|
Decision
system construction, validation and regulatory review. |
|
To
contact us: |
|
Management Consulting Services |
|
Mortgage Decision Technologies, Inc. |
|
Models developed from Subprime mortgages
typically do a good job of rank ordering risk on Prime loans and models
developed from Prime loans perform
fairly well on Subprime mortgages.
Subprime Models are more relevant in that the pricing decision cutoff
occurs so far toward the
subprime portion of the quality
spectrum. A model that
differentiates quite well between the (lowest risk) 5th and 25th percentile
but whose performance
deteriorates in the high risk 80th to 90th percentile is interesting but of
limited practical use in pricing
mortgage loans. Subprime models have
lower “KS” values than their prime counterparts, however performance in the bottom 20% of the
quality distribution is crucial. The
subprime model is often the more
effective tool. Traditional subprime grade assignment
of “A-” through “D” for pricing
purposes provides adequate return on capital for the investor. Traditional grades, however, provide only a general ranking of risk. Empirical
systems classify some loans from traditional “D” risk grade as “A-”
and vice versa. Empirical assignment
concentrates loss in the “C and D” categories and makes the “A-” category
lower risk. Lower risk requires
less severe pricing and may account for lower coupons in the
GSE subprime programs. Empirical classification is the lenders
best defense against allegations of predatory or discriminatory lending. |
