VaR and Expected Shortfall are not competing dashboard ornaments. They answer two different questions about the same loss distribution: where the tail begins, and what the tail costs once the portfolio is already inside it.

01 / The object being measured

Let L be portfolio loss over a defined horizon. VaR_alpha(L) is the alpha-quantile of that loss distribution. At 99%, it gives a threshold that should be exceeded roughly one day out of one hundred under the model. That sentence already contains the fragility: horizon, distribution, sampling window and model regime are doing most of the work.

02 / Why VaR hides severity

VaR tells you the door into the tail. It does not tell you whether the room behind that door is shallow or catastrophic. Two portfolios can have the same 99% VaR while one contains linear exposures and the other contains short convexity, liquidity gaps or crowded unwind risk. The quantile can match while the loss mechanics are completely different.

03 / Expected Shortfall as a tail average

Expected Shortfall_alpha(L) = E[L | L >= VaR_alpha(L)] in the continuous case. In plain terms, it averages the losses beyond the threshold. That makes it more informative for capital, stress and model governance because it penalizes not only the probability of crossing the line, but also the depth of the damage after crossing it.

04 / Estimation discipline

A credible tail metric needs a scenario engine: filtered historical simulation, parametric volatility scaling, Monte Carlo, stressed windows, liquidity horizons and backtesting. The dangerous version is a single number with no diagnostics. The useful version shows exceedances, regime context, tail composition, desk-level contributions and sensitivity to assumptions.

05 / Interface logic

A good risk interface should display VaR as the trigger, Expected Shortfall as the consequence, and stress scenarios as the narrative. The user should immediately see whether risk is coming from volatility, concentration, correlation breakdown, nonlinear payoff shape or liquidity horizon. The metric becomes useful when it points to action.

VaR is the threshold. Expected Shortfall is the loss language beyond the threshold.

Reference note: Basel Committee market risk framework.