Brigo suggests modeling credit intensity by SSRD (shifted squared root diffusion) model. It is a combination of CIR and jump model with shifts added. The advantage is the easy calibration to CDS due to the analytical formulae implied by the model. However, to fit the initial term structure of CDS a term structure of shifts has been added which seems to be a rescue measure. On the other hand we can model credit intensity by BK (black karasinski) which can be implemented in a trinomial tree. Could somebody share her/his practical experience with SSRD for credit intensity, and if possible offer some suggestions about the best practice (models) of credit intensity modelling at the moment? Thanks in advance.

I have not used BK and can only give general comments.BK has some strange features, like the expected bank balance going to infinity for a finite maturity.The two methods are similar in spirit, with the deterministic shift resembling the shifts of the HW tree to match the curve.The desirable feature of SSRD is the jump, although it might produce some unintuitive ATM-forward implied volatilities in my experience. In addition, working with two or more processes can cause problems, as independent jumps will kill the correlation. Also, if the credit spread curve is very steep, then the shift will somewhat dominate the diffusion part.BK will produce more extreme events without the need for a jump, since the lognormal will have a heavier tail than the chi-square.Kyriakos

Thanks, Kyriakos. The interpretation of the parparmeters in SSRD seems to be an issue, for example, very large vols for a flat CDS curve (calibrated to CDS). Given this, do you have experience in pricing hybrid derivatives based on this model? Why should we try to make sure the shifts are positive as this will add burden to the optimization in the calibration? Cheers.