GSEs have in the past relied on a singular credit score to evaluate borrower credit risk, but the FHFA is now considering whether Fannie Mae and Freddie Mac should be allowed to use a multi-score approach. The question is whether this would expand access to credit or simply create greater risk for GSEs and for capital markets.
Friday, April 8, 2022
By Clifford Rossi
The Federal Housing Finance Agency (FHFA) is considering changes to how government-sponsored entities (GSEs), like Fannie Mae and Freddie Mac, apply credit scores in their assessment of mortgage borrower credit risk. It is unclear, however, whether the benefits from adopting a multiple credit score approach for expanding credit access outweigh the many risks to both GSEs and capital markets – in the form of GSE-eligible mortgage securities.
The revisions the FHFA are considering speak to a perennial issue facing the lender and regulatory communities: how to expand equitable access to credit while effectively managing risk. Standard analytical tools, such as automated underwriting models and credit scores, have been mainstays in assessing credit risk for decades; however, for a variety of reasons, a large segment of potential borrowers has largely eluded these tools.
In 2015, the Consumer Financial Protection Bureau (CFPB) released a report indicating that 26 million people were effectively “credit invisible” to credit-reporting agencies, while another 19 million were considered unscorable under the existing credit scoring platforms – due to a lack of credit history.
In an effort to provide greater access to credit for this segment, the FHFA is evaluating whether to require the GSEs to rely on a single credit score or to allow alternative credit scores to be used.
GSEs have used FICO credit scores exclusively to evaluate borrower credit risk since the 1996 debut of automated underwriting systems (AUS). The introduction of credit scores into underwriting models, via AUS, was revolutionary, because it gave GSEs (along with the Federal Housing Administration and the U.S. Department of Veteran Affairs) the power to rank-order borrower credit risk consistently – with a high degree of discriminatory power between good and bad loans.
FICO vs. VantageScore
In more recent times, the three U.S. credit bureaus have developed VantageScore, a version of the FICO score product. VantageScore drew the attention of policymakers, with analysis suggesting that its scoring algorithm could significantly increase access to credit for as many as 10 million borrowers who today remain unscorable.
To no surprise, the arrival of VantageScore on the scene has prompted a sort of credit score death match between FICO and VantageScore, with FICO pushing back on the touted effectiveness of VantageScore to be a silver bullet in expanding credit.
The FHFA took notice of VantageScore, and one of the options under consideration by the agency to modify the way the GSEs use credit scores is to allow the use of both scores. Under such a “waterfall” approach, a primary score would be selected for a credit decision and, if that score was unavailable, a secondary score would be used – again, if available. But this methodology could potentially heighten risks for GSEs, investors in credit risk transfer (CRT) securities and, in the extreme, even taxpayers.
Pros and Cons of the Multiple-Score Approach
Consistent with its primary mission of ensuring that GSEs are safe and sound, before making any decisions about shifting away from the single-score methodology, the FHFA should determine whether a multi-score approach could cause harm by increasing credit risk.
The use of multiple scores invites a range of risks to the GSEs that could adversely impact not only their performance but also that of the mortgage secondary market.
In a study they conducted on the effectiveness of alternative credit score methodologies, the GSEs found that their AUS models had higher discriminatory power than using either FICO or VantageScore alone. This is not surprising, given that the GSEs incorporate many other noncredit risk factors into their scorecards.
A key question, however, is whether GSEs obtain greater “lift” in model performance when they supplement their AUS with either FICO or VantageScore. This is an empirical question that could easily determine the best approach to use.
Conducting a traditional champion-challenger analysis of the two scores in each GSEs’ AUS would identify which approach provides the highest model performance – i.e., the highest discriminatory power and the most “swap ins” of unscorables.
Beyond the potential for a multi-score approach to increase credit risk, several technical model-risk issues lurk in the shadows. For example, differences in the manner in which VantageScore and FICO depict risk in terms of odds ratios can introduce score confusion – and even the potential for lenders to adversely select and arbitrage the GSEs in their borrower acquisition and origination processes.
What’s more, a multi-score methodology invites unhealthy competition among score providers. Nowhere was this more evident than with the credit-rating shopping widely in use on private-label MBS before the GFC.
In addition to potential credit and model risk for the GSEs, a multi-score approach invites operational risk, because it would require an overhaul of both GSEs’ AUS platforms. Even more concerning are potential adverse effects of a multi-score approach on capital market instruments, such as MBS and CRT securities.
Understanding the effects of a multi-score methodology on prepayment risk is critical, given that investors in agency MBS hold interest rate risk on these bonds. A change could adversely affect the value of these instruments, as investors try and sort out the potential impact of a transition to a multi-score approach.
Similarly, investors in CRT securities may have concerns about the impact of such a switch on their tranche of credit risk. Remember, data supplied by the GSEs to investors today to make CRT investment decisions are based strictly on FICO scores. So, a transition to a multi-score approach could make it more difficult for CRT investors to discern credit risk, potentially reducing liquidity in the highly successful CRT program.
The FHFA has much to consider in expanding access to credit to unscorable borrowers. It remains unclear, however, whether moving to a multi-score methodology for the GSEs would result in significantly higher numbers of creditworthy borrowers from previously unscored segments of the population.
At the same time, there are simply too many different potential risks to the GSEs and capital markets from moving to a multi-score model. At the end of the day, the goal should be to expand credit to borrowers, while also ensuring the safety and soundness of the GSEs.
Clifford Rossi (PhD) is a Professor-of-the-Practice and Executive-in-Residence at the Robert H. Smith School of Business, University of Maryland. Before joining academia, he spent 25-plus years in the financial sector, as both a federal-banking regulator and a C-level risk executive at several top financial institutions. He is the former managing director and CRO of Citigroup’s Consumer Lending Group. He also worked at both Freddie Mac and Fannie Mae, where he played role in developing the first automated underwriting systems used by Freddie Mac, the FHA and the VA.