CVaR Node
Conditional Value at Risk - Expected loss beyond VaR
Overview
CVaR (Conditional Value at Risk), also called Expected Shortfall, measures the expected loss when losses exceed the VaR threshold. While VaR tells you "99% of the time, you won't lose more than X," CVaR tells you "when you're in the worst 1% of outcomes, you'll lose an average of Y." This addresses VaR's major flaw: ignoring tail severity.
CVaR is crucial for institutional portfolio management and stress testing. It quantifies extreme downside risk, essential for capital allocation, position sizing, and ensuring portfolios can survive rare but catastrophic events. Unlike VaR, CVaR is a coherent risk measure (satisfies mathematical requirements for sound risk management).
Formula & Calculation
CVaR(95%) = -3.8% (average of those worst 5% of losses)
Interpretation: In worst 5% scenarios, expect to lose 3.8% on average
Parameters
| Parameter | Default | Description |
|---|---|---|
| confidence | 95%, 99% | Confidence level (95% = analyze worst 5% of scenarios) |
| lookback | 252-504 | Number of days for historical sample |
Common Use Cases
1. Extreme Risk Estimation
Quantify tail risk for stress testing. If portfolio CVaR(99%) = -8%, worst 1% of days lose ~8%. Use for capital planning and ensuring reserves can handle extreme losses.
2. Portfolio Stress Testing
Compare CVaR of different portfolio allocations to see which handles tail events better. A conservative portfolio with CVaR(99%)=-6% weatherscrashes better than aggressive portfolio with CVaR(99%)=-12%.
3. Tail Hedge Design
Use CVaR to determine put option sizes needed for protection. If CVaR(99%) = -8%, buy puts to exceed that threshold so losses are capped even in catastrophic scenarios.
4. Risk-Sensitive Capital Allocation
Allocate capital based on CVaR contribution. Strategies with worse CVaR get less capital. Ensures portfolio doesn't have concentration of tail risk in any single position or strategy.
Advantages & Limitations
Advantages
- •Accounts for tail severity
- •Coherent risk measure
- •Better than VaR for optimization
- •Regulatory preferred metric
- •Intuitive interpretation
Limitations
- •Requires large sample size
- •Assumes normal distribution (often violated)
- •Computationally intensive
- •Can be unstable with small lookback
- •Can't model unknown unknowns
Tips & Best Practices
📊 Use Both VaR and CVaR
VaR tells you threshold, CVaR tells you severity. Plot both to understand full risk profile.
🔄 Stress Test Assumptions
Historical CVaR assumes past extremes repeat. Consider adding hypothetical stress scenarios (flash crashes, liquidity crises).
⚡ Monitor Rolling CVaR
Recalculate monthly to track changing tail risk. Increasing CVaR signals changed market regime.
⚠️ Consider Uncommon Tail Events
Add "black swan" scenarios to CVaR calculation - rare events that historical data misses.