Final Rule, USA
Risk-Based Capital Standards:
Advanced Capital Adequacy Framework
Basel II
Loss
given default (LGD)
Under
the proposed rule, a bank would directly estimate an
ELGD and LGD risk parameter for each wholesale exposure
or would assign each wholesale exposure to an expected
loss severity grade and a downturn loss severity grade,
estimate an ELGD risk parameter for each expected loss
severity grade, and estimate an LGD risk parameter for
each downturn loss severity grade.
In
addition, a bank would estimate an ELGD and
LGD risk parameter
for each segment of retail exposures.
Expected loss given default (ELGD)
The
proposed rule defined the ELGD of a wholesale exposure
as the bank’s empirically based best estimate of the
default-weighted average economic loss per dollar of EAD
the bank expected to incur in the event that the obligor
of the exposure (or a typical obligor in the loss
severity grade assigned by the bank to the exposure)
defaulted within a one-year horizon.
The
proposed rule defined ELGD for a segment of retail
exposures as the bank’s empirically based best estimate
of the default-weighted average economic loss per dollar
of EAD the bank expected to incur on exposures in the
segment that default within a one-year horizon.
ELGD
estimates would incorporate a mix of economic conditions
(including economic downturn conditions).
ELGD
had four functions in the proposed rule—as a component
of the calculation of ECL in the numerator of the
risk-based capital ratios; in the EL component of the
IRB risk-based capital formulas; as a floor on the value
of the LGD risk parameter; and as an input into the
supervisory mapping function.
Many
commenters objected to the proposed rule’s requirement
for banks to estimate ELGD for each wholesale exposure
and retail segment, noting that ELGD estimation is not
required under the New Accord.
Commenters asserted that requiring
ELGD estimation
would create a competitive disadvantage by creating
additional
systems, compliance, calculation, and reporting burden
for those banks subject to the U.S.
rule, many of which have already substantially developed
their systems based on the New Accord.
They
also maintained that it would decrease the comparability
of U.S. banks’ capital requirements and public
disclosures relative to those of foreign banking
organizations applying the advanced approaches.
Several commenters also contended that defining ECL in
terms of ELGD instead of LGD raised tier 1 risk-based
capital requirements for U.S. banks compared to foreign
banks using the New Accord’s LGD based
ECL
definition.
The
agencies have concluded that the regulatory burden and
potential competitive inequities identified by
commenters outweigh the supervisory benefits of the
proposed
ELGD risk
parameter, and are, therefore, not including it in the
final rule.
Instead, consistent with the New Accord, a bank must use LGD for the calculation of ECL and the EL component of
the IRB risk-based capital formulas.
Because the
proposed ELGD risk parameter was equal to or less than
LGD, this change generally will have the effect of
decreasing both the numerator and denominator of the
risk-based capital ratios.
Consistent with the New Accord, under the final rule,
the LGD of a wholesale exposure or retail segment must
not be less than the bank’s empirically based best
estimate of the long-run default-weighted average
economic loss, per dollar of EAD, the bank would expect
to incur if the obligor (or a typical obligor in the
loss severity grade assigned by the bank to the exposure
or segment) were to default within a one-year
horizon over a mix of economic conditions, including
economic downturn conditions.
The
final rule also specifies that LGD may not be less than
zero.
The
implications of eliminating the ELGD risk parameter for
the supervisory mapping function are discussed below.
Economic loss and post-default extensions of credit
Commenters requested additional clarity regarding the
treatment of post-default extensions of credit.
LGD is
an estimate of the economic loss that would be incurred
on an exposure, relative to the exposure’s EAD, if the
obligor were to default within a oneyear horizon during
economic downturn conditions.
The
estimated economic loss amount must capture all material
credit-related losses on the exposure (including accrued
but unpaid interest or fees, losses on the sale of
repossessed collateral, direct workout costs,
and an appropriate allocation of indirect workout
costs).
Where
positive or negative cash flows on a wholesale exposure
to a defaulted obligor or on a defaulted retail
exposure (including proceeds from the sale of
collateral, workout costs, and drawdowns of unused
credit lines) are expected to occur after the date of
default, the estimated economic loss amount must reflect
the net present value of cash flows as of the default
date using a discount rate appropriate to the risk of
the exposure.
The
possibility of postdefault extensions of credit made to
facilitate collection of an exposure would be treated as
negative cash flows and reflected in LGD.
For
example, assume a loan to a retailer goes into default.
The
bank determines that the recovery would be enhanced by
some additional expenditure to ensure an orderly workout
process.
One
option would be for the bank to hire a third-party to
facilitate the collection of the loan.
