VantageScore Solutions is releasing a new credit scoring
model which it says will enable lenders and others using the system to pull
scores on 27 to 30 million previously unscorable consumers. The company claims its model, called
VantageScore 3.0 provides up to 25 percent better predictiveness and uses the
consumer-familiar scoring range of 300 to 850.
The model was built using a sample of 45 million anonymous credit files.
Testing of
its new model shows that it provides improves predictive strength across all
industries and applications but in particular within the key prime and
near-prime consumer populations, the most sought after group among mainstream
lenders.
The new
model is said to be able to score more individuals than traditional scoring
models through introduction of a thirteenth scorecard to generate a predictive
score where an individual has little or no recent credit activity. The model factors in non-tradeline credit
data such as collections, public records, and inquiries when active tradeline
data is not available and utilizes tradeline data that is aged more than 24
months but remains predictive. The model
also uses rent, utility, and telecom data when it is available.
The improved predictiveness, which is present
within the same population for originations in the real estate, auto, and
bankcard segments, is demonstrated by nearly identical risk alignment across
all three credit reporting companies (CRC); maximum predictive performance for
both originations and account management, and provides consumer scores which
are within 20 points across all three CRCs for 80 percent of accounts. The
latter feature will allow lenders
to
have confidence
in default probabilities
regardless of which credit reporting company pulls the score.
The company said the key reasons
for the improved performance is using more granular data from all three CRCs
which allowed the model's designers to select 150 of the most predictive
characteristics from among around 900 that were tested. This allows the user to separate first mortgage
from
other mortgage related transactions, facilitating greater
intelligence with regard to a borrower's mortgage-related debt. It also allows more distinct definitions of data, such as the ability to identify student
loan accounts from other
types of installment accounts
and gives more specific measurement
of
delinquency and default
timeframes.
The data sample used to build the
VantageScore 3.0 model
was developed on consumer behavior from two different two-year timeframes: 2009-2011 (during the economic crisis) and 2010-2012.
Each performance timeframe contributed 50 percent of
the
model's development. This reduces the model's sensitivity to consumer behavioral shifts over different
volatile periods. The model also sets negative information to
neutral if it is coded as occurring during a natural disaster but allows
positive information to retain its positive impact.
As part of
the product launch VantageScore Solutions has introduce a consumer oriented
website, www.ReasonCode.org. The site provides a primer on what reason codes are
and
how they are used,
searchable and interactive reason code definitions
and
explanations,
and a glossary of common reason code
terms