Barclays 2007 Annual Report Download - page 94

Risk management
Credit risk management
92 Barclays PLC Annual Report 2007
A key element of the Barclays framework is the Masterscale. This has
been developed to record differences in the probability of default risk
at meaningful levels throughout the risk range (see table below).
In contrast to corporate businesses, retail areas do not bucket exposures
into generic grades or bands for account management purposes
(although they may be used for reporting purposes). Instead, accounts are
managed based on internal, product specific segmentations of accounts,
for instance, deriving from the cut-offs of the associated models. The
cut-offs may be in the form of a score, a probability of default, a measure
of forecast loss or a more sophisticated risk/reward based measure.
Exposure at default (EAD) represents the expected level of usage of the
credit facility when default occurs. At default the customer may not have
drawn the loan fully or may already have repaid some of the principal, so
that exposure is typically less than the approved loan limit. When the
Group evaluates loans, it takes exposure at default into consideration,
using its extensive historical experience. It recognises that customers may
make heavier than average usage of their facilities as they approach default.
For derivative instruments, exposure in the event of default is the
estimated cost of replacing contracts with a positive value should
counterparties fail to perform their obligations.
When a customer defaults, some part of the amount outstanding on their
loans is usually recovered. The part that is not recovered, the actual loss,
together with the economic costs associated with the recovery process
combine to a figure called the loss given default (LGD), which is expressed
as a percentage of EAD.
Using historical information, the Group can estimate how much is likely to
be lost, on average, for various types of loans. To illustrate, LGD is lower for
residential mortgages than for unsecured loans because of the property
pledged as collateral.
The level of LGD depends on: the type of collateral (if any); the seniority
or subordination of the exposure; the industry in which the customer
operates (if a business); and the jurisdiction applicable and work-out
expenses. The outcome is also dependent on economic conditions that
may determine, for example, the prices that can be realised for assets,
whether a businesses can readily be refinanced or the availability of a
repayment source for personal customers.
The Barclays Masterscale (Wholesale)
DG/TTC Default Probability
Band >=Min Mid <Max
1 0.00% 0.010% 0.02%
2 0.02% 0.025% 0.03%
3 0.03% 0.040% 0.05%
4 0.05% 0.075% 0.10%
5 0.10% 0.125% 0.15%
6 0.15% 0.175% 0.20%
7 0.20% 0.225% 0.25%
8 0.25% 0.275% 0.30%
9 0.30% 0.350% 0.40%
10 0.40% 0.450% 0.50%
11 0.50% 0.550% 0.60%
12 0.60% 0.900% 1.20%
13 1.20% 1.375% 1.55%
14 1.55% 1.850% 2.15%
15 2.15% 2.600% 3.05%
16 3.05% 3.750% 4.45%
17 4.45% 5.400% 6.35%
18 6.35% 7.500% 8.65%
19 8.65% 10.000% 11.35%
20 11.35% 15.000% 18.65%
21 18.65% 30.000% 100.00%
The ratings process
The term ‘internal ratings’ usually refers to internally calculated estimates
of PD. These ratings are combined with EAD and LGD in the range of
applications described previously. The ‘ratings process’ refers to the use
of PD, EAD and LGD across the Group. In Barclays, the rating process is
defined by each business. For central government and banks, institutions
and corporate customers many of the models used in the rating process
are shared across businesses as the models are customer specific. For
retail exposures, the ratings models are usually unique to the business
and product type e.g. mortgages, credit cards, and consumer loans.
A bespoke model has been built for PD and LGD for Sovereign ratings.
For Sovereigns where there is no externally available rating we use an
internally developed PD scorecard. The scorecard has been developed
using historic data on Sovereigns from an external data provider covering
a wide range of qualitative and quantitative information. Our LGD model
is based on resolved recoveries in the public domain, with a significant
element of conservatism added to compensate for the small sample size.
To construct ratings for institutions, corporates, specialised lending and
purchased corporate receivables and equity exposures, we use external
models, rating agencies and internally constructed models. External
models employed include Moody’s Credit Edge, rating agency ratings
and Moody’s RiskCalc. The applicability of each of these approaches to
our customers has been validated by us to internal rating standards.
The data used in validating these primary indicators are representative
of the population of the bank’s actual obligors and exposures and its
long-term experience.
Internally built PD models are also widely used. We employ a range of
methods in the construction of these models. The basic types of PD
modelling approaches used are:
– Structural models
– Expert lender
– Statistical
Structural models incorporate in their specification the elements of the
industry accepted Merton framework to identify the distance to default
for a counterparty. This relies upon the modeller having access to specific
time series data or data proxies for the portfolio. Data samples used to
build and validate these models are typically constructed by adding
together data sets from internal default observations with externally
obtained data sets from commercial providers such as rating agencies
and industry gathering consortia.