November 6th, 2006, 4:07 pm
Hi ljcao, I'm trying to do this aswell.The main problem is "What happens to the spread when a default occurs?". We otherwise could fit some arbitrary dynamics to the credit index spread and option prices. Unfortunately, if the spread were not expected to change when default happens, then the skew on the option prices would be much greater than is seen in the market.If we set some arbitrary dynamics to the spread (or the equivalent short-rate), we can fit the model to the loss dynamics of the portfolio. See Schonbucher's paper on modelling loss transitions in a HJM style framework or the (similar) SPA model of Sidenius, Piterbarg and Anderson. However, much work is still needed to fit the model to the index and tranche spreads... never mind the index or tranche vols.Alternatively, we can try to model the individual names within the index - either in a copula or in a factor based approach. See the Markov chain model of Graziano and Rogers. This looks quite promising but there are no promises that it is possible to fit this to the index option prices (which you'd use for hedging).Perhaps the simplest approach is to create a model which is only calibrated to the index - not the tranche rates - and with an externally prescribed change in the spread upon default which can be used to hit the market option prices.You won't be fitting the model to the tranche spreads but the market isn't sophisticated enough to arbitrage index vol skew with equity tranche products yet... there just isn't the completeness required for that.Anybody else got any ideas on this?