Normal quantile and tail-mean multipliers

Locate the percentile threshold

At 95% confidence the standard normal quantile is approximately 1.644854. Multiplying dispersion by this coefficient and adding mean loss gives the normal-model VaR threshold.

Average the selected upper tail

At the same confidence, the upper-tail mean multiplier is approximately 2.062713. It exceeds the threshold because it averages losses above that threshold rather than locating the cumulative crossing.

Compare supported confidence levels

Keep the coefficients dimensionless

Both coefficients multiply a standard deviation in loss units. Neither is a currency amount, a probability percentage or the whole-model mean. At zero dispersion they do not create a spread around the constant loss.