Q 16: The agencies seek comment on and supporting empirical analysis of (i) the proposed rule’s definitions of LGD and ELGD; (ii) the proposed rule’s overall approach to LGD estimation; (iii) the appropriateness of requiring a bank to produce credible and reliable internal estimates of LGD for all its wholesale and retail exposures as a precondition for using the advanced approaches; (iv) the appropriateness of requiring all banks to use a supervisory mapping function, rather than internal estimates, for estimating LGDs, due to limited data availability and lack of industry experience with incorporating economic downturn conditions in LGD estimates; (v) the appropriateness of the proposed supervisory mapping function for translating ELGD into LGD for all portfolios of exposures and possible alternative supervisory mapping functions; (vi) exposures for which no mapping function would be appropriate; and (vii) exposures for which a more lenient (that is, producing a lower LGD for a given ELGD) or more strict (that is, producing a higher LGD for a given ELGD) mapping function may be appropriate (for example, residential mortgage exposures and HVCRE exposures).
Q 17: The agencies seek comment on the extent to which ELGD or LGD estimates under the proposed rule would be pro-cyclical, particularly for longer-term secured exposures. The agencies also seek comment on alternative approaches to measuring ELGDs or LGDs that would address concerns regarding potential pro-cyclicality without imposing undue burden on banks.
Other than for Real Estate exposures, there is no empirical evidence to support the view that loss given defaults are significantly different/higher during downturns. In general, LGD is influenced by collection efficiencies, as well as geographic, product, and customer characteristics; hence, it is preferable for banks to develop internal estimates of LGD for their wholesale and retail exposures. The historical ELGD for most senior, secured exposures is likely to be in the 20% to 25% range.
The proposed mapping rule will shift ‘mean ELGD’ for these exposures in a downturn environment by 25% to 30% (higher). The proposed mapping function is inappropriate for many exposure types. It is not clear that a robust mapping function can be developed that would be applicable to all exposure types and institutions. We think that the proposed mapping rule is an attempt to introduce a level of precision that may not really exist for many exposure types. We recommend that the agencies consider requiring “downturn LGD’s” for capital computations only for selected exposures such as residential mortgage and HVCRE.
More generally, we would support a consistent approach between the US and EU markets, thereby avoiding fragmentation of approach and confusion within Pillar 3.
An over emphasis on the calibration of model parameters to historical data, especially with short time series data, will create pro-cyclical results. Firms should have flexibility to use logical, forward-looking factors in the models to reflect long-run effects/behaviors, as a means of minimizing this.