Stress Testing¶
Difficulty advanced
Overview¶
Stress testing evaluates portfolio performance under extreme but plausible market scenarios. Unlike VaR (which measures normal conditions), stress tests capture tail events.
Historical Scenarios¶
Major Historical Events¶
| Event | Date | S&P 500 | 10Y Treasury | VIX |
|---|---|---|---|---|
| Black Monday | Oct 1987 | -22.6% | - | - |
| Dot-com Burst | 2000-2002 | -49.1% | - | - |
| Financial Crisis | 2008-2009 | -56.8% | +15% | 80 |
| Flash Crash | May 2010 | -9.1% (intraday) | - | - |
| COVID Crash | Feb-Mar 2020 | -33.9% | - | 82.7 |
| SVB Crisis | Mar 2023 | -5% | + | - |
Typing Shocks: Infrastructure vs Regulatory¶
Not all crisis events transmit through the same channel. In crypto (and increasingly in markets with concentrated infrastructure dependencies — DTCC, CHESS, single-cloud exchanges), the empirical evidence separates two regimes:
| Shock type | Examples | Vol impact | Return impact (CAR) |
|---|---|---|---|
| Infrastructure failure | Exchange outage, custodian insolvency, oracle exploit, bridge hack | ~5.7× larger vol response than regulatory | −7.6% |
| Regulatory event | Enforcement action, ban announcement, tax-rule change | Baseline | −11.1% |
The headline finding from a 2026 event-study: vol impact differs by a factor of ~5.7×, but the return responses are statistically indistinguishable (p ≈ 0.81 for the difference). Both punch a roughly similar hole in price; one drives much more two-sided volatility on the way down and back.
Implication for hedge sizing¶
The two shock types call for different hedge instruments, not just different sizes:
| Goal | Infrastructure shock | Regulatory shock |
|---|---|---|
| Protect against price drop | Out-of-the-money puts work for both, similar notional | |
| Protect against volatility itself | Long-vol position (straddle, VIX-equivalent, long variance) | Smaller vol hedge — directional dominates |
| Pre-position cost | Higher (long vol bleeds theta) | Lower |
| Trigger to add | Concentration risk in infra · single-venue dependency | Pending policy events, hostile jurisdiction exposure |
Methodological caveat: the cited study is a 2026 event-study with a relatively small sample of categorized events and a 2018–2025 window. The qualitative ordering (infra > regulatory for vol impact) is robust across specifications; the precise multiplier (5.7×) is sensitive to the event window and the volatility estimator. Treat it as a direction, not a calibration target.
Practical Guidelines¶
- Multiple Scenarios — Test various types of stress
- Correlation Breakdown — Correlations increase in crises
- Liquidity Stress — Can you exit positions?
- Counterparty Risk — What if your broker fails?
- Leverage Amplifies — Small moves become large with leverage
- Regular Testing — Stress test quarterly or after major changes
- Action Plans — Have predefined responses for each scenario
Next Steps¶
- VaR/CVaR — Normal conditions risk measurement
- Hedging Strategies — Protecting against stress
- Drawdown Management — Managing losses