wilmott.com

Hi, assuming we came up with a "new" model to price CDOs (or maybe we have implemented an existing model) but we are not sure if the results make sense. Can anybody come up with an unambigous tests if the model works (besides repricing iTraxx/CDX tranches). Because with repricing tranches you could for example have the problem that you do not use the CDS termstructure instead you are using a flat curve which could make your model look wrong. // Crazyh

try to test who the model behave in the extrem cases of correlation and spreads. for exemple theses cases:1- High correlation with flat spreads.2- High correlation with low spreads3- High correlation + high spreads...results for theses cases can be easly explained...

i would look at the sensitivites as well. make sure for example that the theta on the equity tranche is the lowest among all the tranches (since they have all the gap risk - similar concept as the vol skew going up as maturity of option approaches). also look at deltas - equity deltas have to be highest. tranches below expected loss point have to have negative individual gamma (i.e. convexity with respect to say widest name blowing up 300bps) and positive net gamma (if all spreads widen 1bp).

very interesting points.ASbityakov, do you kow any papers discussing how gama and theta change?

Quote. tranches below expected loss point have to have negative individual gamma (i.e. convexity with respect to say widest name blowing up 300bps) and positive net gamma (if all spreads widen 1bp). would you please explain this with more details? I do not quite capture your view.

you can think about this in terms of idiosyncratic vs systemic risk. equity tranches (typically those whose lower attachment point is below the EL of the portfolio for example 0-3% and 3-7% if the expected loss of the portfolio is 8%) are short idiosyncratic risk and long systemic risk and vice-versa senior tranches are long idiosyncratic risk and short systemic risk. in other words if a single name blows up equity tranches will be hurt more while if all names sell off senior tranches will be hurt more. intuitively this makes sense as senior tranches worry about the overall economy - they get wiped out if there's a massive amount of defaults, not if there's a couple. equity tranches on the other hand get wiped out in a few defaults, they don't care if there's 20 or more because by then they will have been wiped out already.

J,The attached Merrill's paper, though quite old now, has a reasonable bit on tranche Gamma.