What it means
Value at risk (VaR) tells you a loss level that will only be exceeded on a small percentage of days, say 5%. Expected shortfall goes one step further and averages the losses on those 5% of days, which is why it is also called conditional value at risk.
The distinction matters because two portfolios can have identical VaR and completely different disaster profiles. One might lose a little more than the VaR threshold on its bad days, while the other occasionally loses many times that amount, and only expected shortfall distinguishes between them.
Regulators moved towards expected shortfall for exactly this reason, since the measure penalises portfolios with fat tails rather than treating everything beyond the threshold as equivalent. It also has a helpful mathematical property: combining two portfolios can never produce a higher expected shortfall than the sum of the separate figures, which is not always true of VaR.
Calculating it is conceptually simple. Rank the historical or simulated outcomes from worst to best, take the tail beyond your chosen confidence level, and average the losses in that tail.
The limitation is data. Expected shortfall depends entirely on the quality of the tail scenarios, and a model trained on calm periods will produce a comfortable figure that says almost nothing about a genuine crisis.
In practice
Real-world examples.
Example
A hedge fund reports both VaR and expected shortfall to investors each month. The gap between the two widens ahead of a market disruption, giving the risk team an early signal that tail exposure has grown.
Example
A bank sets desk-level limits using expected shortfall rather than VaR, so a trader cannot reduce a measured risk figure by selling far out-of-the-money options that rarely lose but lose enormously when they do.
Example
An insurer models catastrophe exposure across a portfolio of policies and reports the average loss in the worst 1% of simulated years. That figure, not the average year, determines how much reinsurance it buys.
Think of it
“Expected shortfall is average loss in the worst scenarios-what to expect when things go wrong.
Formula
Calculation
Expected shortfall at a given confidence level = the average of all losses that exceed the value at risk threshold at that level.
A trading desk records 100 daily profit and loss outcomes. At 95% confidence, the tail is the worst 5 days out of 100. Those five losses are $180,000, $150,000, $130,000, $120,000 and $120,000. The 95% value at risk is the smallest loss in that tail, $120,000, which is the level exceeded on 5% of days. The expected shortfall is the average of the five: ($180,000 + $150,000 + $130,000 + $120,000 + $120,000) / 5 = $700,000 / 5 = $140,000. So the desk can say that on a bad day it expects to lose at least $120,000, and that the average bad day costs $140,000, which is the figure a risk committee should be planning capital around.Case study
Seen in the real world.
Larkspur Capital Partners is a fictional asset manager created to illustrate the point. Its two strategies both reported a 95% one-day VaR of about $2,000,000, and the investment committee treated them as carrying equivalent risk.
When the risk team added expected shortfall to the monthly pack, the picture changed. The equity strategy showed an expected shortfall of $2,400,000, only slightly above its VaR, while the options strategy showed $6,100,000, because its rare bad days were catastrophic rather than merely poor.
In the illustrative outcome, the committee cut the options strategy's allocation by a third and required tail scenarios to be stress-tested quarterly. Nothing about the underlying positions had changed, only the measure used to look at them, which is the entire argument for reporting expected shortfall alongside VaR.
Watch out
Common mistakes.
- Using expected shortfall and value at risk as if they were the same measure, when one is a threshold and the other is an average beyond that threshold.
- Reporting a single confidence level without stating it, since expected shortfall at 99% is a very different number from the same measure at 95%.
- Trusting a figure built from a short, calm history, which systematically understates the tail it is supposed to describe.
Questions
People also ask.
Is expected shortfall always larger than value at risk?
Yes, at the same confidence level it must be at least as large, because it averages losses that are all at or beyond the VaR threshold.
Does a small business ever need this?
Rarely in this form, though the underlying idea, asking what the average bad outcome costs rather than just how often bad outcomes happen, applies to any concentrated exposure.
What confidence level should be used?
Market risk work commonly uses 97.5% or 99%, with the choice driven by regulation or internal policy rather than by which number looks better.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%