Hi,Base correlation can be used to value bespoke tranches using a given methodology e.g. expected loss scaling, probability matching etc.I have read that adjustments for spread dispersion of the bespoke may also be applied in its valuation.I am thinking that perhaps this adjustment is of the form:*If bespoke spread dispersion < index spread dispersion (whereby the index is used to build the base curve which is an input of the given bespoke valuation methology)*Then deault are more clustered for the bespoke --> a higher correlation for the bespoke and a scaling up of the base correlation used within the given bespoke valuation methology.But how and why is the bespokes spread dispersion relevant, if I have not already answered the question? Any papers on the subject would be appreciated!Also I have heard the larger the number of names in the pool the less the impact of correlation. If this is the case are there adjustment for pool size too?

Dispersion has to do with allocation of value across the capital structure. In a portfolio with low dispersion, all spreads are at about equal level and thus are expected to default with roughly the same probability. Such a portfolio, when compared to a more dispersed one, would have less value in the equity and junior mezz tranches and more value higher up in the capital structure. The impact of this would be a flatter base correlation skew.CDX4 and 5 are a classical example. After the autos blowup in series 4, the portfolio dispersion shot up with the cost of equity protection increasing massively. With the value shift into equities, the base correlation skew steepened and tranche correlations couldn't even be computed for the mezz tranches (you had to skew the attach/detach corrs to match the market values for the tranche regardless of the attach corr). When the index rolled into series 5, the auto's were removed. The value of series 5 equity was lower than series 4, and the base corr curve was a good deal flatter. As far as adjusting base corrs for the skew - there is no precise way of doing this that I know of. Since base correlations are a fudge anyway, I would suggest thinking of impact of skew on the loss distribution.

Good answer man. Also, when dealing with bespokes that have high dispersion [ie. residuals far from mean spread] you have to adjust individual deltas for these credits. In a CDO port with relatively static spreads [low dispersion] you will have a static delta across the port. But, with dispersion, the seller of the CDO/tranche will want to buy more idiosyncratic delta on wider names to hedge default risk.