FRM Exam Part I · Stress Testing

Stress Testing Methodologies and Model Challenges for FRM

Q: What is a reverse stress test?

It starts from a defined failure, such as insolvency or a breach of the capital minimum. You then identify the scenarios and factor moves that would cause it. It exposes vulnerabilities that top-down scenarios can miss.

Q: What is the difference between sensitivity and scenario stress testing?

A sensitivity test shocks a single risk factor, such as rates up 100 bp. A scenario test moves several factors together based on a coherent story, such as a recession. Scenarios capture interactions that sensitivity tests ignore.

Q: Why does correlation breakdown matter in stress testing?

Correlations estimated in calm periods usually understate how assets move together in a crisis. Diversification benefits then shrink and portfolio losses are larger than models suggest.

Q: How is model risk relevant to stress tests?

Stress tests rely on pricing models, parameter estimates and data that may fail under extreme conditions. Weak assumptions can make results unreliable, so models need validation, challenge and use of judgement.

Updated 11 October 2026 · Fact-checked

Stress testing asks what happens to a firm under extreme but plausible conditions. Sensitivity tests shock one risk factor. Scenario tests move many factors together in a coherent story. Reverse stress tests start from a failure outcome and search for the events that cause it. Key challenges are correlation breakdown, liquidity effects, model risk and aggregation.

Understand Stress Testing Methodologies and Model Challenges

A stress test measures loss under conditions that are unlikely but possible. VaR describes normal markets at a confidence level. Stress tests cover what lies beyond, where history gives few observations.

Sensitivity tests change one risk factor at a time by a set amount, for example a 100 bp parallel rate shift or a 10% fall in equity prices. They are simple, fast and easy to explain. They do not show how factors move together, so they can miss the real cause of a loss.

Scenario tests move many factors together in one coherent story. It can be historical (repeat 2008) or hypothetical (a sovereign default plus an oil shock). Historical scenarios are concrete but may not repeat. Hypothetical ones can capture new risks but depend on judgement, and a good one links the factors by an economic narrative.

Reverse stress tests work backwards. You fix a damaging outcome, such as breaching a capital ratio or insolvency, and then ask which scenarios would produce it. They find hidden vulnerabilities and challenge the view that the firm is safe.

Model challenges matter just as much. Correlations that are low in calm markets tend to rise toward one in a crisis, so diversification benefits shrink. Liquidity effects mean positions cannot be sold at model prices: bid-ask spreads widen, funding dries up and collateral calls rise. Model risk comes from simplified pricing models, unstable parameters and poor data. Aggregation across market, credit, liquidity and operational risk is hard because the risks interact and are measured on different bases. Simply adding them overstates the total only if diversification is real, and in stress it often is not.

Key formulas to remember

Sensitivity loss (linear approximation): ΔP ≈ Exposure × Shock; For a bond, ΔP ≈ −DV01 × shock in bp. Use full revaluation for large shocks or options.
Scenario loss: Total loss = Σ (loss from each position under the joint factor moves); All factors move together as the scenario specifies.
Two-asset portfolio variance: σp² = w1²σ1² + w2²σ2² + 2·w1·w2·ρ·σ1·σ2; Shows why a rise in ρ in stress increases portfolio risk.
Reverse stress test logic: Fix outcome (e.g. capital ratio < minimum) → solve for factor moves causing it; Start at the failure point, not the shock.

How to solve Stress Testing Methodologies and Model Challenges questions

Use this method for any question on stress testing types or their weaknesses.

1Identify the test type: one factor shocked (sensitivity), several factors moved together (scenario), or a starting failure outcome (reverse).
2If a calculation is needed, list the exposures and the shocks given in the question.
3Compute each position's loss, using full revaluation if the shock is large or the position is nonlinear. Then add up the losses.
4Check whether the question changes any assumption, such as correlation rising or liquidity costs being added, and recompute.
5For reverse tests, set the target loss and solve for the shock size that reaches it.
6For conceptual items, match the issue to its cause: correlation breakdown, liquidity, model risk or aggregation.
7Pick the answer that fits the exact wording. Eliminate options with absolute words like always or never.

Quickest way: Keyword matching for concept questions

When to use it: Use for conceptual multiple-choice questions where no numbers are needed.

