krot,Assume T=1 year, recovery rate 0, flat discount curve...let s try to price the senor tranche:let s call S the ongoing premium and S(t) the expected survival curve.we have:S * S(T) = 1 - S(T)Now, the definition of S(T) is (TrancheSize-TrancheLoss)/TrancheSize.If the index spread is large enough, S(T) = 0.And if S(T) =0, then S is infinity.do you agree?

QuoteOriginally posted by: krotDiceMan,The (weighted) sum of (non-overlapping which sum to 100%) tranche spreads better equal the spread of the index - otherwise you can just get all the tranches and play against the index.what do you mean by "weighted"?spreads are not additive - expected loss is. so only if you mean "survival probability weighted spread" then it makes some sense.sum of spreads on tranches will be always higher than index spread (due to the shape of expected loss vs. subordination)

hojdard,fair clarification (I should have been more clear). Take DiceMan example, then from the weighted sum of EL, one gets the size-weighted sum of inverses, i.e. sum of 1/(1+si) equals 1/(1+s) (si spread on tranche i).Also true that for huge index spreads S(T) (i.e. survival) can be 0 and then s is infinite.However, realistically: looking at senior and mez tranches, their spreads are small, hence as a quick calculation one can even take the weigthed literally as 1/(1+s) = 1-s for small s (e.g. looking at 9-12, 12-22, 22-100 tranches).Again, if one wants to be precise the proper "weighting" has to be used.Any comments on why the Base Corr as calculated by JPM spreadsheet differ from the ones quoted by dealers?

haven't looked into it, but that may be because the jpm sheet is a 1 timestep cdo and therefore the coupon is an approximation...dealers calculate intermediate timesteps, compute the spread*(remaining notional) at times t<maturity, and therefore get a real premium leg expectation and not the riskyduration approximation.

QuoteOriginally posted by: krothojdard,Any comments on why the Base Corr as calculated by JPM spreadsheet differ from the ones quoted by dealers?one comment would be that nobody uses this s/s....a.k.a nobody uses "Homogenous Large pool model"

I had decided to go away, but I am a little interested in this point, because of course I cannot see what other dealers do, you guys actually know more than me.Are other dealers quoting Base Correlations? Obviously if they are quoting compound correlations they will be different, they are completely different things.Of course if other dealers are quoting Base Correlations but are using a model which they do not reveal there are any number of reasons why the answers will be different. I kinda think it should be down to them to say why it's different, but there are loads of possible model differences: use of non-homogeneous model, different treatment of basis to theoretical, use of correct interest rates, use of full credit curve, use of non-normal copula.The point of the simple large pool model is not that it is intended to be wonderfully accurate, but rather if a dealer does quote such Base Correlations using the simple model, then you can reproduce them. Otherwise you can't. The analogy is CDSW which is the market standard for calculating unwinds for CDS - the default settings which are commonly used use a simple discounting process, and flat credit curves and constant recovery rate. There are many ways to make CDW models more accurate, some of the more interesting recent ones include stochastic recovery rates. But this is a calibration model, it's important to be able to value something else from a given market level.Going back to sleep...RegardsLee

Lee, a number of other major dealers have started quoting base correlation some time ago (in the US, don't know about Europe). However, as one would expect, they don't use the JMP model - at least as far as I know.I cannot agree more that it would be useful to have a standard for quoting tranches in terms of something relatively stable relative to the index level. Base correlation and a simple model may be one solution. However, there are a number of other alternatives...

dealers started to quote base correlation in the US and in Europe. there are 2 different types of base correlation: one coming from JP Morgan paper and one from the Bear paper. dealers tend to quote the base calculated from the JP Morgan one. as far as i am concerned, i do match exactly bid/offer as well as deltas of all these dealers.

Hi Jarod,I don't find the same Base Correlations as JPMorgan. So I have two questions.1. Do you convert the Upfront premium into running one? If so, How? (Upfront/Duration of the Equity + 500 bps ? or another method)2.Do you find the Equity Implied (or Base) Correlation using the following : you don't do the latter conversion but find the correlation thats matches : Upfront + Feeleg(0-3%, BC) = DefLeg (0-3%, BC)When I try the 2 methods I don't have exactly the same results but there is up to 7 points of difference with JPMOf course, when I take the BC for the Equity tranche quoted by JPM, I match the others ones.Thanks,

You should read "Upfront + Feeleg(0-3%, BC) = DefLeg (0-3%, BC)" asUpfront + Feeleg(0-3%, BC, 500bps as premium) = DefLeg (0-3%, BC)Thanks

i do not convert upfront premium into a running one. i keep the standard Running 500bps + Upfront structure. Base correlation and the one implied by the "old model" are obviously the same for an equity piece (0-3% in that case). so it is just a reverse engineeting of the gaussian copula model in that case. u do not match bid/offer correlation shown by brokers?

Thanks Jarod,I haven't seen the Brokers' qutations yet.I will check this

Hi all, a question about interpreting base correlations. I'm looking at the JP large-pool model, on an index like the iTraxx. To start with assume all correlations are flat, say all basecorrs are 15%, and solve for tranche prices. I get s.th. like 40%/300bp/75bp/20bp/3bp. Now take a senior tranche, eg. 9-12%, which should be "long correlation" for just about any interpretation of correlation - in the sense that increased correlation in the portfolio will mean more risk to this tranche. Increase its price, to 25bp say, and reduce the 12-22% spread to 1bp (just to retain a solution for its base corr). Now solve for correlation - the base corr for 0-12% will *drop*.This isn't a surprise when you look at the bootstrapping process (with a higher spread for 9-12% there's more expected loss in the 0-12% 'equity' compared to 15% flat, which corresponds to lower 'equity' correlation). However I find it rather counterintuitive. Am I thinking about this the wrong way, or are basecorrs for senior tranches a little awkward to interpret (because they are so tied in to what's going on underneath them)?

kartsy:the base correlation is not flat it's upward sloping! I think that's why you got confused.

QuoteOriginally posted by: karsty...Now take a senior tranche, eg. 9-12%, which should be "long correlation" for just about any interpretation of correlation - in the sense that increased correlation in the portfolio will mean more risk to this tranche. Increase its price, to 25bp say, and reduce the 12-22% spread to 1bp (just to retain a solution for its base corr). Now solve for correlation - the base corr for 0-12% will *drop*.This isn't a surprise when you look at the bootstrapping process (with a higher spread for 9-12% there's more expected loss in the 0-12% 'equity' compared to 15% flat, which corresponds to lower 'equity' correlation). However I find it rather counterintuitive. I agree, base correlation in that sense is not very intuitive. BTW, it does have the SAME sensitivities regardless of the seniority in the capital structure,QuoteOriginally posted by: Gillkartsy:the base correlation is not flat it's upward sloping! I think that's why you got confused.... and regardless of the BC slope.