Opinion

With competing credit score models, the details matter

On March 1st, the Federal Housing Finance Agency held a public listening session on credit score models. I appreciated the opportunity to share my thoughts along with more than 25 other respected and knowledgeable stakeholders.

The importance — and challenges — of credit scores are very much front-and-center in the current dialogue as we collectively strive for a more efficient, effective, equitable, and inclusive mortgage finance system.

Those who balance sheet loans can use whatever credit score model they prefer — or none if they so choose. However, as the majority of loans exit through the secondary market, most lenders must use a credit approach that meets the requirements of secondary market participants, such as investors or guarantors, within the applicable mortgage channel. Where a credit score is employed, the related score and model must fit within the participants’ respective pricing, risk, and other applicable methodologies.

One of the practical challenges of markets that use competing credit score models is the need to correlate scores from different providers on an apples-to-apples basis in pools or portfolios reflecting a mix of models. For example, it would be quite difficult for investors, rating agencies, lenders, risk managers, regulators, and other mortgage finance ecosystem participants to gauge whether a 700 under “Credit Score Model A” is equal to a 700 or some other — and potentially very different — score under “Credit Score Model B.” We can attempt to compare scores under competing models using analytical tools such as odds tables or odds charts, but if the answer is different for different loans and different borrowers, this would reflect an inconsistent and unreliable correlation. This begs the question: What are the implications and risks of the resultant inadequate disclosures, data, or analytics?

Mortgage finance stakeholders must pay careful attention to “nuts-and-bolts” details that, if ignored or misunderstood, can become landmines when implementing even the most well-intended new policies or practices.

FHFA wisely acknowledged this danger as reflected in its Final Rule on the Validation and Approval of Credit Score Models. The Final Credit Score Rule provides for the evaluation of the potential impact of new or different credit score models across the myriad activities of Fannie Mae and Freddie Mac that involve credit scores. This includes but is not limited to Private Mortgage Insurance Eligibility Requirements, Unified Mortgage-Backed Securities regulation, Credit Risk Transfer transactions, and GSE capital requirements.

One of FHFA’s proposed options with respect to credit score models was to give lenders the choice to select between or among models. In my listening session comments, after noting the need for an “apples-to-apples” dynamic, I stated:

“Even if we could create valid, consistent, reliable odds charts correlating competing credit score models, we must require a lender to identify up front which model it will use to originate a loan. We must not, under any circumstances, allow forum shopping when it comes to competing models. Nothing good comes from forum shopping — never has. If an appraisal comes in too low to support a loan, we don’t allow additional appraisals until one finally allows the loan to move forward. If we don’t follow the same prudent practice for credit scores, we risk saddling the applicant with mortgage credit that he or she can’t sustain, and potentially violate consumer protections and impair systemic safety and soundness."

This is not to say we are unable to solve the various practical challenges that any one credit score model or competing credit score models pose. It is also true that certain credit score models may be better tailored to certain credit applicants. The same goes for other elements of consumer credit analysis such as underlying data sources and underwriting methodologies. For example, many stakeholders now recognize the potential credit impact of taking rental payment history into account for renters and cash flow analysis for gig workers. Innovations like these can help us enhance the market’s ability to define and recognize creditworthiness.

This reflects one of the cornerstones of FHFA’s Final Credit Score Rule: the need to make real progress in expanding access to sustainable credit among the underserved. We know that many people — most notably in communities of color, as well as low-to-moderate income and “credit invisible” communities — may, in fact, be creditworthy, but fall outside the four corners of certain traditionally applied elements of consumer credit analysis. Leveraging financial technology and new data sources to improve our understanding of creditworthiness is crucial to the ultimate goal of a fair, equitable, secure, and inclusive consumer financial system.

But it takes time for markets and market participants to evaluate new technologies and products and even more time to adopt and implement them. For this reason, FHFA requires in the Final Credit Score Rule a minimum review period, not to exceed seven years, for the evaluation and selection of authorized credit score models for a given term until the next review period. This construct allows for stable, functioning markets, but it also risks technological stagnation and a disincentive for competition. In response, and to encourage continued innovation and competition, FHFA should allow the GSEs (and the Federal Home Loan Banks) to establish pilot programs to evaluate new credit score models and technologies in between official review periods. Pilot programs also provide a safer space to test out new products and initiatives while limiting the risk of harm to consumers and adverse market disruptions.

So, what is the crux of it all when it comes to new and competing credit score models?

We need a mortgage finance system in which anyone who is able to sustain mortgage credit can obtain it on fair, equitable, safe, and inclusive terms. Any credit score model we employ must do what it purports to do in a manner that is statistically and methodologically sound, passing stringent analysis and back-testing and clearly and transparently resolving or explaining anomalies. We should also leverage carefully evaluated innovations, including cutting-edge technology and enhanced data sources and analytics, to achieve these goals, making sure to root out any and all embedded discriminatory elements.

At the same time, we must maintain the stability and proper functioning of the secondary mortgage markets as we evaluate, adopt, and implement any credit score model. That means using one credit score model for a reasonable, defined term, supported by pilot programs, unless an alternative model can be safely, soundly, and consistently correlated and integrated. Should a time come when two or more credit score models are in use, we must not allow lenders to employ a second-chance alternative credit score simply because an applicant’s initial credit score rendered them ineligible for mortgage credit. We must instead implement protocols and procedural guardrails around credit score model selection to prevent irresponsible lending practices focused on “chasing volume” at the expense of credit sustainability and consumer protection.

Responsible innovation and implementation of credit score models can enhance our understanding of creditworthiness and help expand access to sustainable credit. But failure to adhere to the principles laid out above could adulterate pricing, risk, and other important methodologies, causing long-lasting damage to consumers and our mortgage finance system — despite the best of intentions.

For reprint and licensing requests for this article, click here.
Credit scores
MORE FROM NATIONAL MORTGAGE NEWS