Providing Greater Certainty Through Enhanced Risk Management

As we noted in last year's perspectives blog on the 2023 Desktop Underwriter® (DU®) V. 11.1 update, Fannie Mae is continually researching new methods and data to assess mortgage credit risk safely and soundly, while ensuring that potential borrowers have access to an affordable, safe mortgage. The recent announcement of DU Version 12.0 reflects this ongoing focus.

With DU V. 12.0, we have made changes to the DU Risk Assessment risk factors, enhanced our assessment of borrowers with limited or no credit, increased the number of borrowers who may be eligible for a Cashflow assessment, and made technical updates that may allow for more frequent adjustments to the DU Risk Assessment in response to changing market conditions. We have also expanded the loan performance and trended credit data in our model development population.

With these updates, our ability to rank order the riskiness of loan applications has improved by 14% and our accuracy metric improved by 4.5x – a remarkable improvement considering DU's substantial history as an industry standard-setting automated underwriting system. This improvement in model accuracy gives us confidence that we and our credit partners can rely on the DU Risk Assessment through all market cycles. Stakeholders may observe that some loans will now receive an Approve/Eligible recommendation that would have previously received an Approve/Ineligible or Refer with Caution recommendation, and vice versa, when compared to the current version of DU.

Risk Factor Updates and Enhancements

With each DU release, we look for opportunities to leverage new sources of data to keep up with changing borrower profiles and better serve current borrowers, while removing or modifying factors that are no longer as predictive. In DU V 12.0:

• When evaluating the debt composition for the borrower, we will no longer consider the composition of revolving debts within the borrower(s)' total monthly expense, nor will we evaluate a borrower(s)' variable income, since these factors were not impactful enough to include once all other risk factors were considered.
• We're also adjusting how we evaluate debt composition for student loan debt, as our research showed that borrowers with student loan debt performed better than borrowers without such debt, assuming similar total debt levels and other risk factors were held constant. For context, 18% of recent applications had borrowers with some amount of student loan debt.
• We are now considering first time homebuyer status as a risk mitigating factor, as our research showed that first-time homebuyers performed better than non-first-time homebuyers when all other risk factors were held constant. 25% of recent applications were to first-time homebuyers.
• We have expanded our use of rent payment history data to use both bank statement data from 12-month asset verification reports, which we implemented in the DU V. 11.0 update in September 2021, and rent reported on a credit report, which Fannie Mae began encouraging multifamily property owners to report to credit bureaus in 2022. We believe this will more than double the availability to borrowers compared to bank statement data alone.
• We are also expanding our Cashflow assessment capability to apply to all borrowers, rather than only borrowers who do not have a credit score. This has the potential to impact about 7% of recent applications, but could eventually impact substantially more borrowers, based on increased availability of 12-month asset report data as lenders increase use of the 12-month asset verification report to enable validation of assets, income, and employment.

As part of these improvements, we will retire the separate model developed to support borrowers without credit scores, as DU V. 12.0 will be able to support borrowers with and without credit scores.

DU V. 12.0 represents Fannie Mae's ongoing commitment to managing mortgage credit risk and serving as a reliable source of mortgage financing for the U.S. housing system. Read more in our whitepaper here.