Basel
ii in the United States of America
from the
Basel ii
Compliance Professionals Association (BCPA)
the largest association of Basel ii Professionals in the
world
Final Rule, USA: Risk-Based Capital Standards:
Advanced Capital Adequacy Framework — Basel II
General quantification principles
The
final rule, like the proposed rule, requires data used
by a bank to estimate risk parameters to be relevant to
the bank’s actual wholesale and retail exposures and of
sufficient quality to support the determination of
risk-based capital requirements for the exposures.
For
wholesale exposures, estimation of the risk parameters
must be based on a minimum of five years of default data
to estimate PD, seven years of loss severity data to
estimate LGD, and seven years of exposure amount data to
estimate EAD.
For
segments of retail exposures, estimation of risk
parameters must be based on a minimum of five years of
default data to estimate PD, five years of loss severity
data to estimate
LGD, and five years
of exposure amount data to estimate EAD.
Default, loss severity, and exposure amount data must
include periods of economic downturn conditions or the
bank must adjust its estimates of risk parameters to
compensate for the lack of data from such periods.
Banks
must base their estimates of PD, LGD, and EAD on the
final rule’s definition of default, and must review at
least annually and update (as appropriate) their risk
parameters and risk parameter quantification process.
In all
cases, banks are expected to use the best available data
for quantifying the risk parameters.
A bank
could meet the minimum data requirement by using
internal data, external data, or pooled data combining
internal data with external data.
Internal data refers to any data on exposures held in a
bank’s existing or historical portfolios, including data
elements or information provided by third parties
regarding such exposures.
External data refers to information on exposures held
outside of the bank’s portfolio or aggregate information
across an industry.
For
new lines of business, where a bank lacks sufficient
internal data, a bank likely will need to use external
data to supplement its internal data.
The
agencies recognize that the minimum sample period for
reference data provided in the final rule may not
provide the best available results.
A
longer sample period usually captures varying economic
conditions better than a shorter sample period.
In
addition, a longer sample period will include more
default observations for LGD and
EAD estimation.
Banks
should consider using a longer-than-minimum sample
period when possible.
However, the potential increase in precision afforded by
a larger sample size should be weighed against the
potential for diminished comparability of older data to
the existing portfolio.
Portfolios with limited data or limited defaults
Many
commenters requested further clarity about the
procedures that banks should use to estimate risk
parameters for portfolios characterized by a lack of
internal data or with very little default experience.
In
particular, the GAO report recommended that the agencies
provide additional clarity on this issue.
Several commenters indicated that the agencies should
establish criteria for identifying homogeneous
portfolios of low risk exposures and allow banks to
apportion expected loss between LGD and PD for those
portfolios rather than estimating each risk parameter
separately.
Other
commenters suggested that the agencies consider whether
banks should be permitted to use the New
Accord’s standardized approach for credit risk for such
portfolios.
The
final rule requires banks to meet the qualification
requirements in section 22 for all portfolios of
exposures.
The
agencies expect that banks demonstrating appropriately
rigorous processes and sufficient degrees of
conservatism for portfolios with limited data or limited
defaults will be able to meet the qualification
requirements.
Section 22(c)(3) of the final rule specifically states
that a bank’s risk parameter quantification process
“must produce appropriately conservative risk parameter
estimates where the bank has limited relevant data.”
The
agencies believe that this section provides sufficient
flexibility and incentives for banks to develop and
document sound practices for applying the IRB approach
to portfolios lacking sufficient data.
The
section of the preamble below expands upon potential
approaches to portfolios with limited data.
The
BCBS publication “Validation of low-default portfolios
in the Basel II Framework”
also
provides a resource for banks facing this issue.
The
agencies will work with banks through the supervisory
and examination processes to address particular
situations.
Portfolios with limited data.
The
final rule, like the proposal, permits the use of
external data in quantification of risk parameters.
External data should be informative of, and appropriate
to, a bank’s existing exposures.
In
some cases, a bank may be able to acquire and use
external data from a third party to estimate risk
parameters until the bank’s internal database meets the
requirements of the rule.
Alternatively, a bank may be able to identify a set of
data-rich internal exposures that could be used to
inform the estimation of risk parameters for the
portfolio for which it has insufficient data.
The
key considerations for a bank in determining whether to
use alternative data sources will be whether such data
are sufficiently accurate, complete, representative and
informative of the bank’s existing exposures and whether
the bank’s quantification of risk parameters is
rigorously conducted and well documented.
For
instance, consider a bank that has recently extended its
credit card operations to include a new market segment
for credit card loans and, therefore, has limited
internal data on the performance of the exposures in
this new market segment.
The
bank could acquire external data from various vendors
that would provide a broad, market-wide picture of
default and loss experience in the new market segment.
This
external data could then be supplemented by the bank’s
internal data and experience with its existing
credit
card operations.
By
comparing the bank’s experience with its existing
customers to the market data, the bank can refine the
risk parameters estimated from the external data on the
new market segment and make those parameters more
accurate for the bank’s new market segment of exposures.
Using
the combination of these data sources, the bank may be
able to estimate appropriately conservative estimates of
risk parameters for its new market segment of exposures.
If the
bank is not able to do so, it must include the new
market segment of exposures in its set of aggregate
immaterial exposures and apply a 100 percent risk
weight.