Another option would be for the bank to extend
additional credit directly to the defaulted obligor to
allow the obligor to make an orderly liquidation of
inventory.
Both
options represent negative cash flows on the original
exposure, which must be discounted at a rate that is
appropriate to the risk of the exposure.
Economic downturn conditions
The
expected loss severities of some exposures may be
substantially higher during economic downturn conditions
than during other periods, while for other types of
exposures they may not.
Accordingly, the proposed rule
required banks to use an LGD estimate that reflected
economic downturn conditions for purposes of calculating
the risk based capital requirements for wholesale
exposures and retail segments.
Several commenters objected to the requirement that LGD
estimates must reflect economic downturn conditions.
Some of these commenters stated that empirical evidence
of correlation between economic downturn and LGD is
inconclusive, except in certain cases.
A few
noted that estimates of expected LGD include
conservative inputs, such as a conservative estimate of
potential loss in the event of default or a conservative
discount rate or collateral assumptions.
One commenter
suggested that if a bank can demonstrate it has been
prudent in its LGD estimation and it has no evidence of
the cyclicality of LGDs, it should not be required to
calculate downturn LGDs.
Other
commenters remarked
that the requirement to incorporate downturn conditions
into LGD
estimates should not be used as a surrogate for proper
modeling of PD/LGD correlations.
Finally, a number of commenters supported a pillar 2
approach for addressing LGD estimation.
Consistent with the New Accord, the final rule maintains
the requirement for a bank to use an LGD estimate that
reflects economic downturn conditions for purposes of
calculating the risk-based capital requirements for
wholesale exposures and retail segments.
More
specifically, banks must produce for each wholesale
exposure (or loss severity rating grade) and retail
segment an estimate of the economic loss per dollar of
EAD that the bank
would expect to incur if default were to occur within a
one-year
horizon during economic downturn conditions.
For
the purpose of defining economic downturn conditions,
the proposed rule identified two wholesale exposure
subcategories – high-volatility commercial real estate (HVCRE)
wholesale exposures and non-HVCRE wholesale exposures
(that is, all wholesale exposures that are not HVCRE
exposures) – and three retail exposure subcategories –
residential mortgage exposures, QREs, and other retail
exposures.
The
proposed rule defined economic downturn conditions with
respect to an exposure as those conditions in which the
aggregate default rates for the exposure’s entire
wholesale or retail subcategory held by the bank (or
subdivision of such subcategory selected by the bank) in
the exposure’s national jurisdiction (or subdivision of
such jurisdiction selected by the bank) were
significantly higher than average.
The
agencies specifically sought comment on whether to
require banks to determine economic downturn conditions
at a more granular level than an entire wholesale or
retail exposure subcategory in a national jurisdiction.
Some
commenters stated that the proposed requirement is at a
sufficiently granular level.
Others
asserted rement should be eliminated or made less
granular.
Those
commenters
favoring less
granularity stated that aggregate default rates for
different product
subcategories in different countries are unlikely to
peak at the same time and that requiring economic
downturn analysis at the product subcategory and
national jurisdiction level does not recognize potential
diversification effects across products and national
jurisdictions and is thus overly conservative.
Commenters also maintained that the proposed granularity
requirement adds complexity and implementation burden
relative to the New Accord.
The
agencies believe that the proposed definition of
economic downturn conditions incorporates an appropriate
level of granularity and are incorporating it unchanged
in the final rule.
The
agencies understand that downturns in particular
geographical subdivisions of national jurisdictions or
in particular industrial sectors may result in
significantly increased loss rates in material
subdivisions of a bank’s exposures.
The
agencies also recognize that diversification across
those subdivisions may mitigate risk for the overall
organization.
However, the agencies believe that the required minimum
level of granularity at the subcategory and national
jurisdiction level provides a suitable balance between
allowing for the benefits of diversification and
appropriate conservatism for risk-based capital
requirements.
Under
the final rule, a bank must consider economic downturn
conditions that appropriately reflect its actual
exposure profile.
For
example, a bank with a geographical or industry sector
concentration in a subcategory of exposures may find
that information relating to a downturn in that
geographical region or industry sector may be more
relevant for the bank than a general downturn affecting
many regions or industries.
The
final rule (like the proposed rule)
allows banks to
subdivide exposure subcategories or national
jurisdictions as they deem appropriate given the
exposures held by the bank.
Moreover, the agencies note that the exposure
subcategory/national jurisdiction granularity
requirement is only a minimum granularity requirement.
Supervisory mapping function
The
proposed rule provided banks two methods of generating
LGD estimates for wholesale exposures and retail
segments.
First,
a bank could use its own estimates of
LGD for a
subcategory of exposures if the bank had prior written
approval from its
primary Federal supervisor to use internal estimates for
that subcategory of exposures.
