08 Practice exam continuously using 1.96 for 95% confidence

[hsuwang]

Hello David,

I've been working through the 08 practice exam and have found many cases where the answers were using 1.96 for a single tailed 95% confidence. I'm not sure if it is GARP's mistake..

for example, for the calculation of LVaR, 08' Practice I #36

Consider an asset worth USD 1 million whose 95th percentile VaR is USD 100,000. Suppose the bid-ask spread on the asset has a mean of USD .1 and stdev of USD .3. What is the 95t% liquidity adjusted VaR?

The answer takes 100,000 + 1mm .5(.1+.3(1.96))

I'm sure 1.96 is wrong for a one-tailed 95% test but for some reason I've seen this mistake all over the 08' practice... since these exam questions actually come from past exams, this is really frustrating because if I were to use 1.64, there won't be any choices that will match the answer..

Thanks you.

 

[David Harper CFA FRM]

Hi Jack

It is GARP's mistake. Actually, that question has two problems. Please see:
http://forum.bionicturtle.com/viewthread/764/
(i explicitly challenged this question in regard to the one- or two-tail and the response was ultimately "we agree with you" ...my challenge on this point survived considerable scrutiny frankly)
...so your instincts are just fine ... it should indeed be a one-tail 1.645 instead of 1.96
...don't even get me started on how much it frustrates me
...but the good news is: I know GARP knows about this historical mistake

the correct answer is illustrated in this spreadsheet: https://www.editgrid.com/bt/admin/sample_2008_q36

btw, the other mistake illustrate why I prefer to use Dowd's absolute VaR formula; i.e.,
-return + (volatility)*(deviate)

it looks silly at first glance (why not just use volatilty* deviate - return) ... but, it turns out that the +/- signage can lead to errors like committed in the '08 practice exam: they end up netting the additional spread in the wrong direction ... but Dowd's approach is more robust to the *addition* of the LVaR term (IMO), it is totally "safe" to *add* one-half the spread

Thanks, David