I am currently reading Michael Lewis’ newest book,” The Big Short”
BTW…i recommend it. On par with Liar’s Poker.
OK, here’s my point, apparently several of the Wall Street brokers, Goldman, Bear, LEH…the usual suspects…. were taking CDO’s repackaging the CDO’s, cramming additional subprime in the CDO and having the new and riskier product re-rated by the jokers at the rating agencies and actually having the rating improved, often under the guise of improved diversification.
They clearly do not look at diversification as I do.
When insufficient price data exists to capture meaningful relationship data, diversification can still be measured. You just need a new dataset. The dataset in this case is all of the bond characteristics.
How much of the bond has FICO scores less than XX?
What is the current default rate?
What are the sources of origionation?
What is the average duration?
What is the geographic breakdown?
What is the current prepayment rate?
etc, etc…..If you can name a meaningful characteristic, you can improve the measurement.
These characteristics are dimensions. Just like data reduction techniques used in time series data, some dimensions are more important than others. We have some mojo for doing the math on these cases.
The takeaway here, is that diversification can be measured on practically anything. Brainstorming a list of characteristics provides an alternative to the information contained in the time series. Contrasting the two techniques can also be a valuable exercise to test the completeness of your analysis.