ABSTRACT

The research objective is to develop a multifactor endogenous mortgage rate model. To obtain accurate predictions of the mortgage rate, a model needs to consider many state factors such as multiple factor interest rates, real estate prices, unemployment rates, etc. The conventional approach to solve such a multifactor model suffers from the curse of dimensionality and is, therefore, computationally intractable. This research will develop a computationally tractable algorithm by a fundamental shift in the formulation of the problem: instead of computing the mortgage rate that corresponds to a multidimensional state factor, the algorithm computes the set of state factors that correspond to a given mortgage rate. This approach eliminates the curse of dimensionality: the complexity of the algorithm is independent of the number of factors used to model the mortgage rate. The research will develop regression based Monte Carlo methods and a new randomized quasi-Monte Carlo sequence to implement the algorithm.

The accurate modeling of the mortgage rates, and thus accurate modeling of the prepayment behavior, improves the stability of the financial markets, reduces risk for mortgage investors, and, consequently, lowers mortgage rates for homeowners. At present, the common approach to model mortgage rates by the financial industry is based on a very simple heuristic: the 10-year Treasury yield plus an exogenous constant. Recent empirical evidence suggests that this heuristic is seriously flawed in certain economic situations. This research will develop a computationally efficient algorithm to solve an accurate mortgage rate model.

Please report errors in award information by writing to: awardsearch@nsf.gov.