorporate material discount effects, the PRA expects firms’ methods for discounting cash flows to take account of the uncertainties associated with the receipt of recoveries with respect to a defaulted exposure, for example by adjusting cash flows to certainty equivalents or by using a discount rate that embodies an appropriate risk premium; or by a combination of the two.

13.12

If a firm intends to use a discount rate that does not take full account of the uncertainty in recoveries, we would expect it to be able to explain how it has otherwise taken into account that uncertainty for the purposes of calculating LGDs. This can be addressed by adjusting cash flows to certainty equivalents or by using a discount rate that embodies an appropriate risk premium for defaulted assets; or by a combination of the two.

13.13

In addition to the above expectations, firms should ensure that no discount rate used to estimate downturn LGD is less than 9%.

13.13A

For the purpose of estimating long-run average LGD, the PRA expects firms to use a discount rate of Sterling Overnight Index Average (SONIA) at the moment of default plus 5% for exposures denominated in GBP. For the purpose of estimating long-run average LGD for exposures denominated in currencies other than GBP, firms should use a comparable liquid interest rate in the currency of that exposure.

13.13B

For defaulted exposures, the PRA expects firms to use a discount rate of SONIA at the moment of default plus 5% for estimating Best Estimate of Expected Loss[10]; and ensure that no discount rate used to estimate LGD in-default is less than 9%.

Footnotes

  • 10. Best Estimate of Expected Loss is also known as ‘BEEL’ and ‘ELBE’.

13.13C

For the discount rate to be used for defaults that occurred before 2 January 1997 (ie the first SONIA rate available from the Bank of England), the PRA expects firms to develop an approach, for example, an extrapolation based on available data, or use an appropriate alternative for that period, for example, the relevant central bank rate.

(CRR Article 5(2))

13.13D

The PRA expects that the amount of recoveries that can be recognised as a cash flow and discounted should not be higher than the amount of recoveries the firm is contractually entitled to retain for the exposure.

Wholesale LGD

13.14

The PRA expects firms using AIRB approaches to have done the following in respect of wholesale LGD estimates:

  1. (a) applied LGD estimates at transaction level;
  2. (b) ensured that all LGD estimates (both downturn and non-downturn) are cautious, conservative and justifiable, given the paucity of observations. In accordance with CRR Article 179(1)(a), estimates must be derived using both historical experience and empirical evidence, and not be based purely on judgemental consideration. We expect the justification as to why the firm thinks the estimates are conservative to be documented;
  3. (c) identified and explained at a granular level how each estimate has been derived. This should include an explanation of how internal data, external data, expert judgement or a combination of these has been used to produce the estimate;
  4. (d) clearly documented the process for determining and reviewing estimates, and the parties involved in the process in cases where expert judgement was used;
  5. (e) demonstrated an understanding of the impact of the economic cycle on collateral values and be able to use that understanding in deriving their downturn LGD estimates;
  6. (f) demonstrated sufficient understanding of any external benchmarks used and identified the extent of their relevance and suitability to the extent that the firm can satisfy itself that they are fit for purpose;
  7. (g) evidenced that they are aware of any weaknesses in their estimation process and have set standards, for example related to accuracy, that their estimates are designed to meet;
  8. (h) demonstrated that they have sought and utilised relevant and appropriate external data, including through identifying all relevant drivers of LGD and how these will be affected by a downturn;
  9. (i) ensured, in most cases, estimates incorporate effective discrimination on the basis of at least security type and geography. In cases where these drivers are not incorporated into LGD estimates then we would expect the firm to be able to demonstrate why they are not relevant;
  10. (j) have put in place an on-going data collection framework to collect all relevant internal loss and exposure data required for estimating LGD and a framework to start using these data as soon as any meaningful information becomes available; and
  11. (k) ensured it can articulate the data the firm intends to use from any industry-wide data collection exercises in which it is participating, and how the data will be used.

(CRR Section 6)

LGD models for low default portfolios

13.15

We have developed a framework for assessing the conservatism of firms’ wholesale LGD models for which there are a low number of defaults. The framework is set out in Appendix C and does not apply to sovereign LGD estimates which are floored at 45%. We are in the process of using this framework to assess the calibration of firms’ material LGD models for low-default portfolios.

13.16

In the following cases, the PRA expects firms to determine the effect of applying the framework set out in Appendix C to models which include LGD values that are based on fewer than 20 ‘relevant’ data points (as defined in Appendix C):

  1. (a) the model is identified for review by the PRA; or
  2. (b) the firm submits a request for approval for a material change to its LGD model.

13.17

In such cases firms should contact their supervisor to obtain the relevant data templates that should be populated and submitted to the PRA.

LGD – use of external data for residential mortgages

13.17A

The PRA expects that, for residential mortgages, where a firm’s internal experience of defaults for a rating system is low, the firm may use external data to supplement internal data when modelling LGD.

