LGD with a distribution

FinEng

New Member
Hi David,

Can you clarify the formula for LGD on the bottom of page 51 of the formula sheets.
It states PD=F*N(.)-Ve^(mu*T)*N(.).

Should it not be LGD=F*N(.)-Ve^(mu*T)*N(.) since PD = N(.) and we are trying to determine E(LGD|PD).

Faisal.
 

David Harper CFA FRM

David Harper CFA FRM
Subscriber
Faisal,

Yes, you are correct. It should be LGD. Apologies. See comments. I uploaded a corrected version. I still need to add errata entry (but this is only one so far surfaced for formula sheet). David
 

FinEng

New Member
Thanks David. You may also want to correct the notes.

Also, what is the best way to inform you of some of these errors or typos. A couple of months ago I had sent you an email but I didn't get a response so I'm thinking you may have not received it. Is there a central place where you would like me to post the corrections?

Also thanks for putting all of this together, especially the ipod versions. With a busy schedule and little time to study, the ipod versions make it easy and keep me occupied on the road.
 

Yun chiang Tai

New Member
Hello David:
the formula about the long term mean volatility of GARCH(1,1)(p.18 of 2008 formula) is a little wrong...
Persistence
GARCH (1, 1) is unstable if the persistence > 1. A persistence of 1.0 indicates no mean reversion.
A low persistence (e.g., 0.6) indicates rapid decay and high reversion to the mean.
The average, unconditional variance in the GARCH (1, 1) model is given by:
VL=Alpha0/(1-Alpha1+Beta) (wrong)
VL=(Alpha0)/(1-Alpha1-Beta) (correct)
I think you must want to write it in the first place like this one
VL=(Alpha0)/(1-(Alpha1+Beta))
( the 2007 version is wrong too, so I guess you have forgotten to change it...)
Chris

 
 

David Harper CFA FRM

David Harper CFA FRM
Subscriber
Hi FinEng: I try to respond to emails but i admit I prioritize the forum, so posting here in the forum is great! Please note, per my message above, I'll relay errors to the ERRATA section here. Thanks for liking the ipod version - I am glad you are telling me this because downloadable content is very easy for people to steal but i am glad to do it if paying customers get a benefit.

Hi Chris: You are absolutely right, Geez, that escaped me both years, my mistake. Ugg. See errata. Thanks.

David
 

Yun chiang Tai

New Member
Hello David:
The formula of LGD derived from the PD in some sence shows that they are not indepedent, right?
But in the DeSergvigny and others we often assume that PD and LGD are independent, right??
But as the formula similar to BSM model, it shows that they are not...Right?
How can we think about it intuitively?
Thank you...
Chris
 

David Harper CFA FRM

David Harper CFA FRM
Subscriber
Chris,

Re: "The formula of LGD derived from the PD in some sense shows that they are not independent, right?"
Yes, true, that is an EXCELLENT point (I had not considered that)! The conditional LGD here is implicity correlated with PD.

But, also true in regard to de Servigny and Ong: otherwise we ASSUME INDEPENDENCE
I would add: as does Basel II IRB treat them as independently

In short, we are generally treating PD & LGD as independent.

"How can we think about it intuitively?" I do not have a way, sorry. Here is why i have not bothered to try: it is a totally unreliable method (I feel GARP should drop this AIM). The LGD generated by this structural approach is typically much too low to be realistic. Consider Stulz' own example, it is something like $100K LGD on $100MM, if i recall correctly...

David
 
Top