I recently had the opportunity to participate in a panel discussion: “What is happening with underwriting and credit standards” during The Future of Housing Finance Conference at George Washington University. One of the topics covered, which garnered a great deal of interest from the audience, was how a more predictive score will allow lenders to safely qualify more consumers for a mortgage.
More specifically, the interest was around the analytic advancements within FICO® Score 9. As we’ve shared previously, new features in the score include a multi-faceted modeling approach and a more refined treatment of third-party collections (differentiating between medical and non-medical collections, and ignoring paid collections), which significantly enhance mortgage origination prediction.
We recently conducted a research study to illustrate the benefit of using FICO® Score 9 for mortgage originations. We used a swap set analysis that captures both the change in consumer distribution across the FICO® Score range and the improved predictive value over the prior version of the score in use today by the industry (FICO® Score 5). The analysis is based upon mortgage loans booked between November 2011-April 2012 and subsequent performance on those loans through October 2013. For the analysis, we used the standard “bad” definition of 90+ days past due.
The chart below can be interpreted as follows: The first column represents the FICO® Score cutoff, and the next two columns – under “FICO® Score 5” – represent the percentage of mortgages booked that were above or below the specified score cutoff. Columns 4 and 5 under “Swap Set” show the percentage of booked accounts that would have “swapped” above or “swapped” below using FICO® Score 9.
This information shows where the scores disagreed, however it doesn’t tell you which version of the FICO® Score was more accurate in predicting risk. That is what the last two columns reveal: under “Swap Set Odds,” we see the performance odds (good to bad ratio) of the actual booked mortgage loans. Across the score spectrum, we see consumers shifting to higher score ranges that have higher odds of repayment and those shifting to lower scores that have lower odds of performance, demonstrating that FICO® Score 9 further refines classifying risk prediction.
Using a FICO® Score 5 cutoff of a 680 as an example, 87.5% of the mortgages booked had a FICO® Score 5 of 680 or higher, and 12.5% had a score below the 680 cutoff. Analyzing the swap set, we see a net increase of 1.5% of the population that would have moved above a FICO® Score of 680 as a result of switching to FICO® Score 9 – that is, the difference between the 3.9% of the booked population that swapped above and the 2.4% of the population swapped below to a lower score range, resulting in more consumers being approved for credit.
In fact, using the same FICO® Score cutoff of 680, 186,000 additional consumers could have been approved for a mortgage during the same time period.
How did we get that number? We quantified the size of both the applicant pool and booked loans during the November 2011 to April 2012 time period:
- Applicant pool – More than 190 million consumers inquired for credit during this time period.
- Booked loans – Of those inquiries for credit, we identified a smaller subset of 70 million inquiries from the real estate, bank and finance industries. From the credit data alone, one can’t precisely identify a mortgage inquiry, so we conservatively estimated that 24.8 million were for mortgage credit. This estimate uses an industry average assumption of at least two applications for every one closed loan during a time period when there were 12.4 million booked loans.
The resulting calculation is then 1.5% of this 12.4 million, which equates to 186,000 additional booked loans.
This study illustrates that not only will all participants in the industry – from consumers to lenders to investors – benefit from adopting the more predictive FICO® Score 9, but the housing market and recovery will benefit as well.
