Stress testing – central to risk control

By now it should be clear that VaR is inadequate as a measure of risk by itself. Risk
management must provide a way of identifying and quantifying the effects of extreme
price changes on a bank’s portfolio. A more appropriate risk measurement methodology
for dealing with the effect of extreme price changes is a class of methods known
as stress testing. The essential idea behind stress testing is to take a large price
change or, more normally, a combination of price changes and apply them to a
portfolio and quantify the potential profit or loss that would result. There are a
number of ways of arriving at the price changes to be used, this chapter describes
and discusses the main methods of generating price changes and undertaking stress
testing:

Ω Scenario analysis: Creation and use of potential future economic scenarios to
measure their profit and loss impact on a portfolio
Ω Historical simulation: The application of actual past events to the present
portfolio. The past events used can be either a price shock that occurred on a
single day, or over a more extended period of time.
Ω Stressing VaR: The parameters, which drive VaR, are ‘shocked’, i.e. changed and
the resultant change in the VaR number produced. Stressing VaR will involve
changing volatilities and correlations, in various combinations.
Ω Systematic stress testing: The creation of a comprehensive series of scenarios
that stress all major risk factors within a portfolio, singly and in combination. As
with the first two methods, the desired end result is the potential profit and loss
impact on the portfolio. The difference with this method is the comprehensive
nature of the stress tests used. The idea is to identify all major scenarios that
could cause a significant loss, rather than to test the impact of a small number
scenarios, as in the first two methods above.

One of the primary objectives of risk management is to protect against bankruptcy.
Risk management cannot guarantee bankruptcy will never happen (otherwise all
banks would have triple A credit ratings) but it must identify the market events that
would cause a severe financial embarrassment. Note that ‘event’ should be defined
as an extreme price move that occurs over a period of time ranging from one day to
60 days. Once an event is identified the bank’s management can then compare the
loss implied by the event against the available capital and the promised return from
the business unit.

The probability of an extreme event occurring and the subsequent assessment of
whether the risk is acceptable in prevailing market conditions has until now been
partly subjective and partly based on a simple inspection of historic return series.
Now, however, a branch of statistics known as extreme value theory (EVT) holds out
the possibility of deriving the probability of extreme events consistently across
different asset classes. EVT has long been used in the insurance industry but is now
being applied to the banking industry. An introduction to EVT can be found below.
Given the low probabilities of extreme shocks the judgement as to whether the loss
event identified is an acceptable risk will always be subjective. This is an important
point, particularly as risk management has become based increasingly on statistical
estimation. Good risk management is still and always will be based first and foremost
on good risk managers, assisted by statistical analysis.

Stress testing must be part of the bank’s daily risk management process, rather
than an occasional investigation. To ensure full integration into the bank’s risk
management process, stress test limits must be defined from the bank’s appetite for
extreme loss. The use of stress testing and its integration into the bank’s risk
management framework is discussed in the final part of this chapter.