13.17B

Where external data are used, the PRA expects firms to apply additional margins of conservatism in order to:

  1. (a) recognise the difference between downturn recoveries from established firms with the experience and processes to realise high recoveries, and those from firms with more limited experience and less established processes;
  2. (b) recognise any differences in portfolio comparability between the external data and the firm’s lending; and
  3. (c) address unobservable differences that relate to risk drivers or risk characteristics that cannot be derived from external data.

13.17C

The PRA expects the level of added conservatism to be significant until sufficient internal data are available to support the firm’s reduction.

13.17D

Firms using external data in their LGD estimates should run a Forced Sale Discount (FSD) model and Probability of Possession Given Default (PPGD) model with appropriate governance and monitoring requirements. Firms with no internal repossession data for use in their FSD modelling could rely on external data, along with an internal expectation on costs and an additional margin of conservatism, as part of their FSD estimation.

13.17E

The PRA considers that firms would be unlikely to be able to demonstrate that third-party recovery data from non-UK legal regimes are comparable to UK data. The PRA therefore expects only UK data to be used when estimating LGD for UK residential mortgage exposures. For non-UK mortgage exposures, the PRA expects firms to demonstrate that data are representative for the local mortgage market in order to be used to supplement internal data where appropriate.

13.17F

The PRA expects firms to incorporate internal data as it builds up.

(CRR Articles 171, 179 and 181)

Unexpected loss (UL) on defaulted assets[11]

[This section was deleted in its entirety in May 2020]

Unsecured LGDs where the borrowers’ assets are substantially used as collateral

13.21

The extent to which a borrower’s assets are already given as collateral will affect the recoveries available to unsecured creditors. If the degree to which assets are pledged is substantial this will be a material driver of LGDs on such exposures. Although potentially present in all transactions, the PRA expects firms to be particularly aware of this driver in situations in which borrowing on a secured basis is the normal form of financing, leaving relatively few assets available for the unsecured debt. Specialist lending (including property), hedge funds, some SME/mid-market lending are examples of such cases.

13.22

The PRA expects firms to take into account the effect of assets being substantially used as collateral for other obligations when estimating LGDs for borrowers for which this is the case. The PRA expects firms not to use unadjusted data sets that ignore this impact, and note that it is an estimate for downturn conditions that is normally required. In the absence of relevant data to estimate this effect, conservative LGDs – potentially of 100% – are expected to be used.

(CRR Articles 171(2), 179(1)(a))

Probability of Possession Given Default (PPGD) for UK residential mortgage exposures

13.22A

The PRA expects firms to ensure that PPGD estimates appropriately reflect economic downturn conditions. The PRA expects (as in paragraphs 13.8 and 13.9 above) downturn PPGD estimates to be consistent with a fall in the value of property due to house price deflation not lower than 25% from the previous peak price, and not lower than 5% from the current price.

13.22B

Firms should reflect these economic downturn conditions in their PPGD models by ensuring that the:

  1. (a) allocation of exposures to rating grades is consistent with the reductions in property values set out above.
  2. (b) calibration of possession rates for a given rating grade is based on data reflecting the reductions in property values set out above. If the data reflect reductions in property values that are lower than either reduction above, firms should appropriately adjust their calibration within grades to be consistent with these property values.

13.22C

If a firm’s PPGD model is not sensitive, or less sensitive, to falls in property values, for example if the model uses values at origination and not current values to assign exposures to rating grades, the firm should ensure that its calibration of possession rates reflects economic conditions where property values are at least 25% below their peak values. The firm should also demonstrate to the PRA that the model achieves similar outcomes as it would if it was using current property values to assign exposures to rating grades, including in stressed scenarios.

13.22D

Firms with limited data from a downturn should apply an additional margin of conservatism.

13.23

For firms with low internal experience of possessions, the PRA expects firms to assess the appropriate margin of conservatism in the calculation of PPGD against PRA reference points.

13.24

The PRA believes the following reference points to be appropriate:

  1. (a) PPGD reference point of 100% where there are very low default volumes, regardless of the length of observed outcomes; and
  2. (b) PPGD reference point of 70% where firms are able to demonstrate they have greater, but still not considerable, volume and history of data to estimate future possession rates.

13.25

The PRA expects firms to assess whether, on a case-by-case basis, they can apply a PPGD level above or below the reference point relevant to their circumstances. Indicators supporting a PPGD level set higher than 70% include: high LTV lending; non-owner occupied lending (ie buy-to-let); and levels of default data towards the lower end of the mortgage lenders cohort. Indicators supporting a PPGD level set lower than 100% or 70% include: low LTV lending; owner-occupied lending; and more data than typical of the cohort. The PRA will consider a firm’s proposal to use a lower level of PPGD than the relevant reference point on a case-by-case basis.

13.26

As required by the CRR, firms using the PRA reference points as a basis for calculating PPGD margins of conservatism will still need to run an LGD model subject to appropriate governance and monitoring requirements. As a firm gains additional data, and the modelled PPGD estimates rely upon internal data to a greater extent, the PRA expects the appropriate margin of conservatism to decline.