In
approving a bank’s use of internal estimates of LGD, a
bank’s primary Federal supervisor would consider whether
the bank’s internal estimates of LGD were reliable and
sufficiently reflective of economic downturn conditions.
The
supervisor would also consider whether the bank has
rigorous and well-documented policies and procedures for
identifying economic downturn conditions for the
exposure subcategory, identifying
material adverse correlations between the relevant
drivers of default rates and loss rates given default,
and incorporating identified correlations into internal
LGD estimates.
If a
bank had supervisory approval to use its own estimates
of LGD for an exposure subcategory, it would use its own
estimates of LGD for all exposures within that
subcategory.
As an
alternative to internal estimates of LGD, the proposed
rule provided a supervisory mapping function for
converting ELGD into LGD for risk-based capital
purposes.
A bank
that did not qualify to use its own estimates of LGD for
a subcategory of exposures would instead compute LGD
using the linear supervisory mapping function:
LGD
= 0.08 + 0.92 x ELGD.
A bank
would not have to apply the supervisory mapping function
to repo-style transactions, eligible margin loans, and
OTC derivative contracts (defined below in section V.C.
of this preamble).
The
agencies proposed the supervisory mapping function
because of concerns that banks may find it difficult to
produce internal estimates of LGD that are sufficient
for risk-based capital purposes because LGD data for
important portfolios may be sparse, and there is limited
industry experience with incorporating downturn
conditions into LGD estimates.
The
supervisory mapping function provided a pragmatic
methodology for banks to use while refining their LGD
estimation techniques.
In
general, commenters viewed the supervisory mapping
function as a significant deviation from the New Accord
that would add unwarranted prescriptiveness and
regulatory burden to the U.S. rule. Commenters requested
more flexibility to address problems with LGD
estimation, including the ability to apply appropriate
margins of conservatism as contemplated in the New
Accord.
Commenters expressed concern that U.S. supervisors would
employ an unreasonably high standard for allowing own
estimates of LGD, forcing banks to use the supervisory
mapping function for an extended
period
of time.
Commenters also expressed concern that supervisors would
view the output of the supervisory mapping function as a
floor on internal estimates of LGD.
Commenters asserted that in both cases risk-based
capital requirements would be increased at U.S. banks
relative to their foreign competitors, particularly for
high-quality assets, putting U.S. banks at a competitive
disadvantage to foreign banks.
In
particular, many commenters viewed the supervisory
mapping function as overly punitive for exposure
categories with relatively low loss severities,
effectively imposing an 8 percent floor on LGD.
Commenters also objected to the proposed requirement
that a bank use the supervisory mapping function for an
entire subcategory of exposures even if it had
difficulty estimating LGD only for a small subset of
those
exposures.
The
agencies continue to believe that the supervisory
mapping function is a reasonable aid for dealing with
problems in LGD estimation.
The
agencies recognize, however, that there may be several
valid methodologies for addressing such problems.
For
example, a relative scarcity of historical loss data for
a particular obligor or exposure type may be addressed
by increased reliance on alternative data sources and
data enhancing tools for quantification and alternative
techniques for validation.
In
addition, a bank should reflect in its estimates of risk
parameters a margin of conservatism that is
related to the likely range of uncertainty.
These
concepts are discussed below in the quantification
principles section of the preamble.
Therefore, the agencies are not including the
supervisory mapping function in the final rule.
However, the agencies continue to believe that the
function (and associated estimation of the long-run
default-weighted average economic loss rate given
default within a one-year horizon) is one way a bank
could address difficulties in estimating
LGD.
However it chooses to estimate LGD, a bank’s estimates
of LGD must be reliable and sufficiently reflective of
economic downturn conditions, and the bank should have
rigorous and well-documented policies and procedures for
identifying economic downturn conditions for each
exposure subcategory, identifying changes in material
adverse relationships between the relevant drivers of
default rates and loss rates given
default, and incorporating identified relationships into
LGD estimates.
Pre-default reductions in exposure
The
proposed rule incorporated comments on the ANPR
suggesting a need to better accommodate certain credit
products, most prominently asset-based lending programs,
whose structures typically result in a bank recovering
substantial amounts of the exposure prior to the default
date – for example, through paydowns of outstanding
principal.
The
agencies believe that actions taken prior to default to
mitigate losses are an important component of a bank’s
overall credit risk management, and that such actions
should be reflected in LGD when banks can quantify their
effectiveness in a reliable manner.
In the
proposed rule, this was achieved by measuring LGD
relative to the exposure’s EAD (defined in the next
section) as opposed to the amount actually owed at
default.
