The Credit-Scoring Model Used for Mortgage Lending is Outdated

By: Laurie Goodman
Date: August 10, 2017

The credit scoring model currently used by mortgage lenders has not been updated in nearly 20 years! Over the years, more advanced credit scoring models (including the birth of the VantageScore credit score) have come to the market since FICO 4, which was developed in the late 1990s. If the GSEs and mortgage lenders were to integrate more recent models, such as VantageScore 3.0, there are various benefits for both lenders and consumers. The benefits include: competition for a fairer market, transparency, ability to serve creditworthy borrowers seeking a mortgage more fully, and more.

VantageScore Solutions is also rolling out VantageScore 4.0 this fall, the first tri-bureau credit scoring model to leverage machine learning, trended credit data, and align with NCAP.

In a recent article, Urban Institute cataloged what the difference is between the outdated models and the newer models.Below is an excerpt and link to the full article:

“• More granular data: In the late 1990s, credit bureau data was much less granular than it is now. For example, student loan debt was included in installment debt. There was no differentiation between first and second mortgages. The use of more granular data in the newer models has allowed for improved credit modeling.

Better information on student loans: There is much better information in the newer models on the performance of student loans, and how this performance impacts the performance of other types of debt. In the late 1990s, there was less than $100 billion of student loan debt outstanding. Federal Reserve Bank of New York data indicates that in Q1, 2003, there was $240 billion in student loan debt outstanding; as of Q1, 2017, that number was $1.34 trillion, a 458 percent increase. By comparison, total consumer debt has increased from $7.23 trillion to $12.75 trillion in this same period, a 76 percent increase.

More consumer-friendly treatment of collections data: Paid collections were included in the FICO 4 family of models but are ignored in more recent FICO and VantageScore models in recognition of the limited probative value of paid collections. In addition, in the early FICO models there was no differentiation between unpaid medical collections and unpaid non-medical collections. In more recent models, unpaid medical collections are weighted less heavily, as they 2 are found to be less predictive of future performance. These are features that affect many borrowers; Urban research has shown that 35 percent of those with a credit score have debt in collection. Research by the Consumer Financial Protection Bureau has shown the difficulties of relying on medical collections data as an indication of creditworthiness.

More consistent, robust information: FICO 4 is actually three different models, one for each credit bureau, estimated in slightly different time periods between 1995-2000. (The three models are TransUnion FICO Classic 4, Equifax Beacon 5.0, and Experian/Fair Isaac Risk Model v2.) Lenders are instructed to obtain a credit score for each borrower from each of the three credit bureaus. The middle of the three scores is used for lending/pricing purposes. More recent FICO and VantageScore models use identical time periods for the estimation, and more closely align the models. In the case of VantageScore, the models are completely aligned.”

Click here to read the full article from urban.org.