Contents Management report Corporate governance Consolidated financial statements Financial statements Pillar 3
Back testing
Back testing is a risk management technique to evaluate
the quality and accuracy of internal VaR models. In essence,
back testing is a routine comparison of model generated risk
measures (daily VaR) with the subsequent trading outcomes
(hypothetical or actual profit or loss). It is expected that the
calculated VaR will be larger than all but a certain fraction of
the trading outcomes, where this fraction is determined by the
confidence level assumed by the VaR measure.
Among others, the performance of the VaR model is dependent
on the risk factors covered by the VaR framework, the accuracy
of the methodology applied and on the quality of the market
data used to generate the historical scenarios. Inaccuracies
in these items can lead to an abnormal number of outliers
which could be an indication of inadequate quality of the
internal model. Another source for outliers are technical issues.
Using back tests, the quality of the VaR model can be assessed,
both in terms of the distributional assumptions, historical
market data validation and transaction or position registration.
In line with regulation, Rabobank uses the 99% confidence
level, 1-day holding period VaR for the purpose of back testing.
Back tests are carried out at consolidated level and at book
level, using both actual profit or loss and hypothetical profit
or loss. Back testing results are reported to the regulator on
a quarterly basis. Outliers are reported and individually analysed
if they exceed an operational threshold (50,000 for books with
a VaR smaller or equal 500,000, and 0.15 times VaR for books
with a VaR larger than 500,000).
The number of outliers over a rolling window of one year
determines an additional charge to the capital multiplier used
on the VaR and stressed VaR in the RC calculation. If the number
of outliers on consolidated level exceeds a given number based
on the 99% confidence level of the VaR measure, a capital
multiplier has to be applied. In 2015, there were 5 actual and
4 hypothetical P&L outliers which exceeded the consolidated
VaR. On 31 December, the additional charge is 0.4 for Rabobank
as the number of outliers was above the threshold of four
CRR (CRD IV).
Valuation
The valuation of the trading portfolio positions is based on, or
derived from, observable prices or curve inputs. The availability
of observable prices or curve inputs varies by product and
market, and may change over time. In some markets or for
certain products, observable prices or inputs are not available,
and fair value is determined using valuation techniques
appropriate for the particular product.The data sources are
consistent between various products and independent of any
Rabobank activity. If in exceptional cases, in a start-up phase,
curves are not independently maintained, a minimum of
a monthly independent validation is a requisite. These curves
will be subjectto a yearly review in orderto define if database
providers have caught up with the market evolution and
allow independent monitoring in the future.The department
accountable for the valuation process is independent of the
front office.
Rabobank aims to include all liquid inputs of its valuation
models as risk factors in its market risk models and to achieve
maximum alignment between the valuation and risk models.
For instance, for interest rate portfolios a large variety of
forecasting and discounting curves are used to value the
products within them. All these curves are treated as separate
risk factors in the risk models. The historical yield curve data
which is used to generate the historical scenarios for the VaR
calculation is derived from the yield curves which are used to
value the trades. By doing so Rabobank also aims to achieve
maximum correlation between the actual profit or loss and
hypothetical profit or loss figures.
Stress testing
EDTF25 Rabobank recognizes that VaR, due to its underlying
statistical assumptions, must be complemented by
stress testing for a more complete risk assessment. Stress testing
is used to measure events that are not captured by the VaR
model. It is instrumental in gauging the impact of extreme, yet
plausible predefined moves in market risk factors on the P&L of
individual trading and investment portfolios. Rabobank
designed a large number of global scenarios based on book
composition and current macro/economic financial markets
situations. Risk drivers captured by these scenarios include
among other things: tenor basis swap spreads, interest rates,
credit spreads, volatility and interest rate rotation.The shocks
applied are determined using historical calibration or are based
on expert judgment. The scenarios are global and
homogeneous for all geographical regions.
Rabobank uses sensitivity stress scenarios for the following risk
factor categories:
Interest rates;
Interest rate volatility;
Interest rate curve rotation;
Credit spreads;
Commodities;
Commodity volatility;
FX rates;
FX volatility;
Equities;
359 9. Market risk