Expected shortfall, also called conditional value at risk (CVaR), is the average loss in the worst scenarios, those beyond the Value at Risk threshold. It answers the question VaR leaves open: when losses are that bad, how bad are they on average?
A worked example
Take 100 daily returns and look at the worst five (the 5% tail). They are -4.0%, -3.5%, -3.0%, -2.8% and -2.5%.
| Measure | Calculation | Result |
|---|---|---|
| 95% one-day VaR | The 5th worst return | 2.5% loss |
| 95% expected shortfall | (4.0 + 3.5 + 3.0 + 2.8 + 2.5) / 5 | 3.16% loss |
The numbers are made up. Both describe the same tail, but ES shows how deep it goes. See value at risk.
Why it is preferred
- It sees the tail. Two portfolios can share a VaR but have very different losses beyond it.
- It is coherent. Artzner et al. (1999) showed that ES is sub-additive, so diversifying never increases measured risk, which VaR cannot guarantee.
- Regulators use it. Bank trading-book rules under Basel's market risk framework moved from 99% VaR to 97.5% expected shortfall.
How it is computed
Average the returns beyond the cut-off in historical data, or use a parametric or Monte Carlo model. Because it relies on few observations in the tail, it needs more data than VaR to estimate well.
Limits
- It is only as good as the tail data or model.
- It is harder to back-test than VaR.
- It still describes loss in a fixed period, not a path. For paths see drawdown.
See also risk management.
Frequently asked questions
What is expected shortfall?
It is the average loss on the days when losses exceed the VaR threshold, a measure of the severity of the tail.
What is the difference between VaR and expected shortfall?
VaR is the loss level not expected to be exceeded at a given confidence. Expected shortfall is the average loss when that level is exceeded.
Is CVaR the same as expected shortfall?
Yes. Conditional value at risk is another name for expected shortfall.
Why do regulators prefer expected shortfall?
Because it captures the size of tail losses and is a coherent risk measure, unlike VaR.