RBS 2012 Annual Report Download - page 247

RBS GROUP 2012
245
During 2012, an improved methodology was implemented for interest
rates, to more realistically represent the distribution of rate changes. The
enhanced model introduces a level-dependent scaling methodology for
interest rates, which removes the overestimation of rate fluctuations in
regimes of declining rates and leads to a swifter adaptation to changing
circumstances in times of increasing rates. At the point of implementation
the impact on the trading VaR was a decrease of £3.9 million, while the
interest rate VaR saw an increase of £1.4 million. The non-trading total
and interest rate VaR decreased by £0.5 million and £1.9 million
respectively.
SVaR - is applied to the trading portfolio and utilises data from a specific
one year period of stress. As with VaR, the technique produces estimates
of the potential change in the market value of a portfolio over a specified
time horizon at given confidence level. For the purposes of calculating
regulatory SVaR, a time horizon of ten trading days is assumed and a
confidence level of 99%.
In December 2012, the FSA confirmed the European Banking Authority
guidelines relating to SVaR. The FSA now requires the use of ‘Dynamic’
SVaR, where the worst one year period of stress is determined on a daily
basis.
Risks not in VaR (RNIV) - The RNIV framework has been developed to
quantify those market risks not adequately captured by VaR and SVaR
methodologies. The RNIV approach is used for market risks that fall
within the scope of VaR, but which are insufficiently captured by the
model methodology, for example due to the lack of sufficient historical
data. These risks are therefore assessed outside the VaR model.
The Group adopts two approaches to the quantification of risks not in
VaR (RNIVs):
x Some RNIVs are quantified using a (standalone) VaR approach. For
these RNIVs, two values are calculated: (i) the VaR RNIV; and (ii)
the SVaR RNIV.
x Some RNIVs are quantified using a stress scenario approach. For
these RNIVs, an assessment of ten-day extreme, but plausible,
market moves is used in combination with position sensitivities to
give a stress-type loss number - the stressed RNIV value.
For each legal entity covered by the FSA VaR model waiver, all RNIVs
are aggregated to obtain the following three measures: (i) Total VaR
RNIV; (ii) Total SVaR RNIV; and (iii) Total stressed RNIV.
In each case, no allowance is made for potential diversification in respect
of material RNIVs.
Incremental risk charge (IRC) - The IRC model aims to quantify the
impact of defaults and rating changes on the market value of bonds,
credit derivatives, and other related positions held in the trading book. It
is calculated over a one year horizon to a 99.9% confidence level, and
therefore represents a 1-in-1,000 loss over the following year. The
modelling framework differentiates between the liquidity of different
underlying instruments, with a minimum liquidity horizon of three months.
It also captures basis risks between different products referencing the
same underlying credit (e.g. bonds and credit default swaps (CDS)), and
between similar products with different contractual terms (e.g. CDS in
different currencies). The portfolio impact of correlated defaults and rating
changes is assessed with reference to the resulting market value change
of positions, which is determined using stressed recovery rates and
modelled credit spread changes. The average liquidity horizon at the year
end was 4.6 months.
In 2012, the IRC model was enhanced further; i) to better capture the risk
characteristics of sovereign exposure migrations and defaults; and ii) to
align the recovery rates for sovereign exposures to the banking book
internal ratings based approach.
All price risk (APR) - The APR model is applied to the corporate credit
correlation trading portfolio, subject to certain eligibility constraints
(principally that the underlying names are liquid corporate CDS
positions). The measure is calibrated to a 99.9% confidence level over a
one year time horizon. All material price risks, including defaults and
credit rating changes, are within the scope of the model. Of these, the
most significant are credit spread risk, credit correlation risk, index basis
risk, default risk, and recovery rate risk. In addition, losses due to both
hedging costs and hedge slippage are modelled. The overall APR capital
charge is floored at 8% of the corresponding standard rules charge for
the same portfolio. The average liquidity horizon at the year end was 12
months.
Model validation - A model assessment is performed before a new or
changed model element is implemented, and before a change is made to
a market data mapping. Depending on the results, it may be necessary to
notify the FSA before implementation. The form of internal validation
depends on the type of model and the materiality of the change.
In the case of VaR models, the following steps are considered. In some
cases, for example a minor change to a market data mapping, it will not
be necessary to perform all of the steps. However, in all cases there will
be an independent review and validation.
x Perform accuracy testing of the valuation methods used within VaR
on appropriately chosen test portfolios. Ensure that tests capture the
effect of using external data proxies where these are used.
x Back-test the approach using the relevant portfolio.
x Back-test the approach using hypothetical portfolio(s) where this is
helpful for isolating the performance of specific areas of the model.
x Identify all risks not adequately captured in VaR, and ensure that
such risks are captured via the risks not in VaR process.
x Identify any model weaknesses or scope limitations, their effect and
how they have been addressed.
x Identify ongoing model testing designed to give early warning of
market or portfolio weakness becoming significant.
x Perform impact assessment. Estimate the impact on total one-day
and ten-day 99% VaR at the total legal entity level and the major
business level, and individual risk factor level one-day and ten-day
99% VaR at the total legal entity level.