Skip to content

Loss-quantile jump at a cumulative boundary

A confidence level just above 95% moves the original model’s lower quantile from 10 to 50.

On this page

Hold outcomes and probabilities unchanged

Losses remain −10, 0, 10 and 50 with probabilities 30%, 40%, 25% and 5%. F(10)=95% and F(50)=100%. The only changed input is confidence.

Compare exact and just-above levels

Original lower-quantile boundary
ConfidenceVaR loss unitsES loss units
95% exactly1050
95.0001%5050
97.5%5050

Explain the constant tail average

At 95% the selected 5% mass is entirely loss 50. Just above 95%, the selected mass becomes slightly smaller and remains inside the same loss-50 atom. Every included unit of probability has the same loss amount, so ES remains 50.

Identify discretisation instead of a new scenario

The VaR jump is a feature of the declared finite distribution and quantile convention. No economic loss outcome changed. Retain confidence precision and the sample rule when comparing reports, rather than describing the jump as a newly observed market event.

Further references