Skip to content

Negative loss quantiles in an all-gain model

Under a positive-loss convention, an all-gain model can produce negative VaR and expected shortfall.

On this page

State the original gain outcomes

Suppose losses are −10 with probability 90% and −5 with probability 10%. They represent gains of 10 and 5 units, respectively. The larger loss number −5 is the worse economic outcome in this gain-only model.

Select the worse tail

At 95% confidence, the lower loss quantile is −5. The worst 5% mass lies inside the 10%-probability outcome at −5, so expected shortfall is also −5. Mean loss of the whole model is −9.5, a different summary.

Original gain-only outputs
MeasureLoss units
Lower 95% quantile−5
95% expected shortfall−5
Whole-model mean loss−9.5

Keep any floor separately defined

Reporting zero instead would impose a nonnegative floor. That may be a distinct convention for another purpose, but it is not the signed measure calculated here. State a floor explicitly if a different task requires one.

Keep the declared-model scope

The example assumes only the two stated outcomes. It does not establish that a real position cannot lose money or that future risk is absent.

Further references