Mortgage backed securities, duration and implied prepayments

The primary purposes of this study are to provide an accurate measure of risk for mortgage backed securities and to infer the market's expectations regarding future prepayment rates from actual mortgage backed security price data. The first major objective is to accurately estimate the implied duration of a mortgage backed security, using an elasticity concept. The results of this study clearly show that the proposed multiple regression model does a significantly better job at estimating implied duration than the proposed simple regression model. The multiple regression model provides an extremely strong relationship between Macaulay's duration and total return. This finding should be of interest to investors, since it provides a means of measuring the risk of mortgage backed securities.

The second major objective is to use implied duration to estimate prepayments. Using an estimate of implied yield, an iterative process is employed to find the prepayment rate that provides the implied duration. With this procedure, prepayment rates are implied by market prices. While the results of the study indicate the multiple regression model generally works better than the simple regression model, neither model seems to work well in predicting prepayments. Implied prepayment is found to be a biased predictor of actual prepayment, indicating both models overestimate the prepayment rate.

Although the multiple regression model approach is effective at estimating the risk of a mortgage backed security, it should not be used to estimate constant prepayments associated with the mortgage pool. None of the models do well in predicting prepayments, indicating that the constant prepayment rate assumption is, in all likelihood, incorrect. During times of volatile interest rate movements, which is the time of this study, mortgage prepayments probably follow some nonconstant prepayment pattern.