FRM Part II · FRM Exam Part II · Backtesting VaR
A bank moves from a 99% to a 95% confidence level for backtesting the same VaR model over 250 days, keeping the test on the number of exceptions. What is the main effect on the backtest's reliability?
Expected exceptions rise from 2.5 to 12.5, so the binomial test has more power to separate good from bad models. The trade-off is that a 95% level reveals less about extreme tail losses than a 99% level.
- ALower power is gained because fewer exceptions are expected
- BThe expected number of exceptions rises from 2.5 to 12.5, making the test better able to distinguish accurate from inaccurate modelsCorrect
- CThe expected exceptions fall from 12.5 to 2.5, reducing power
- DReliability is unchanged because the sample length is identical
Explanation
At 99%, expected exceptions are 250 x 0.01 = 2.5; at 95% they are 250 x 0.05 = 12.5. More expected exceptions make the count less noisy relative to its mean, so the test discriminates better, though it says less about extreme tail behaviour.
Did you get it right without looking?
One question tells you little. A timed set on Backtesting VaR shows your real accuracy, how long you take and where you lose marks.
More Backtesting VaR questions
- A bank's internal model produces 99% one-day VaR. Over the last 250 trading days it records 7 exceptions. Under the Basel traffic light fram…
- A risk manager reviews the Basel traffic light framework and asks why the green zone upper limit is set at 4 exceptions for a 99% VaR backte…
- A bank backtests its 99% one-day VaR over 250 days. To reduce the chance of wrongly rejecting a sound model, the validation team raises the …
- A bank's 99% one-day VaR model produces 9 exceptions over 250 trading days. Review shows that on 6 of these days the trading desk had booked…
- A bank's 99% one-day VaR shows 9 exceptions over 250 days. Investigation shows the VaR model prices an exotic option using a delta-only appr…
- A bank backtests a 99% VaR model over 250 days and rejects it if there are 5 or more exceptions. The true exception probability is 1%, and t…