same OAS system. The OAS levels will, of course, be different from usual OAS – they become prOAS. Our research has proven that valuation with a risk-neutral prepay model better explains MBS market dynamics and OAS variability, predicts Greeks and, hence, value changes.

AD&Co maintains a two-factor risk view with refinancing and turnover factors; our risk-neutral tuning recommendations are published weekly. Here is news for those who did not know: our risk-neutral calibrator is a part of a valuation system. Hence, users can tune our prepay model to the market at their leisure and frequency.

AD&Co Cutting-Edge Prepayment Model

To make a prepay forecast more transparent, we have extracted the “enhanced-features” part from the library; it can be directly fed from an external information system by passing loan sizes, LTVs, FICOs, geography, etc. We identified relevant Bloomberg fields and programmed VBA links that make the work less cumbersome. Each prepay model can be tuned separately, for risk-neutrality or as a corrective measure. Each position (pool) can be tuned on its own, following enhanced collateral information (automated).

Applications and Case Studies: Pay-ups, Old Loans, MSR

For most of these case cases, using the enhanced prepay model and data feeds is key. Information about specified collateral characteristics (such as low loan balance or “slow geography”) first gets converted into pool-level prepay tunings that are reported. Hence, the way our analytics crack this data is transparent, very informative, but still fully automatic.

• “Theoretical” pay-up is easy to compute as the same-OAS (or same-prOAS) difference between a fair value of a specified pool and the matching TBA quoted price. However, theoretical pay-ups usually overstate market levels; the market implies limited-period holding. We first introduced our “practical” pay-up methodology in the November 2004 article (http://www.ad-co.com/newsletter/Nov04/Valuation.htm). Since that article, we have compiled a practical pay-up matrix (pay-up in ticks for each feature) that resembles the market. We included these results in an upcoming MBS Handbook edited by F. Fabozzi.

 

>>>