Skip to content

FRM Part II · FRM Exam Part II · Validating Bank Holding Companies' Value-at-Risk Models for Market Risk

A bank's 99% one-day VaR model is backtested over 250 days and shows 6 exceptions. The expected number is 2.5. Using the Kupiec proportion-of-failures logic, a validator notes that exceptions are also clustered in a single volatile week. Which conclusion best reflects the limits of the exception count test?

An unconditional exception count ignores timing. Clustered exceptions violate independence and show the model reacts too slowly to changing volatility, a weakness that would remain even if the total count were acceptable. Independence or conditional coverage tests are needed to detect it.

  1. AThe unconditional count test ignores timing, so clustering suggests the model fails to adapt to volatility even if the total count were acceptableCorrect
  2. BClustering is irrelevant because only the total count matters for validation
  3. CClustering proves the model is too conservative
  4. DClustering means the confidence level should be lowered to 95%

Explanation

Counting exceptions ignores when they occur. Clustering shows that exceptions are not independent, which indicates VaR does not respond quickly to volatility changes. Conditional coverage or independence tests address this. Clustering signals underestimation in stress, not conservatism.

Did you get it right without looking?

One question tells you little. A timed set on Validating Bank Holding Companies' Value-at-Risk Models for Market Risk shows your real accuracy, how long you take and where you lose marks.

Practise Validating Bank Holding Companies' Value-at-Risk Models for Market RiskFree diagnostic mock

More Validating Bank Holding Companies' Value-at-Risk Models for Market Risk questions