Thanks very much for your comments. My question was intended to relate to a specific tranche, market premium and market base correlation, rather than a more general limit for minimum correlation that only depends on the number of assets.I suppose my question is this: given the recent volatility in the index tranche market, can anyone think of a circumstance where the standard models people use will all stop working. For example, in the last 2 weeks, random factor loading ("RFL") models stopped working, tranche correlation models also stopped working. Banks that were using these models, I imagine, have suffered.I am trying to understand what type of market move would prevent the standard (no RFL, gaussian copula) base correlation model from working, and, how likely such a market move would be.Cheers

the base correlation "model" is not really a model in the usual sense. it's an interpolation and translation tool. so, as long as most people use this tool to quote and look at the base correlation skew, it should "work".

QuoteOriginally posted by: amitabhGuys, With recent break down in the Equity/Mezz ..Tranches in CDX/ITRAXX., does that sound the deathknell of Base Correlation and market moves on to Compound Correlation. Can any prop guy here, give us the real deal on what exactly happenedI do not know the background of this story. Would you please tell me what happens about breaking down in the Equity/Mezz?

QuoteOriginally posted by: Herbiein the last 2 weeks, random factor loading ("RFL") models stopped working, tranche correlation models also stopped workingHi Herbie. Quite interesting stuff here. What do you mean exactly by "stopped working"?Y.

Hi,Apologises late reply. -1/(n-1) is for the average, so non-homogeneity shouldn't be an issue. If correlation were to go through this level, or as you rightly put it, the protection price above, one would sell the option/protection if comfortable with the Gaussian Copula assumptions. Under it's assumptions, the market is complete, and so the arbitrage would flow through the hedge.Of course, in reality, the market is not complete and there would be no arbitrage: one would have to seriously question the assumptions. I can imagine a large number of relative value ideas though.Interestingly enough, I think the key assumption that fails is actually not the coupling itself, Gaussian or whatnot, but more so the implied assumptions on the marginals: this is where the market fails to provide adequate hedges. The standard model is in essence a structural model where defaults can only happen through a diffusion to a barrier. While jumps-to-default may not really exist, some kind of jump in the asset process definitely exists: the standard model, in the way marginals are calibrated to CDS spreads, does not account for this. Then, there is also the issue of time homogeneity.Regards