Commenters agreed that the IRB approach should allow
banks to recognize in their risk parameters the benefits
of expected pre-default recoveries and other expected
reductions in exposure prior to default. Some commenters
suggested, however, that it is more appropriate to
reflect pre-default recoveries in EAD rather than LGD.
Other
commenters
supported the proposed rule’s approach or asserted that
banks should have
the
option of incorporating pre-default recoveries in
either LGD or EAD.
Commenters discouraged the agencies from restricting the
types of pre-default reductions in exposure that
could be recognized, and generally contended that the
reductions should be recognized for all exposures for
which a pattern of pre-default reductions can be
estimated reliably and accurately by the bank.
Consistent with the New Accord, the agencies have
decided to maintain the proposed treatment of
pre-default reductions in exposure in the final rule.
The
final rule does not limit the exposure types to which a
bank may apply this treatment.
However, the agencies have clarified their requirement
for quantification of LGD in section 22(c)(4) of the
final rule.
This section states that where the bank’s
quantification of LGD directly or indirectly
incorporates estimates of the effectiveness of its
credit risk management practices in reducing its
exposure to troubled obligors prior to default, the bank
must support such estimates with empirical analysis
showing that the estimates are consistent with its
historical experience in dealing with such exposures
during economic
downturn conditions.
A
bank’s methods for reflecting changes in exposure during
the period prior to default must be consistent with
other aspects of the final rule.
For
example, a bank must use a default horizon no longer
than one year, consistent with the one-year default
horizon incorporated in other aspects of the final rule,
such as the quantification of PD.
In
addition, a pre-default reduction in the outstanding
amount on one exposure that does not reflect a reduction
in the bank's total exposure to the obligor, such as a
refinancing, should not be reflected as a pre-default
recovery for LGD quantification purposes.
The
following simplified example illustrates how a bank
could approach incorporating pre-default reductions in
exposure in LGD. Assume a bank has a portfolio of
asset-based loans fully collateralized by receivables.
The
bank maintains a database of such loans that have
defaulted, which records the exposure at the time of
default and the losses incurred at and after the date of
default.
After
careful analysis of its historical data, the bank finds
that for every $100 of exposure on a typical asset-based
loan at the time of default, properly discounted average
losses are $80 under economic downturn conditions.
Thus,
the bank may assign an LGD estimate of 80 percent that
is based on such evidence.
However, assume that the bank division responsible for
collections reports that the bank’s loan workout
practices generally result in exposures on the
asset-based loans being significantly reduced between
the time the loan is identified internally as a problem
exposure and the time when the obligor is in default for
risk-based capital purposes.
The
bank studies the pre-default paydown behavior of
obligors that default within the next one-year horizon
and during economic downturn conditions.
In
particular, the bank uses rnal historical data to map
exposure amounts for asset-based loans at the time of
default to exposure amounts for the same loans at
various points in time prior to default and confirms
that the pattern of pre-default paydowns corresponds to
reductions in the bank’s overall exposures to the
obligors, as opposed to refinancings.
Robust
empirical analysis further indicates that pre-default
paydowns for assetbased loans to obligors that default
within the next one-year horizon during economic
downturn conditions depend on the length of time the
loan has been subject to workout.
Specifically, the bank finds that the prospects for
further pre-default paydowns diminish markedly the
longer the bank has managed the loan as a problem credit
exposure.
For
loans that are not in workout or that the bank has
placed in workout for fewer than 90 days, the bank’s
analysis indicates that pre-default paydowns on loans to
obligors defaulting within the next year during economic
downturn conditions were, on average, 50 percent of the
current amount owed by the obligor.
In
contrast, for asset-based loans that have been in
workout for at least 90 days, the bank’s analysis
indicates that any further pre-default recoveries tend
to be immaterial.
Thus,
provided this analysis is suitable for estimating LGDs
according to section 22(c) of the final rule, the bank
may appropriately assign an LGD estimate of 40 percent
to asset-based loans that are not in workout or that
have been in workout for fewer than 90 days.
For
asset-based loans that have been in workout for at least
90 days, the bank should assign an LGD of 80 percent.
Receive the New Member
Orientation Newsletters.
Understand Basel II
You will have the
opportunity to learn what members registered before you have
already learned. Understand better the Basel II environment,
projects, careers, challenges and opportunities.
Return
to Table of Contents
Return to
Index
Read more
about our
Certified Basel
ii Professional (CBiiPro)
program
Read more
about our
Certified Pillar 2 Expert
(CP2E)
program
Read more about our
Certified Pillar 3 Expert
(CP3E)
program
Read
more about our
Certified
Stress Testing Expert (CSTE)
program
 | |
