Skip to content

FRM Exam Part II · Case Study: Model Risk and Model Validation

Mitigating and Quantifying Model Risk for FRM Part II

Updated 11 October 2026 · Fact-checked

Model risk is the chance of loss or poor decisions from a model that is wrong or misused. You mitigate it with validation, controls, conservative adjustments, reserves and limits on use. You quantify it by comparing alternative models, varying assumptions and sizing the output range. Then you hold a buffer for that uncertainty.

Understand Mitigating and Quantifying Model Risk

A model is a simplified view of reality. Every model can be wrong in its inputs, its assumptions, its code or its use. Model risk is the loss that follows from acting on a flawed or misused model. Supervisory guidance such as SR 11-7 says you cannot remove it. You can only identify it, limit it and manage it.

Mitigation works in layers. First, controls: independent validation, effective challenge, change control, documentation and an inventory of models. Second, limits on use: a model is approved only for stated products, markets and conditions, and may carry size limits or conditions. Third, adjustments: when a model is known to be weak, you apply a conservative adjustment or an overlay, such as a wider spread, a higher volatility or a haircut to the output.

The fourth layer is financial. A model reserve (or valuation adjustment) reduces the reported value of a position to reflect model uncertainty. Capital buffers cover losses that may arise beyond the reserve. Reserves protect earnings and valuations. Capital protects solvency.

To size any of this you must measure model uncertainty. Common approaches are: run several plausible models and look at the spread of results; vary the key parameters and see how outputs move (sensitivity analysis); compare to benchmark or challenger models; and use stress scenarios in which assumptions break. A wide spread means more uncertainty and a larger adjustment. A good answer links the measure, the action taken and the reason.

Key formulas to remember

Model range (uncertainty spread)
Range = Max(model values) − Min(model values)
Simple measure of uncertainty across competing models. It ignores how likely each model is.
Model reserve from a range
Reserve = Reported value (or mid-model value) − Conservative value
Policies often set the reserve as some fraction of the range. Use the fraction the question states.
Model-risk adjusted risk measure
Adjusted VaR = Base VaR × (1 + add-on %)
A multiplicative overlay. For example, a 10% add-on turns a 20 into 22.
Sensitivity of output to a parameter
Sensitivity = Change in output ÷ Change in parameter
A large value means the output depends heavily on a hard-to-estimate input.
Capital versus reserve
Reserve covers expected valuation error; capital buffer covers unexpected model error
A rule to recall for interpretation questions.

How to solve Mitigating and Quantifying Model Risk questions

Use this order for any question on mitigating or quantifying model risk.

  1. 1Identify the source of the model risk: input data, assumption, implementation or use outside its design.
  2. 2Decide whether the question asks for mitigation (control, limit, adjustment) or quantification (range, sensitivity, reserve).
  3. 3For mitigation, match the action to the source: bad assumptions need challenge or conservative adjustment; misuse needs limits on use.
  4. 4For quantification, list the alternative models or parameter values and compute the outputs.
  5. 5Compute the spread, the sensitivity or the reserve using the rule given in the question.
  6. 6Check the direction: a conservative adjustment must lower asset values or raise risk and liabilities.
  7. 7State the interpretation: what the number means and who must be told, such as model owner, validator or senior management.

Quickest way: Match source to remedy, then do the arithmetic

When to use it: Use when four options look plausible and time is short.

  1. Name the weakness in one phrase, such as illiquid input or untested assumption.
  2. Pick the remedy that acts directly on that weakness.
  3. Reject options that remove validation or let the model owner approve their own model.
  4. If numbers are given, compute max minus min or base times one plus add-on.
  5. Choose the answer that is conservative in direction.

Common mistakes in Mitigating and Quantifying Model Risk

  • Saying model risk can be eliminated by validation.

    Students treat validation as a cure.

    Fix: Validation reduces and reveals model risk. Residual risk always remains and needs limits, adjustments or buffers.

  • Confusing a model reserve with regulatory capital.

    Both are described as buffers.

    Fix: A reserve reduces valuation or earnings for expected uncertainty. Capital absorbs unexpected losses.

  • Applying a conservative adjustment in the wrong direction.

    Students focus on the number and ignore the sign.

    Fix: Lower asset values, raise liability values and raise risk measures. Check the sign before answering.

  • Treating the model range as a probability-weighted error.

    Range looks like a standard deviation.

    Fix: Range ignores model likelihood. It is a rough bound, so state that it is a simple indicator.

  • Letting the model developer validate the model.

    Developers know the model best.

    Fix: Effective challenge needs independence, competence and authority. Validation sits outside development.

  • Using a model beyond its approved scope without limits.

    The model seems to work for similar products.

    Fix: Restrict use to approved products and conditions. Extension needs fresh validation or a temporary limit and overlay.

Worked examples

Example 1

Three models value an illiquid exotic derivative at USD 10.2 million, USD 10.8 million and USD 11.4 million. The bank's policy books a reserve equal to 50% of the range below the highest value. The desk reports at USD 11.4 million. What reserve and adjusted value result? (Take the policy reading as: reserve = 50% × range.)

Show the solution
  1. Range = 11.4 − 10.2 = USD 1.2 million.
  2. Reserve = 50% × 1.2 = USD 0.6 million.
  3. Adjusted value = 11.4 − 0.6 = USD 10.8 million.
  4. The direction is conservative: the asset value falls.

Answer: Reserve USD 0.6 million; adjusted value USD 10.8 million.

Example 2

A bank's VaR model for a new product has a base 99% 10-day VaR of USD 40 million. Validation finds weak tail assumptions and requires a 15% overlay. Separately, the model's use is limited to positions of USD 500 million notional, and the desk holds USD 650 million. What is the adjusted VaR, and what action is needed on the position?

Show the solution
  1. Adjusted VaR = 40 × (1 + 0.15) = USD 46 million.
  2. Compare notional: 650 − 500 = USD 150 million above the limit.
  3. The position breaches the model-use limit, so it must be reduced or escalated for approval.
  4. Until then, the overlay applies and the breach is reported to risk governance.

Answer: Adjusted VaR is USD 46 million; the position exceeds the model-use limit by USD 150 million and needs reduction or formal escalation.

Exam tips

  • Read the verb: mitigate asks for controls, limits or overlays; quantify asks for a range, sensitivity or reserve.
  • Check the sign of every adjustment. Conservative means lower asset values and higher risk.
  • Pick answers that keep validation independent from development.
  • Expect case-style questions on London Whale or LTCM-type failures. Link the failure to the missing control.
  • Separate reserve (valuation, expected) from capital buffer (solvency, unexpected).

Practice questions from Case Study: Model Risk and Model Validation

Mitigating and Quantifying Model Risk: frequently asked questions

What is the difference between a model overlay and a model reserve?

An overlay adjusts a model's output, such as raising VaR by a percentage, because the model is known to be weak. A reserve is a valuation deduction that reduces the reported value of a position. Both are conservative responses to model uncertainty.

How do banks quantify model uncertainty?

They compare results from alternative or challenger models, vary key parameters and observe output changes, and run stress scenarios where assumptions fail. The spread of results indicates how much uncertainty exists. That spread guides the size of reserves or add-ons.

Can model risk be fully eliminated?

No. Validation, controls and limits reduce it but residual risk remains. Institutions manage it through conservative adjustments, reserves and capital buffers.

Why must model validation be independent?

Effective challenge needs people who are not invested in the model's success. Independent validators have the standing and incentive to find weaknesses. Without independence, errors and optimistic assumptions can go unchallenged.