FINC6023 · Financial Risk Management
Operational Risk, Stress Testing & Scenario Analysis
Operational Risk, Stress Testing & Scenario Analysis covers the losses VaR cannot see. Operational risk is loss from failed internal processes, people, systems or external events (the famous rogue-trader cases live here). Because VaR only describes 'normal' conditions and gives only the minimum tail loss, you supplement it with stress testing (identify situations that cause extraordinary losses, with no probability attached) and scenario analysis (historical, prospective or portfolio-driven shocks). The key conceptual distinctions are statistical vs non-statistical and objective vs subjective probability.
What this chapter covers
- 01Operational risk definition and major loss events (Barings, SocGen, UBS, NAB)
- 02Why VaR is insufficient: describes 'normal' conditions and only a minimum tail loss
- 03Stress testing: non-statistical, no probability attached
- 04Scenario analysis: historical / prospective / portfolio-driven
- 05Moving one variable at a time ignores correlations (= sensitivity testing)
- 06Typical shocks: ±100bps yield curve, ±20% equity/vol, ±20bps spreads
- 07Objective vs subjective probability; the incentive to water down scenarios
Build a stress scenario and read why it differs from VaR
- 1 markEquity leg: a 20% fall on $5,000,000 = 0.20 × 5,000,000 = $1,000,000 loss.
- 2 marksBond leg: ΔP = −D·P·Δy = −4 × 8,000,000 × 0.01 = −$320,000, i.e. a $320,000 loss for a 100bps rise.
- 1 markCredit leg: 5% of $2,000,000 = $100,000 loss.
- 1 markTotal stress loss = 1,000,000 + 320,000 + 100,000 = $1,420,000.
- 1 markExplain: this is a deliberately severe, low-probability joint scenario chosen for its damage, whereas the 99% VaR only describes the loss exceeded 1 day in 100 under normal conditions — so a stress loss can be much larger.
Key terms
- Operational risk
- The risk of loss from inadequate or failed internal processes, people, systems, or external events. It excludes strategic and reputational risk and is illustrated by rogue-trader cases (Barings/Leeson, SocGen/Kerviel, UBS/Adoboli).
- Stress testing
- Identifying situations that would cause extraordinary losses, without attaching a probability. It is non-statistical and supplements VaR, which only describes normal conditions and the minimum tail loss.
- Scenario analysis
- Evaluating the portfolio under specified scenarios — historical, prospective or portfolio-driven. Composite VaR can blend stress and historical scenarios by assigning probabilities.
- Objective vs subjective probability
- Objective probabilities are estimated from data (backward-looking); subjective probabilities are judgmental forecasts. Stress and scenario work often relies on subjective probabilities, which invites a regulatory incentive to understate severity.
Operational Risk, Stress Testing & Scenario Analysis FAQ
Why isn't VaR enough on its own?
VaR describes only 'normal' market conditions and reports the MINIMUM loss in the tail (the quantile), not how deep the tail goes. It cannot capture rare, joint, extreme events. Stress testing and scenario analysis fill that gap by deliberately examining severe situations VaR was never designed to see.
What is the difference between scenario analysis and sensitivity testing?
Scenario analysis shocks several variables together in a coherent story (e.g. a recession). Sensitivity testing moves one variable at a time, which ignores correlations between factors. Moving variables one-by-one understates the joint damage of a real crisis.
Is a probability attached to a stress test?
No — pure stress testing is non-statistical and attaches no probability; it simply asks how bad the loss would be if a scenario occurred. Composite approaches can later weight scenarios by subjective probabilities, but the core stress test is judgment-driven, not distributional.
Exam move
Be able to name the marquee operational-loss events and state the operational-risk definition precisely. Practise computing a multi-leg stress loss and articulating, in one line each, why VaR is insufficient and how stress testing differs from sensitivity testing.
