Any banks out there actually use RFL for CDO pricing?Is there some specific distribution assigned to correlation?

Who cares how to price dead assets?

I do thats why I am asking?

Also isn't there still trading activity in index tranches. So this isn't a dead asset, yet.

QuoteOriginally posted by: GamalWho cares how to price dead assets?btw there are CDO's which are alive and well not backed by mortgages

Not all CDOs have been unwound, so they need to be valued.From what I know, GS uses RFL internally.

katastrofa/Yossarian22,Agree with you on your points. I think Gamals comment was rather dismissive.katastrofa I also heard GS have an RFL model that works too i.e. it fits the market. Do you have any more detail on this? I thought RFL couldn't fit the market across the capital structure even when introducing random recovery as well as correlation?Cheers, Sam.

Sent you a PM.

Would appreciate a PM on this too Thanks v much.

Right now, I think, RFL is dead. Not worth bothering with.

my contacts suggest that Gaussian copula with stochastic recovery rates is the industry standard at the moment.

@mj100% correct. Base Correlation turned out to be the Black-Scholes model of CDOs. Everything is built on top of that.

QuoteOriginally posted by: mjmy contacts suggest that Gaussian copula with stochastic recovery rates is the industry standard at the moment.how lovely is this looking back guys I'd add that now the point is often how to model the stochastic recoveryand BTW, RFL is so dead that I started working later than Aug 2009 (when this thread was born) and I've never even heard of it

If you are still insterested in synthetic pricing. Then not only sochastic/random recovery + base correlation is important but you should look at:* eliminating model arbitrage of first and second derivatives of expected loss curve EL[K] where K is the upper attachement of base/equity tranche* interpolation/extrapolation methodology of EL[K] to ensure EL[K]'>0 and EL[K]''<0* in the loss distribution pdf you want to eliminate areas of negative density. some funky optimization + good interpolation may help with this* fitting top part of capital stucture i.e. 60-100 tranches will manipulate the variance of loss of the pdf as you want sensible prices across the capital stucture

How are firms now dealing with CDO^2, considering that there is reasonable agreement on using base correlation + stoch recovery (Hitier?) for synth CDOs ?