Value Commentary

The Andrew Davidson & Co., Inc. Implied Prepayment Model is a method of solving for prepayment model tuning factors that compensate for the risk premium demanded by investors in discount and premium MBS. It does so by attempting to reduce the variability of OAS across coupons. Generally, this will result in a model that exaggerates the fears of the market, namely that refinancing will be faster, turnover slower, and burnout less powerful than expected. While the tuned model is not useful for predicting prepayments along any particular rate path, it will, in theory, give an analyst durations that more closely match observed durations by taking into account the risk premiums.

In this article, we test the theory that the implied durations will more closely match observed market durations. We do so by computing the tuned and untuned durations for several GNMA TBAs on both July 3 and July 18, taking the average, and comparing them to the actual sensitivities observed over this period. The untuned durations were computed using the ADCo OAS system with all tuning parameters set to 1.0 except for burnout, which was set to 1.1 as recommended in the February Pipeline. For the tuned durations, we using the ADCo OAS system with the tuning parameters listed in Table 1. The empirical sensitivities were computed by downloading from Bloomberg the daily closes for TBA prices and mortgage current coupon yields and regressing the former versus the latter. The beta from the regression equation was then divided by the average TBA price for the period then multiplied by -100. The results of the analysis are shown in Table 2 on the next page.

It seems that the implied tuning durations are closer to the market observations than the untuned durations. The tuned durations on the low coupons are higher than the untuned durations, and thus closer to the market observations. The reverse is true for the high coupons - the tuned durations are lower than the untuned and closer to the market. This is because the implied prepayment model is pessimistic. By exaggerating the markets fears that turnover will be slower, refinancing faster, and burnout less powerful than the model predicts, the implied model lengthens the durations on low coupons and shortens them on high coupons.

Andrew Davidson & Co., Inc. is currently working on a valuation method that explicitly takes into account the price of risk without the need for an implied prepayment model. This model is more soundly grounded in theory, and we predict it will better predict interest rate sensitivity, but, evidently, the implied model can does a satisfactory job in the interim. >>>