FRM Part II · FRM Exam Part II · Case Study: Model Risk and Model Validation
A risk manager values an exotic option using three internally approved models. The valuations are USD 4.10 million, USD 4.30 million and USD 4.80 million. The bank decides to quantify model risk as the difference between the highest and the lowest valuation produced by the alternative models, and to hold a reserve equal to half of that range. What is the reserve?
The reserve is USD 0.35 million. The valuation range across the models is 4.80 minus 4.10, or USD 0.70 million, and the policy holds half of it. Using the full range or only adjacent model pairs would give other figures that do not match the stated policy.
- AUSD 0.35 millionCorrect
- BUSD 0.70 million
- CUSD 0.25 million
- DUSD 0.20 million
Explanation
Range = 4.80 - 4.10 = 0.70 million. Half of the range = 0.35 million. The USD 0.70 million option forgets to halve the range; 0.25 is half of 4.80-4.30 (wrong pair of models); 0.20 is half of 4.30-4.10 (wrong pair).
Did you get it right without looking?
One question tells you little. A timed set on Case Study: Model Risk and Model Validation shows your real accuracy, how long you take and where you lose marks.
More Case Study: Model Risk and Model Validation questions
- A bank uses a vendor-supplied option pricing model as a black box. The vendor refuses to disclose the code and documentation is minimal. Whi…
- A risk manager reviews a fund that, like LTCM, has equity of USD 4.7 billion and assets of USD 125 billion. Its leverage ratio (assets/equit…
- A validation team reviews a bank's VaR model and finds that a spreadsheet macro used to aggregate position data silently drops positions who…
- In the London Whale case, the CIO's trading positions grew while VaR limits were exceeded, and the 10-Q VaR was temporarily reported using t…
- A validator wants to assess how sensitive a pricing model's output is to a key assumption, such as a volatility input, holding the model str…
- Two models, A and B, price the same loan portfolio. Both pass backtesting, but they give materially different fair values because each rests…