Prediction of residential mortgage contract rates

In this thesis, we have examined the predictability of the effective yield of residential mortgage loan rates as a function of economic variables which would be available to a loan officer preparing to set a loan rate. Of the possible explanatory regressor variables, the 10-year interest rate is most highly correlated with effective yield. Since this is a case where the sample sizes (number of loans made per month) vary widely, the major new statistical idea introduced in this thesis is that of the weighted autoregressive first order model [WAR(1)]. The procedure used in this thesis to obtain the [WAR(1)] estimates is based on iteratively reweighted least squares methods.