Verification improves predictive performance
VantageScore Solutions released the first of two new white papers that demonstrate superior predictive performance when using one generic credit score to verify an initial risk assessment made by another credit score. “The Predictive Value of Credit Score Consistency” white paper also demonstrates that verifying VantageScore 3.0 credit scores at two separate nationwide credit reporting companies (CRCs) yields a higher percentage (75 percent) of score alignment than using two proprietary credit scoring models, because the VantageScore model yields more consistent credit scores.
The white paper examines the practice of verification — the alignment of risk assessments from two credit scoring models — to ensure consistency and accuracy in the assessment of a population, which ultimately can result in minimizing the default rates for lenders. In this study, VantageScore data scientists sought to answer two questions when validating the need for verification:
- Was there predictive performance improvement (i.e., Gini improvement) when verifying any credit score with another, different brand of credit score from a separate CRC?
- How often do credit scores “verify” risk assessment?
Through testing, the study found that verified credit scores (regardless of the brand or version) deliver Gini improvements of at least two points higher on the verified population than if only one credit score is used on that same population. Conversely, the population with credit scores that are not verified by the other credit score (i.e., those with conflicting verification evaluations) shows a drop in Gini performance of four to five points, meaning risk assessment for these consumers is less reliable. Thus, consumers with unverified credit scores should be managed with greater scrutiny than consumers with verified credit scores.
Additionally, the study uncovered that when two VantageScore 3.0 scores (from two independent CRCs) were used, the credit scores of 75 percent of the consumers in the population were verified. However, when the verification was conducted using two proprietary CRC credit scores from two independent CRCs, just 30 percent of the consumers’ credit scores were verified. The improved verification performance using VantageScore 3.0 is a function of the consistent algorithm used across the three CRCs, which is a patented signature feature of all VantageScore models.
A second, related white paper that examines the upside of a verification process for those consumers with limited credit histories will soon be available.
While VantageScore 3.0 was used for this study, results would be similar if other versions of VantageScore were used, including VantageScore 4.0.
For more details on “The Predictive Value of Credit Score Consistency” white paper, visit: www.vantagescore.com/consistencyWP