Glossary term
Conditional Value at Risk (CVaR)
Conditional value at risk is a tail-risk measure that estimates the average loss in the worst outcomes beyond a chosen value-at-risk threshold.
Updated
Read time
What Is Conditional Value at Risk (CVaR)?
Conditional value at risk is a tail-risk measure that estimates the average loss in the worst outcomes beyond a chosen value-at-risk threshold. It is also commonly called expected shortfall. While value at risk asks, “How bad could losses be at a specified confidence level?” CVaR asks, “If losses are worse than that threshold, how bad are they on average?”
That difference matters because extreme losses are not all the same. A portfolio might have a 95% value at risk of $1 million, but the average loss in the worst 5% of scenarios could be $1.6 million, $3 million, or more. CVaR tries to measure the severity inside that tail rather than stopping at the cutoff.
Key Takeaways
- CVaR measures expected loss beyond a value-at-risk threshold.
- It is often called expected shortfall, average value at risk, or expected tail loss.
- It gives more information about severe losses than VaR alone.
- CVaR depends heavily on the model, confidence level, data window, and assumptions used.
- Risk managers use it for portfolios, trading books, stress analysis, and tail-risk comparisons.
The Basic Formula
A simplified CVaR expression is:
L represents loss, α is the confidence level, and VaRα is the loss threshold at that confidence level. Read plainly, CVaRα is the expected loss conditional on losses being at or beyond the VaRα cutoff.
For example, assume a portfolio's 95% VaR is $500,000. That means the model estimates losses will exceed $500,000 in the worst 5% of scenarios. If the average loss across those worst 5% scenarios is $780,000, the 95% CVaR is $780,000.
What It Adds to VaR
VaR is useful but incomplete. It identifies a threshold, not the depth of losses beyond the threshold. Two portfolios can have the same VaR and very different tail profiles. One may have losses clustered just beyond the cutoff; another may have rare but catastrophic losses. CVaR helps reveal that difference.
This is why CVaR is often favored for tail-risk management. It is more sensitive to the shape of the loss distribution after the cutoff, especially when returns are skewed, fat-tailed, leveraged, or exposed to liquidity stress.
How to Read It
CVaR should not be read as a promise. It is a model result. Historical simulation, Monte Carlo simulation, parametric assumptions, liquidity adjustments, and stress windows can all produce different numbers. A calm historical period can understate tail risk, while a crisis-heavy data window can make risk look persistently higher.
The measure is most useful when comparing portfolios under the same method. If one portfolio has a materially higher CVaR than another at the same confidence level, it may carry deeper expected losses in stress scenarios. But the comparison is only as credible as the model behind it.
CVaR is especially useful for portfolios with options, leverage, credit exposure, illiquid assets, or concentrated positions. Those portfolios may behave normally most of the time and then lose sharply in stress. A threshold-only measure can miss how much damage sits beyond the cutoff.
It is also useful in limit setting. A firm may allow a desk or strategy to take risk up to a CVaR limit, then require reduction, hedging, or approval if projected tail losses rise. The number becomes a governance tool, not just a statistical output.
The confidence level should always be stated. A 95% CVaR and a 99% CVaR answer different questions, and the higher-confidence measure usually focuses on rarer, more severe outcomes. Without the confidence level and time horizon, the number is incomplete.
The Bottom Line
Conditional value at risk looks past the loss threshold and into the tail. It helps investors and risk managers ask a sharper question: not just how often severe losses may occur, but how damaging they may be when they do.