Freddie Mac 2006 Annual Report Download - page 126

cash Öow approach with market input assumptions extracted from the dealer quotes provided on the more liquid products,
reduced by an estimated liquidity discount.
For 2004, we calculated the Guarantee asset fair value using an expected cash Öow approach. SpeciÑcally, Monte Carlo
projections were used to forecast Guarantee asset-related future cash Öows. The forecasted cash Öows were then discounted
using factors that were derived from modeled forward interest rates for each scenario path, to which we then applied a
trailing average option-adjusted spread of up to 24 months that was based on trust IO security prices.
Recognized Guarantee obligation
Our approach for estimating the fair value of the Guarantee obligation makes use of third-party market data as
practicable. We divided the credit aspects of our Guarantee obligation portfolio into three primary components: performing
loans, non-performing loans and manufactured housing. For each component, we developed a speciÑc valuation approach
for capturing its unique characteristics.
For performing loans, we use capital markets information and rating agency models to estimate subordination levels and
dealer price quotes on proxy securities with collateral characteristics matched to our portfolio to value the expected credit
losses and the risk premium for unexpected losses related to our guarantee portfolio. We segmented the portfolio into
distinct loan cohorts to diÅerentiate between product types, coupon rate, seasoning, and interests retained by us versus those
held by third parties.
Because typical structured securitizations of single-family collateral only include performing loans, we utilize a separate
method for estimating the fair value of the Guarantee obligation for non-performing loans. For loans that are extremely
delinquent and have been purchased out of pools, we obtained dealer indications that reÖect their non-performing status. To
value delinquent loans remaining in PCs, we started with the market driven performing loan and non-performing whole loan
values and used empirically observed delinquency transition rates to interpolate the appropriate values in each phase of
delinquency (i.e., 30 days, 60 days, 90 days).
We evaluated market sources to determine the appropriate credit costs associated with the Guarantee obligation for the
manufactured housing portfolio, approximately 1 percent of our total guarantee portfolio and 19 percent of the fair value of
the Guarantee obligation, and determined that there is not suÇciently reliable market data. As a result, we used our
judgment to develop an alternative approach for estimating the incremental credit costs associated with the manufactured
housing portfolio. SpeciÑcally, we calculated the ratio of realized credit losses for performing loans and manufactured
housing loans to determine a loss history ratio. We then applied the loss history ratio to market implied performing loan
Guarantee obligation fair value estimates to calculate the implied credit costs for the manufactured housing portfolio. This
approach grounded the Guarantee obligation related to manufactured housing in performing loan market prices, while
adjusting for the loss history reÖected in empirical data. We undertook a similar process for estimating the fair value of
seriously delinquent manufactured housing loans.
The components of the Guarantee obligation associated with administering the collection and distribution of payments
on the mortgage loans underlying a PC are estimated based upon amounts we believe other market participants would
charge. Finally, we use our models to estimate the present value of net cash Öows related to security program cycles. This
estimate is included in the Guarantee obligation valuation.
For 2004, the Guarantee obligation fair value was calculated using internal models to estimate future cash Öows using a
Monte Carlo simulation. The components of estimated cash Öows associated with the Guarantee obligation included
estimates of expected future credit losses using statistically based models that were benchmarked periodically to the non-
conforming loan, or jumbo, securitization market. For all periods our estimates included costs to administer the collection
and distribution of payments on the mortgage loans underlying a PC and considered net cash Öows due to security program
cycles.
Recognized PC residuals
The fair value of recognized PC residuals is determined in a manner that is consistent with the approach described
above for the recognized Guarantee asset and Guarantee obligation.
Key assumptions used in the valuation of the Guarantee asset
Table 2.2 summarizes the key assumptions associated with the fair value measurements of the recognized Guarantee
asset. The assumptions included in this table for 2004 relate to those used in our internal models. For 2006 and 2005, the fair
values at the time of securitization and the subsequent fair value measurements were estimated using third party
information. However, the assumptions included in this table for those years are those implied by our fair value estimates,
with the Internal Rates of Return, or IRRs, adjusted where necessary to align our internal models with estimated fair values
determined using third party information. Prepayment rates are presented as implied by our internal models which are
benchmarked periodically to market prepayment estimates.
114 Freddie Mac