Critical z-scores for bid-ask spread component of LVAR

[LeeBrittain]

Hi David,

In your notes you specifically mention that we should assume a one-tailed test for the z-score component when we are valuing the exogenous spread cost, however in the Schwesser practice exams they use a two tailed score. I assume that they are incorrect in using this? Has GARP confirmed that the right measure to use is a one-tailed score?

Thanks,
Lee

 

[David Harper CFA FRM]

Hi Lee,

Yes, one-tail for the spread just like the VaR. And GARP definitely knows, and agreed although it hasn't surfaced lately (e.g., http://forum.bionicturtle.com/threa...ontinuously-using-1-96-for-95-confidence.1733 ). I actually established this back in 2007 and 2008, so it's not an issue i've had to raise in years.

I do notice the 2012 FRM Handbook, see example question 26.5, correctly employs a 99% one-tailed 2.33 deviate on the spread. So I think it's settled. (The error came from a 2001 Culp reading.)

There are quite a few old threads on this (the links are pre our new forum so the internal links aren't good, but a search will work), but I still think common sense tells us that the spread needs to be be one-tail: e.g., if the mean spread is 1.0% with volatility of 0.5%, our LVaR is not interested in the favorable side of the distribution where the spread contracts to 1% -0.5%*deviate, but only the unfavorable side where the spread expands to 1%+0.5%*sigma. As it's a LC added to the VaR, I never could see how it could not be one-tailed deviate on the spread. Thanks,