CDO Random Recovery - Calibration Recovery Distribution

MaxCohen · June 15th, 2009, 10:40 am

I want to calibrate a base correlation + random recovery model.The recovery distribution R~Beta(alpha,beta) follows a beta distribution, where I have one free parameter beta. Alpha is defined so the mean of R matches that of the underlying pool.How to choose alpha?I would like to fit the 60-100 tranche, so need to make guess on its value. Also do the same for the 30-60 tranche. Is there some simple method used to imply 30-60 and 60-100 tranche spreads. Cheers, Sam.

katastrofa · June 15th, 2009, 6:44 pm

You can try to match the historical variance, if you're into that kind of thing. But I wouldn't bother and would just choose alpha of such value that there is enough variance to be able to get base correlations for all tranches.

MaxCohen · June 16th, 2009, 8:28 am

it is not good enough just to fit the standard detachments, if thats what you mean when you say all tranches?. need to make sure senior bespokes are priced appropriately, which means possibly making judgement on 60-100 and 30-60 in the case of CDX.

JoeyD123 · June 16th, 2009, 4:34 pm

For purposes of calibrating, you could always take a certain percentage (say 20-30%) of the 30-100% expected loss and assign it to 60-100%, and keep the remainder for 30-60%. To then convert that to running, just keep in mind that 60-100 will have a shorter duration than 30-100 because of the recoveries hitting a thinner tranche. 30-60 should have a duration a bit longer than 15-30...Hope this helps - this is more art than science this high up in the cap structure.

katastrofa · June 17th, 2009, 6:06 pm

QuoteOriginally posted by: samkhalilianit is not good enough just to fit the standard detachments, if thats what you mean when you say all tranches?. need to make sure senior bespokes are priced appropriately, which means possibly making judgement on 60-100 and 30-60 in the case of CDX.I mean all standard detachments.