Stress testing evaluates how a portfolio or institution would perform under extreme but plausible conditions — a 2008-style credit crisis, a sudden 200-basis-point rate shock, a sector-specific collapse — rather than relying on statistical models calibrated to normal markets.
Why It Matters
Regulators require it (DFAST and CCAR for large U.S. banks, similar regimes elsewhere) precisely because VaR and standard deviation are calibrated to normal conditions and systematically understate tail risk. A firm that only manages to its VaR number is flying blind on exactly the scenarios that produce solvency-threatening losses.
How It Works in Practice
- 1Define scenarios: historical replay (2008, 2020), hypothetical (a specific rate/credit/FX shock), or regulator-mandated (severely adverse macro scenarios)
- 2Reprice every position in the portfolio under each scenario's assumed market moves
- 3Aggregate the losses across positions, accounting for how correlations and liquidity typically deteriorate in a real crisis
- 4Compare the result against capital buffers or risk limits and report to the board and, where required, the regulator
Common Pitfalls
Scenarios are only as good as the imagination behind them — every crisis to date has featured at least one mechanism the prior generation of stress tests didn't model
Static stress tests don't capture how a firm might actually behave mid-crisis (forced deleveraging, hedging, liquidity hoarding), which can amplify or dampen the modeled loss
Reverse stress testing (starting from 'what scenario would cause failure?' rather than 'how bad is this specific scenario?') is underused despite often surfacing more relevant risks