One factor changed: sensitivity. Several factors in a story: scenario. Starts from failure: reverse.
Correlations jump in a crisis: correlation breakdown, so diversification is overstated.
Cannot sell at quoted prices or funding vanishes: liquidity effects.
Simplified pricing or unstable parameters: model risk.
Combining risk types: aggregation, which is difficult because of interactions.

Common mistakes in Stress Testing Methodologies and Model Challenges

Treating a reverse stress test as a scenario test with a bigger shock.
Both involve extreme outcomes, so they look alike.
Fix: A reverse test starts with the failure outcome and finds the causes. A scenario test starts with the cause and finds the loss.
Saying sensitivity tests capture interactions between risk factors.
Candidates assume any stress test is multi-factor.
Fix: Sensitivity tests move one factor at a time and ignore joint moves.
Assuming correlations stay at their normal-period values in stress.
Models use historical averages from calm periods.
Fix: Expect correlations to rise in crises. Diversification benefits shrink and loss estimates rise.
Ignoring liquidity in stressed valuations.
Mark-to-model prices look precise.
Fix: Add wider bid-ask spreads, longer liquidation periods, funding stress and margin calls.
Believing historical scenarios are the most severe possible.
Past crises feel like the worst case.
Fix: Past events limit imagination. Use hypothetical and reverse tests to cover events that have not happened.
Adding risk-type losses and calling it a full firm-wide figure without noting interactions.
Simple summing seems conservative.
Fix: Understand that market, credit and liquidity risks interact and reinforce each other in stress, so a simple sum can be wrong in either direction.

Worked examples

Example 1

A bond portfolio has a DV01 of $42,000 per bp. Using a sensitivity test, a parallel 150 bp rise in yields is applied. The portfolio value is $500 million. What is the approximate loss, and as a percentage of value?

Show the solution

Formula: ΔP ≈ −DV01 × shock in bp.
Loss = 42,000 × 150 = $6,300,000.
Percentage = 6,300,000 ÷ 500,000,000 = 1.26%.

Answer: Approximate loss is $6.3 million, or 1.26% of portfolio value. Convexity would make the true loss slightly smaller.

Example 2

Two assets each have a $100 million position and a stressed one-day standard deviation of 4%. In normal times ρ = 0.2. In a crisis ρ = 0.9. What is the portfolio standard deviation in dollars in each case, and the difference?

Show the solution

Each asset: σ = 4% × 100m = $4m.
σp = √(4² + 4² + 2 × 0.2 × 4 × 4) = √(16 + 16 + 6.4) = √38.4 ≈ 6.197.
Crisis: σp = √(16 + 16 + 2 × 0.9 × 16) = √(32 + 28.8) = √60.8 ≈ 7.797.
Difference = 7.797 − 6.197 ≈ 1.60.

Answer: Normal: about $6.20 million. Crisis: about $7.80 million. Correlation breakdown adds about $1.60 million of risk.

Exam tips

Know the three test types by their starting point: one factor, a joint story, or a failure outcome.
Questions often ask which limitation applies. Link correlation to diversification, liquidity to exit prices and funding, and model risk to parameters and pricing.
For calculations, check whether the shock is small enough for a linear DV01 approximation or needs full revaluation.
Beware of absolute statements such as always or never. Correct answers are usually qualified.

Practice questions from Stress Testing

Stress Testing Methodologies and Model Challenges in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Stress Testing Methodologies and Model Challenges: frequently asked questions

What is a reverse stress test?

It starts from a defined failure, such as insolvency or a breach of the capital minimum. You then identify the scenarios and factor moves that would cause it. It exposes vulnerabilities that top-down scenarios can miss.

What is the difference between sensitivity and scenario stress testing?

A sensitivity test shocks a single risk factor, such as rates up 100 bp. A scenario test moves several factors together based on a coherent story, such as a recession. Scenarios capture interactions that sensitivity tests ignore.

Why does correlation breakdown matter in stress testing?

Correlations estimated in calm periods usually understate how assets move together in a crisis. Diversification benefits then shrink and portfolio losses are larger than models suggest.

How is model risk relevant to stress tests?

Stress tests rely on pricing models, parameter estimates and data that may fail under extreme conditions. Weak assumptions can make results unreliable, so models need validation, challenge and use of judgement.