Stress Testing Methodology for Traders

Statistical risk measures like VaR and Expected Shortfall describe what the recent past implies about normal-conditions risk. Stress testing does the opposite: it asks what would happen under specific, severe, often implausible scenarios and reports the loss. It is the risk method that catches what models miss, because the next crisis rarely looks like the average of the last decade.

The purpose

1. Identify hidden concentrations that look diversified under recent correlations but collapse together under stress.
2. Size the worst plausible loss in a way VaR cannot, because stress tests are not constrained by the historical sample.
3. Pre-commit to actions — predefined responses that reduce the impulse to freeze or double down when the scenario arrives.

Three families of stress test

Historical scenarios replay actual past crises on the current portfolio: the 2008 Lehman week, the 2010 Flash Crash, the 2015 Swiss franc unpeg, the 2020 COVID crash, the 2022 UK gilt crisis. Apply the actual percentage moves of each asset class to current positions and read the loss. The strength is realism — these moves actually happened. The weakness is that the next crisis will differ.

Hypothetical scenarios specify a coherent narrative and shock the relevant risk factors: a 200 basis point rate hike in a week, a 30% equity decline over a month, a 20% currency devaluation, a 50% spike in implied volatility. Specify shocks consistently across factors, including their correlations, then reprice the portfolio.

Reverse stress testing inverts the question: what scenario would cause the portfolio to lose a critical amount, say 30%? Working backward from a ruin threshold identifies the combinations of shocks that would break the book, often revealing vulnerabilities no forward test would have proposed.

Building a stress test program

• Catalog exposures by risk factor: equity beta, duration, credit spread, FX, volatility, commodity, liquidity, funding.
• Map each position to its factor sensitivities — delta, gamma, vega, duration, spread DV01.
• Apply shocks to factors and reprice positions, including nonlinear payoffs and option greeks, then aggregate to portfolio loss including any funding or margin effects.

Pitfalls

• Using average correlations during stress. Correlations converge to 1 in crises; use stress-specific correlations.
• Ignoring second-order effects. A rate shock changes option vega and gamma, not just delta.
• Forgetting funding and margin. A solvent portfolio can be liquidated by a margin call before the position recovers.
• Choosing scenarios that flatter the book. If every stress test shows a 5% loss, the scenarios are not stressful.

For each scenario that crosses a critical loss threshold, predefine a response: reduce position, hedge, raise cash, or halt new entries in that factor. Document the trigger and the action before the crisis, because during the crisis judgment is impaired.