|Stochastic Systems Group|
Dr. Bill Irving
A central challenge in the valuation of mortgage-backed securities (MBSs) is to characterize prepayment behavior. Prepayments arise because the loans collateralizing an MBS generally give the borrowers the right to prepay some or all the principal at any time, usually without penalty. Wanted or not, these unscheduled principal payments are passed back to the MBS holders, which means the holders are short a complex American-style call option. To value this option -- and by extension the MBS itself -- it is necessary to project future prepayments under a whole spectrum of possible interest-rate scenarios. Given that the MBS market is a multi-trillion dollar market, the stakes in this modeling enterprise are huge. In response, Wall Street dealers and a few buy-side firms have expended a great deal of resources over the past seventeen years to build increasingly sophisticated prepayment models. These models quantitatively characterize the individual homeowner's decision process, taking into account the effects of interest rates and the robustness of the housing market on the dollar value of the incentive to refinance or relocate. In turn, prepayment models are embedded in bigger systems for valuation and control of interest-rate risk. This talk provides a glimpse into the fascinating field of MBS valuation, with a particular emphasis on prepayment modeling. This field continues to be actively researched as it uniquely brings together elements of finance, economics, statistics, and signal processing.
Bill Irving is a quantitative analyst in the Fixed-Income division of Fidelity Investments. His research focus is on valuation of mortgage-backed securities, and in particular on prepayment modeling. He is an alumnus of the Stochastic Systems Group, having done his doctoral work on multiscale stochastic realization under the direction of Alan Willsky. After graduating, he worked from 1995-1999 for Alphatech where his research interests included SAR ATR, multisensor fusion, and license-plate character recognition. Since 1999, he has been with Fidelity.
Problems with this site should be emailed to firstname.lastname@example.org