JPMorgan has come up with a new way of calculating the correlation of single-tranche collateralized debt obligations that promises to simplify the process. Dealers agree that the current method is flawed, but are split on how to rectify it. Some favor creating a new correlation model while others are suggesting the market should use a different distribution of data.
The problem with the most-widely adopted model, known as the Gaussian copula model, is that the correlation of the first mezzanine tranche can vary significantly with small movements in spread. Another problem is that it gives too low a value to the senior tranches, explained dealers.
Lee McGinty, head of credit derivatives strategy at JPMorgan in London, said the new method, dubbed base correlation, treats all the credits in the tranche as equal in terms of spread and recovery rates and enters a spread for the whole index rather than each individual credit. It then calculates a correlation for the first loss to the 3, 6, 9, a 12% detatchment points.
The market norm is to price a single-tranche CDO by taking a correlation for a specific tranche, but McGinty argues that because of correlation skew it is more accurate to take specific correlations and then interpolate a value for a tranche. This borrows the thinking behind pricing equity call spreads, in which dealers price each call separately rather than take an average implied volatility.
Some credit strategists, however, argue that it is not the model, but the distribution of data that need to be changed. They favor the so-called student t-distribution because it gives more weight to extreme events.
