Credit risk: Probability of Default and Loss Given Default estimation May 2020 12
an observed impact for a considered downturn period, as outlined above, the downturn
LGD should be calibrated using a haircut approach or an extrapolation approach, or a
combination of both approaches; and
downturn LGD estimation where observed or estimated impact is not available (Section 7
of the GL on downturn LGD): where sufficient data are not available to quantify downturn
LGDs for the downturn period under consideration based on observed or estimated
impact using the two approaches outlined above, firms still have to estimate downturn
LGD. Firms are permitted to estimate downturn LGD using any other approach, but the
downturn LGD estimates plus an appropriate margin of conservatism (covering the lack of
data and methodological deficiencies) must be higher than the corresponding long-run-
average LGDs plus 15 percentage points (capped at a final downturn LGD estimate level of
105%). To use this approach, the institution must justify to the satisfaction of the
competent authority that it can apply neither of the two approaches outlined above.
2.36 In CP21/19, the PRA proposed that firms should be able to adopt a modelling approach in
line with Section 5 or 6 of the GL on downturn LGD, and that it is unlikely that a firm would be
able to justify using an approach in line with Section 7 of the GL on downturn LGD.
2.37 Some respondents argued that some portfolios could have insufficient data to use a
method in line with Section 5 or 6 of the GL on downturn LGD. These approaches could be
disproportionately complex for small data sets and may not produce a credible or robust
output that is fit for purpose, or usable for sound risk management.
2.38 In response, the PRA clarifies that even if an approach under Section 7 of the GL on
downturn LGD were applied, firms would still be required to produce a downturn LGD model
that is fully compliant with the CRR and the relevant RTSs, GLs and PRA SSs. Under Section 7,
while the downturn LGD estimate cannot be lower than the long-run average LGD + 15
percentage points, this add-on cannot be used as a substitute for CRR compliant modelling.
Also, paragraph 36 of the GL on downturn LGD requires a justification to the satisfaction of the
supervisor that the firm cannot calibrate downturn LGD by using an approach under Section 5
or 6. In addition, approaches under Section 7 need to meet CRR requirements to have robust
and representative data.
2.39 In addition, the PRA still considers it unlikely that a firm would be able to justify using an
approach in line with Section 7 of the GL on downturn LGD. Therefore, the PRA has decided to
maintain the proposed policy. Approaches under Section 7 were designed to be used in
exceptional cases only. Section 7 was not designed to be used extensively or for material
portfolios. The PRA also considers that the add-on of 15pp may be insufficient for certain
portfolios.
2.40 Therefore, the PRA expects firms to apply an approach under Section 5 or 6. For low
default wholesale portfolios, this may involve applying the PRA’s wholesale LGD framework for
low default portfolios as set out in SS11/13. If a firm does not have sufficiently robust data for
a large proportion of its portfolio, it is questionable whether it can build a compliant IRB model
for that portfolio. In such cases, it may be more appropriate for the firm to use a non-modelled
LGD approach, such as the FIRB approach or the standardised approach to credit risk.
2.41 The PRA has added paragraph 13.7B(b) in SS11/13 to reflect the above expectation.