A few messages since I've been away... Although a lot less controversial...We should separate discussions of Base Correlation vs. Large Pool Model. The point of the Large Pool Model is to create a transparent, distributable model that takes very few inputs to enable all market participants to have a reference point for correlations. It is (obviously) not the most accurate model in the world, but I don't see a multi input model becoming a standard - one of the reasons is the next point.As to the input "index spread", this is the traded index level. The standardised portfolios trade where they trade, sometimes away from the average level of spreads. Given that the typical hedge for (for example) a DJ iTraxx tranche would be DJ iTraxx, there are all sorts of interesting questions about what you should use... (for instance there are lots of questions about what you do with a model that takes 125 single name spreads, when your hedge is trading 10 tighter than the theoretical level of these.)In terms of single solutions. The point more relevant is that there is never more than one solution. An analogy might be to consider the implied vol on a 50 strike call option priced at 5, when the underlying market is at 60. Base Correlations tend to have no solutions when the total losses approach or pass the total losses on the whole portfolio in the same way. (By the way, I know this isn't quite correct, that is what I mean by an analogy). I think now that we have full merger(s) of the underlying indices globally, we are likely to get to a point where there are always solutions. Regards[edited silly typo - which meant my point was just wrong! And spelling mistake]Lee

Hi, I am curious about the value of base correlation given that people in industry may not use a single-correlation number for pricing/hedging. In my view, base correlation is a better way to present correlation - in the limited one-factor world. As the JP article shows, while base correlation can be useful for comparing liquid tranches like trac-x, it does not extend to bespoke tranches. It doesn't say much about hedging either. Lee, did u mention that you are working on extending the base correlation framework to bespoke tranches? what's your view on that?thx

Hi everybody,I am interested in the Hedging question too.I would like to know if you use the methodology explained by Bear Stearns in their paper distributed in Madrid. They compute the hedge for a Mezzanine tranche as the difference between the hedges of the Equity "Base" tranches. The 3-7% Hedge is diveded by 2 compared to the "implied correlation" HedgeAnd what about the correlation reserve ? (This question has already been mentioned in this thread by CreditGuy)Thanks!------------FBA

FBA,in my view, hedges based on compound correlation are not very useful. I think the alternatives are hedges based on base correlation, i.e. computing hedges based on first loss tranches, or using one correlation structure for each portfolio (possibly term dependent).Obviously, your correlation structure for bespoke portfolios would differ to that for the DJ CDX portfolio. The question now is, how to estimate the correlation structure for a bespoke portfolio based on market information for DJ CDX. The ATM correlation concept might be usedful here. Anyone got some good ides?

How do I flip this around to look at recovery correlations for distressed loan portfolios? Not enough historical data and what is available would be a poor proxy anyway. Someone help me, my brain too small and low-wattage for all of this.....help!

like I was saying, you need a pro <in a 'zen' style>when you do not have experience to guide youyou must begin by studyingthe question. Recoveries occurin a foreign land of mysterious marketsnot commonly observed. Yet the inhabitants of this land seekthe same hopes and dreamsas you and I.So, your questionmay be answered.</zen>

A-ha, wondering when you'd zing me - I'll PM asap.Now that kr is able but unwilling to be helpful , any other bright sparks out there willing help poor non-quant me...

Hi all,does anyone have the Bear Stearns paper that has been distributed in Madrid?Thanks,SurferD

Hi!i m trying to replicate the iTraxx quotes with a jpmorgan type model.From the base correlations i find upfront premium/on going premiums which are sort of in between bid and ask.But my i cant replicate the deltas.Anyone has had the same problem/know what formula are being used for computing delta?Thanks!

Hi Lee,I have a simple question: the Large Pool Model (i.e. the spreadsheet you kindly published some time ago) does not depend on the number of names in the pool - simply assuming it's large enough, right?Thanks for the help,SurferD

I can't get the JPM Base Corr model to even come close to an iTraxx tranche quote/corrs. Can anybody enlighten my feeble mind?example quote for 5Y index @ 43 midbid offer bidCorr offerCorr27.5 29.5 23.5 20.5168 176 4.3 570 78 15.5 16.541 46 23 2519 23 30 33

Wibble,I had the same problem. I got a copy of the model from the forum, played with it a bit and here are some questions/issues:1) The level of Base Corr derived from JPM Xl model varies significantly from some quotes for ITRAXX (I get ITRAXX and IBOXX info from BS, but only ITRAXX comes with BASE CORR); I'd love to get Base Corr quotes from other dealers and for IBOXX (CDA.NA.IG) as well. E.g. today's Lower Upper Upfront Spread Base Corr0 0.03 0.2625 0.05 20.40.03 0.06 0.016 30.30.06 0.09 0.0062 37.10.09 0.12 0 0.0039 41.70.12 0.22 0 0.0018 54.10.03 1 0 0.0013 Index @42What I get from JPM spreadsheet is :27.83%37.79%45.39%51.38%66.96%I don't expect different dealers to use the same, but I thought they'd be much closer.2) Most senior tranche in the JPM spreadsheet has pretty much 30-100 hard-wired into the Expected Loss calculation (cell Q25 I believe). It subtracts the "ZerotoSub2UpperLoss" instead of looking at the Senior Lower Attachment. However, even with the change the model recalculates the Fair Spread for 3-100 and replaces the quoted 13 bps with 16 bps. The reason being, I believe, in the Porfolio Loss for Senior tranche (AZ column) it takes the SeniorCorr into account, while:a) Senior Corr is not defined b) it is not calculatedc) hence, zero is inputedd) theoretically should be 1 by constructionThis came up when I was trying to see that all the spreads sum up to the index.Lee, I would be curious if somebody already noticed and/or adressed these issues. Any comments?Thanks!

krot,why should the sum of spreads sum up to the index?Just look at the senor tranche and suppose its expected survival is zero (*).Then its corresponding spread is infinity and you therefore can t have sum of spreads = index spread.(*) This happens when tranchel oss is equal to tranche size.

DiceMan,I'm not sure I get the zizzi part...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.

QuoteOriginally posted by: krotThe (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.won't work.consider 5-6 defaults=>equity wiped outyou lose the ultra huge spread on equity and only gain 0.05 on the index pos.