What is the Implied Prepayment Model?
In this premiere issue, we will explain what the Implied
Prepayment Model (IPPM) is, what we are trying to achieve
by running it, and where to find the results on our website.
For future installments in this section, we will comment on
relative value in the MBS Market by looking at effective duration,
option-adjusted spreads and other measures of relative value.
When we run the IPPM, we are trying to find market-implied
values for four of the tuning factors: scale, burnout, turnover,
and refi. To do this, we calculate values for these tuning
parameters that generate valuations reflected in the prices
of actively traded TBAs. We run the IPPM separately for our
Ginnie Mae, Fannie Mae, and Freddie Mac 30-year MBS Prepayment
Models to determine distinct tuning solutions for each of
these models. We post all the results on our website, at www.ad-co.com.
Methodology
For each of the three model types, we have a basket of MBS
that we use as input to the IPPM. The basket consists of MBS
with coupons ranging from 5.5 to 8.0 in 50 basis point increments.
For each MBS, we poll Bloomberg for current market price and
use these prices as inputs to our OAS model. We first run
the model untuned (with each tuning factor set to the default
of 1.0). This gives us the "Untuned OAS" curve shown
on our website. (The OAS curve is a graph with coupon on the
x-axis and OAS on the y-axis.) The IPPM, itself, is a non-linear
optimization technique that attempts to flatten the OAS curve,
that is minimize the variance of OAS across coupons, by changing
the tuning factors. We expect the OAS to be equal across all
liquid coupons for a given model type, as there is no difference
in credit quality, and OAS takes into account the different
option costs of different securities.
Results
We have found that, in most cases, we can reduce the variance
on Fannie and Freddie OASs by 70% or more and on Ginnies by
90% or more. Check out the Implied Prepayment Model section
of our website today to see tuned and untuned OAS curves,
as well as the resulting tuning solutions.
