Liquidity Adjusted VaR

[Sunil Natarajan]

Hi David,

In Culp's chapter (Identifying,Measuring &Monitoring;liquidity risk page 427).
LVaR=V(Mean-(confidence level*std dev)+0.5Spread)
LVaR(confidence level,confidence level#)=V{(Mean-(confidence level*std dev)+{0.5[(mean of spread)+(confidence level#*std dev of spread)]}.
In the second formulae there are 2 confidence levels. Could you please throw some light on this formulae.Could you please upload a spread sheet on the second formulae.

Regards,
Sunil Natrajan

 

[David Harper CFA FRM]

Hi Sunil,

Here is <10 min briefcast that explains both.
I just uploaded, to member section, the underlying spreadsheet used. LVaR sheet here.

Please note: in my opinion, Culp's formula is mistaken in regard to the +/- sign. See thread.

Thanks,
David

 

[Sunil Natarajan]

Hi David,
I went through through your excel sheet.The formulae in excel sheet is different from that given in Culp's chapter.
The formulae given in Culp's chapter is different. In the second case
LVaR=V{(Mean-(confidence level*std dev)+{0.5[(mean of spread)+(confidence level#*std dev of spread)]}
From what I can understand is that there are two confidence level alpha & alpha*. So the worst case spread applies a different confidence level say alpha* and not the original confidence level of alpha as used in the first part.
Is there any other author in our FRM curriculum list who has solved LVaR in a different way.

Regards,
Sunil Natrajan

 

[David Harper CFA FRM]

Hi Sunil,

I deliberately re-used the confidence level. Of course, a difference confidence level can be used for the spread; e.g., 95% VaR and 99% confidence for the deviate. Jorion has two different alphas. But frankly difference confidences here would strike me as pedagogical, why would somebody do that? But sure, a line can be added to take a second confidence level...I do not think this is particularly relevant. The point (i.e., to increase the spread) is conveyed by the model.

Now John raised question #36. The question implies the spread is two-tailed versus one-tailed. That is a second issue, my model doesn't do that either.

The more important different between mine and Culp's is that Culp's subtract the volatility then adds the spread. It is no point have a different confidence on the spread if the formula is incorrect in the first place I'd say the relevant difference btwn mine and Culp's is that Culp's has a glaring error.

David