Conditional Value-at-Risk (Expected Shortfall)
Also known as: CVaR, expected shortfall, average value-at-risk, tail VaR, Koşullu Riske Maruz Değer (CVaR / Expected Shortfall)
Conditional Value-at-Risk (CVaR), also called Expected Shortfall, is a coherent tail-risk measure that quantifies the conditional expectation of losses beyond the Value-at-Risk threshold. It was introduced for optimization by Rockafellar and Uryasev (2000) and shown to be coherent by Acerbi and Tasche (2002), and it has replaced VaR as the regulatory standard under Basel III/IV.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use CVaR when you need a risk measure for continuous financial loss or return series and you care specifically about the severity of extreme tail losses, not just a single threshold. It is appropriate for portfolio risk management and optimization because, as a coherent measure, it satisfies monotonicity, positive homogeneity, translation invariance, and sub-additivity, so it correctly rewards diversification. It assumes a reasonably long, stationary loss series: the source recommends at least 100 observations, and below 250 the tail estimate becomes unreliable so simpler VaR is preferred. On non-stationary series the estimate is misleading and the data should first be made stationary.
Strengths & limitations
- Coherent risk measure: it satisfies monotonicity, positive homogeneity, translation invariance, and sub-additivity, so it correctly captures the diversification benefit of combining positions.
- Measures the average severity of losses beyond VaR, giving a fuller picture of tail risk than VaR alone.
- CVaR optimization can be reformulated as a convex linear program, making large portfolio problems tractable.
- Adopted as the regulatory standard (expected shortfall) under Basel III/IV.
- Needs a fairly long history: below about 250 observations the tail estimate is unreliable and plain VaR is preferred.
- On non-stationary series the CVaR estimate is misleading; the data must first be differenced or otherwise made stationary.
- Historical CVaR depends entirely on the observed sample tail and may understate risks never seen in the data.
- Harder to back-test directly than VaR because it is a conditional tail expectation rather than a single quantile.
Frequently asked
How is CVaR different from VaR?
VaR is the loss threshold you are unlikely to exceed at a given confidence level, but it says nothing about how large the loss is once that threshold is breached. CVaR is the average of those breaching losses, so it always equals or exceeds VaR and describes the severity of the tail, not just its cutoff.
Why is CVaR called a coherent risk measure?
Acerbi and Tasche (2002) showed that Expected Shortfall satisfies the four coherence axioms: monotonicity, positive homogeneity, translation invariance, and sub-additivity. Sub-additivity in particular means a diversified portfolio is never riskier than the sum of its parts, a property VaR can violate.
How much data do I need?
The source recommends at least 100 observations, but warns that below 250 the tail estimate becomes unreliable. With short samples a simpler VaR estimate is preferred over CVaR.
Can CVaR be used for optimization?
Yes. Rockafellar and Uryasev showed that minimising CVaR can be written as a convex problem that reduces to a linear program over the loss scenarios, so it scales to large portfolios and is widely used for tail-risk-aware portfolio construction.
Sources
- Rockafellar, R. T. & Uryasev, S. (2000). Optimization of Conditional Value-at-Risk. Journal of Risk, 2(3), 21-41. DOI: 10.21314/JOR.2000.038 ↗
- Acerbi, C. & Tasche, D. (2002). On the Coherence of Expected Shortfall. Journal of Banking & Finance, 26(7), 1487-1503. DOI: 10.1016/S0378-4266(02)00283-2 ↗
How to cite this page
ScholarGate. (2026, June 1). Conditional Value-at-Risk (Expected Shortfall). ScholarGate. https://scholargate.app/en/finance/conditional-value-at-risk
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- ARIMAEconometrics↔ compare
- EGARCHEconometrics↔ compare
- Quantile RegressionEconometrics↔ compare
- Realized VolatilityFinance↔ compare
- Value at RiskFinance↔ compare