Expected Shortfall

Unlike Value-at-Risk, Expected Shortfall satisfies the property of sub-additivity, making it a coherent risk measure. This means the risk of a combined portfolio is always less than or equal to the sum of the individual assets' risks.

• Mathematical guarantee: ES(A + B) ≤ ES(A) + ES(B)
• Diversification benefit: ES correctly reflects that merging portfolios does not increase total risk
• Regulatory alignment: This property is why the Basel Committee on Banking Supervision mandated ES over VaR for market risk capital calculations

Expected Shortfall provides a complete picture of the loss distribution tail by averaging all losses beyond the VaR threshold, rather than just identifying a single quantile boundary.

• Captures tail fatness: ES increases as the tail of the distribution becomes heavier, even if VaR remains unchanged
• No cliff effect: A portfolio with a 99% VaR of $10M and extreme losses of $100M beyond that threshold will have a high ES, alerting risk managers to hidden leverage
• Monotonic with risk: ES always ranks riskier distributions higher, unlike VaR which can paradoxically favor riskier positions

Expected Shortfall lacks the statistical property of elicitability on its own, meaning there is no natural scoring function to directly backtest ES forecasts without additional assumptions.

• Joint elicitability: ES is jointly elicitable with VaR, requiring both measures to be evaluated together for rigorous backtesting
• Practical workarounds: Risk managers use quantile regression and calibration tests to validate ES models
• Regulatory response: The Basel Committee introduced the P&L Attribution Test and Backtesting Exceptions framework specifically to address ES validation in the Fundamental Review of the Trading Book (FRTB)

Expected Shortfall can be expressed as a spectral risk measure, where losses are weighted by a risk-aversion function that emphasizes the tail.

• Weighting function: ES at confidence level α weights all losses beyond the α-quantile equally, unlike VaR which assigns zero weight to losses beyond the threshold
• Continuous generalization: This connects ES to the broader family of distortion risk measures, where the probability distribution is transformed to overweight adverse outcomes
• Practical interpretation: ES represents the expected loss in the worst (1-α)% of scenarios, providing an intuitive dollar-value interpretation for capital allocation

Expected Shortfall is a convex function of portfolio weights, enabling its use in large-scale portfolio optimization problems where VaR's non-convexity causes computational intractability.

• Global optimum guaranteed: Minimizing ES subject to linear constraints yields a unique global solution using standard convex solvers
• Scenario-based formulation: ES can be expressed as a linear programming problem over discrete scenarios, making it tractable for multi-asset portfolio construction
• Real-world application: Institutional asset managers use ES minimization to construct tail-risk-constrained portfolios that explicitly limit expected losses during crisis periods

The largest peak-to-trough decline in a portfolio's cumulative returns over a specified period. While Expected Shortfall is a forward-looking statistical measure, Maximum Drawdown represents the worst-case realized historical loss.

• Captures the duration and magnitude of underwater periods
• A 50% drawdown requires a 100% gain to recover, highlighting the asymmetry of losses
• Often used alongside Expected Shortfall to provide a complete picture of historical vs. probabilistic tail risk

A simulation technique that projects portfolio losses under hypothetical, severe scenarios that are plausible but historically unprecedented. Unlike Expected Shortfall, which relies on historical or modeled statistical distributions, stress testing imposes specific narrative shocks.

• Scenarios include geopolitical crises, sudden correlation breakdowns, and liquidity freezes
• Complements Expected Shortfall by capturing regime shifts that statistical models may miss
• Required by CCAR and DFAST regulatory frameworks for systemically important financial institutions

A self-reinforcing cycle of forced selling where declining asset prices trigger margin calls, leading to further sales and a rapid evaporation of market depth. Expected Shortfall estimates can severely underestimate losses during these events.

• Market impact costs become non-linear and explosive during cascades
• Feedback loops between leverage, collateral, and price amplify tail losses
• Understanding cascades is essential for adjusting Expected Shortfall estimates to account for endogenous liquidity risk

A phenomenon during market crises where historically uncorrelated or negatively correlated assets suddenly move in the same downward direction. This nullifies diversification precisely when it is most needed, causing Expected Shortfall estimates based on normal-period correlations to be dangerously optimistic.

• Diversification benefits tend to evaporate in the left tail
• Requires modeling of tail dependence using copulas or regime-switching frameworks
• Critical for multi-asset portfolio Expected Shortfall calculations under stressed conditions