Pricing Mortgage-Backed Securities in a Multifactor Interest Rate Environment : A Multivariate Density Estimation Approach

There have been recent, well-documented cases of financial institutions and investment groups incurring huge monetary losses on their mortgage-backed security (MBS) portfolios. This vulnerability is partly due to the complexity of MBS pricing. Homeowners have the option to prepay their mortgages and refinance the property at any time; hence, MBS investors are implicitly buying a fixed-rate bond and writing a call option on this bond. Furthermore, prepayment of mortgages (and thus the timing and magnitude of the MBS's cash flows) can occur not only when rates fall, but also for reasons not related to the interest rate option. This paper develops a nonparametric, model-free approach to the pricing of MBSs, using multivariate density estimation (MDE). Our goal is to investigate the relation between MBS prices and interest rates. While the usual methods for valuing MBSs are highly dependent on specific assumptions about interest rates and prepayments, this method yields informative results without requiring such assumptions. The MDE estimation suggests that weekly MBS prices from January 1987 to May 1994 can be well described as a function of the level and slope of the term structure. We analyze how this function varies across MBSs with different coupons and investigate the sensitivity of prices to the two factors