 |
 |
More on Interest Rate Model Selection
The “Conscientious Choice” paper celebrated its 5th anniversary earlier this year. The paper was reproduced by the Journal of Portfolio Management in 2004; the subject was revisited again last year and submitted for one of upcoming Frank Fabozzi’s publications. The main outcome–“Stay Normal!”–still seems very relevant (it is advisable for other purposes too). The normal model does not have to be a single-factor one though, and this is the current message.
Please re-read the February 2005 Pipeline article on the role and importance (or lack of it) of two-factor modeling. Provided that both the one-factor normal model (AKA the Hull-White model) and a normal multi-factor model are calibrated to the same set of vanilla options, the difference in value will only be measured for MBS that are convex to the curve’s twist. In option theory jargon, we say that instruments have a “curve option.”
The February 2005 paper showed that most MBS will be priced within a close range, no matter how many factors are employed. This paradox stems from the fact that AD&Co., as well as others, model prepayments as a sequence of European payoffs. Pass-throughs’ values are almost “nailed” by rates and volatilities. However, some CMOs present notable exceptions. Here is a short list of exceptions that investors and analysts should be concerned about:
A. CMO classes that are much shorter or much longer than their collateral carry the curve option
-Short fixed-rate CMOs: induced curve put (+)
-Long fixed-rate CMOs: upsized curve call (-)
B. Capped amortizing floaters carry the curve option (-)
C. Many IO and inverse IO tranches contain features (A) or (B) above
In the above abbreviation, (+) means that a two-factor model increases the value, (-) means it decreases
