Expected shortfall (ES)

Expected shortfall answers the question value at risk leaves open: given that a loss has exceeded the VaR threshold, how bad is it on average? It is the mean of all losses in the tail beyond the VaR cutoff, rather than the single boundary value VaR reports.

For a given confidence level, say 95%, expected shortfall is calculated by averaging all losses that fall within the worst 5% of outcomes, using historical, parametric, or Monte Carlo simulation-based methods to generate that tail. This makes it sensitive to the shape and severity of extreme losses in a way VaR, which only marks where the tail begins, is not.

Because it accounts for the magnitude of tail losses rather than just their frequency, expected shortfall is considered a more conservative and theoretically better-behaved risk measure — it is "coherent" in the technical sense that VaR is not, meaning it never understates the combined risk of a portfolio relative to its parts. Regulatory frameworks for bank trading books have shifted toward expected shortfall for this reason, often used alongside maximum drawdown. Its main pitfall is that it requires a larger, more reliable sample of tail events to estimate accurately, so it can be noisier than VaR when data on extreme losses is scarce.