liangfeng

Just picture a normal distribution.

We're looking at the left tail of this distribution.

95% - there's 5% to the left of the VAR number. (say the VAR number = -1,000,000)

99% - there's 1% to the left of the VAR number. (say the VAR number = -1,500,000)

What does that mean? The 99% VAR occurs further in the tail i.e. it's lower in numeric terms.

But VAR tells you about the maximum loss at a given % confidence so for 95%, your maximum loss is 1m but for 99%, your max loss is 1.5m.

HENCE, higher VAR for higher confidence level.

Unforseen

it depends on the wording. The example given was not precise enough. VAR will increase in MAGNITUDE if you lower the probability. In other words, the lower the probability the higher the expected loss.

susana

bpdulog, I don't agree

VAR gives you the MINIMUM expected loss given a certain period and %confidence level, loss can alway be higher than stated VAR

as the confidence level increases, so does the expected loss

as the period (days -> weeks -> months ->....) increases, so does VAR

both is logical and easy to remember, as the very highest loss (with either method) is @100% and you get closer if you move towards this

...and of course you would expect to get to this highest loss more probable in the next 10 years then by next tuesday

re negative and positive values: the thingie is calle "Value at risk", so it must be a positive number ($$$ or % @risk)

otherwise, it would be called "gain at risk" (which would be negative) ;-)

of course, if you develop VAR from historical figures f.e., you look at negative numbers, but the VAR would still be stated positively

Unforseen

I'm talking about absolute values here, you guys are over analyzing.

NO EXCUSES