Making the most of score differences

I previously blogged about how a consumer’s FICO® Score may vary from one CRA to another, due mostly to differences in data reported to each of the credit reporting agencies. This begs the question: when pulling scores from more than one CRA, which should be used for credit decisioning?

The answer largely depends on a lender’s internal strategy, operational constraints, compliance goals and risk tolerance. A lender focused on portfolio growth may choose to accept an applicant based on the lowest of the three FICO® Scores. A lender focused primarily on risk mitigation may take a more conservative approach and accept an applicant based on the highest FICO® Score. In the mortgage industry, it's common to use the middle score for individual applicants and the lower of the two middle scores for joint applicants.

Multiple case studies indicate that averaging scores provides better results than simply using the one of the scores. Indeed, we've seen greater predictive value in using the average—modest gains to KS were observed.

Averaging scores ensures that information from each score is being factored in, particularly in cases where the difference between the scores is large. In other words, more information helps the score be more predictive—something FICO has espoused for a long time. Using the average is one way to leverage three different, albeit highly correlated, pieces of information. (Note: while this approach may be more predictive, it's important to run any new strategies by your compliance and operations teams since there could be adverse action notification and dispute handling considerations.)

This stronger performance is intuitive given that each score is designed to capture as much predictive lift as possible within that CRA’s available data. So to the extent that there are differences in the data, combining scores could provide more lift than any individual score.