Portfolios
with limited defaults.
Commenters indicated that they had experienced very few
defaults for some portfolios, most notably margin loans
and exposures to some sovereign issuers, which made it
difficult to separately estimate PD and LGD.
The
agencies recognize that some portfolios have experienced
very few defaults and have very low loss experiences.
The
absence of defaults or losses in historical data does
not, however, preclude the potential for defaults or
large losses to arise in future circumstances. Moreover,
as discussed previously, the ability to separate EL into
PD and
LGD is a key
component of the IRB approach.
As
with the cases described above in which internal data
are limited in all dimensions, external data from some
related portfolios or for similar obligors may be used
to estimate risk parameters that are then mapped to the
low default portfolio or obligor.
For
example, banks could consider instances of near default
or credit deterioration short of default in these low
default portfolios to inform estimates of what might
happen if a default were to occur.
Similarly, scenario analysis that evaluates the
hypothetical impact of severe market disruptions may
help inform the bank’s parameter estimates for margin
loans.
For
very low-risk wholesale obligors that have publicly
traded financial instruments, banks may be able to glean
information about the relative values of PD and LGD from
different changes in credit spreads on instruments of
different maturity or from different moves in credit
spreads and equity prices.
In all
cases, risk parameter estimates should incorporate a
degree of conservatism that is appropriate for the
overall rigor of the quantification process.
Other
quantification process considerations.
Both
internal and external reference data should not differ
systematically from a bank’s existing portfolio in ways
that seem likely to be related to default risk, loss
severity, or exposure at default.
Otherwise, the derived PD, LGD, or EAD estimates may not
be applicable to the bank’s existing
portfolio.
Accordingly, the bank must conduct a comprehensive
review and analysis of reference data at least annually
to determine the relevance of reference data to the
bank’s exposures, the quality of reference data to
support PD, LGD, and EAD estimates, and the consistency
of reference data to the definition of default in the
final rule.
Furthermore, a bank must have adequate internal or
external data to estimate the risk parameters PD,
LGD,
and EAD (each of which incorporates a one-year time
horizon) for all wholesale
exposure and retail segments, including those originated
for sale or that are in the securitization pipeline.
As
noted above, periods of economic downturn conditions
must be included in the data sample (or adjustments to
risk parameters must be made).
If the
reference data include data from beyond the minimum
number of years (to capture a period of economic
downturn conditions or for other valid reasons), the
reference data need not cover all of the intervening
years.
However, a bank should justify the exclusion of
available data and, in particular, any temporal
discontinuities in data used.
Including periods of economic downturn conditions
increases the size and potentially the breadth of the
reference data set.
According to some empirical studies, the average loss
rate is higher during periods of economic downturn
conditions, such that exclusion of such periods would
bias LGD or EAD estimates downward and unjustifiably
lower risk-based capital requirements.
Risk
parameter estimates should take into account the
robustness of the quantification process.
The
assumptions and adjustments embedded in the
quantification process should reflect the degree of
uncertainty or potential error inherent in the process.
In
practice, a reasonable estimation approach likely would
result in a range of defensible risk parameter
estimates. The choices of the particular assumptions and
adjustments that determine the final estimate, within
the defensible range, should reflect the uncertainty in
the quantification process.
More
uncertainty in the process should be reflected in the
assignment of final risk parameter estimates that result
in higher risk-based capital requirements relative to a
quantification process with less uncertainty.
The
degree of conservatism applied to adjust for uncertainty
should be related to factors such as the
relevance of the reference data to a bank’s existing
exposures, the robustness of the models, the precision
of the statistical estimates, and the amount of judgment
used throughout the process.
A bank
is not required to add a margin of conservatism at each
step if doing so would produce an excessively
conservative result.
Instead, the overall margin of conservatism should
adequately account for all uncertainties and weaknesses
in
the
quantification process.
Improvements in the quantification process (including
use of more complete data and better estimation
techniques) may reduce the appropriate degree of
conservatism over time.
Judgment will inevitably play a role in the
quantification process and may materially affect the
estimates of risk parameters.
Judgmental adjustments to estimates are often necessary
because of limitations on available reference data or
because of inherent differences between the reference
data and the bank’s existing exposures.
The
bank’s risk parameter quantification process must
produce appropriately conservative risk parameter
estimates when the bank has limited relevant data, and
any adjustments that are part of the quantification
process must not result in a pattern of bias toward
lower risk parameter estimates.
This
does not prohibit individual adjustments that result in
lower estimates of risk parameters, as both upward and
downward adjustments are expected.
Individual adjustments are less important than broad
patterns; consistent signs of judgmental decisions that
materially lower risk parameter estimates may be
evidence of systematic bias, which is not permitted.
In
estimating relevant risk parameters, banks should not
rely on the possibility ofU.S. government financial
assistance, except for the financial assistance that the
U.S. government has a legally binding commitment to
provide.
4.
Optional approaches that require prior supervisory
approval
A bank
that intends to apply the internal models methodology to
counterparty credit risk, the double default treatment
for credit risk mitigation, the IAA for securitization
exposures to ABCP programs, or the IMA to equity
exposures must receive prior written approval from its
primary Federal supervisor.
The
criteria on which approval will be based are described
in the respective sections below.
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
 | |
