The GSEs Are Making Bold Moves to Support the Housing System and Taking on New Risks to Do It

Fannie Mae and Freddie Mac have been taking big swings to support the housing system at the direction of their conservator, FHFA Director Bill Pulte. We are facing a uniquely challenging time in housing finance and lenders are staring down the barrel of an escalating interest rate environment, after persevering through already elevated mortgage rates since 2022. In this environment, it makes sense for housing policymakers to work aggressively to support borrowers and lenders. While credit score modernization and the GSEs repurchases of their own mortgage-backed securities operate at different ends of the housing finance food-chain, both are aimed at making it easier and, ideally, more affordable for lenders to extend credit to potential homeowners. Borrowers and lenders seemingly stand to benefit from these initiatives. But these two actions share another commonality, as they have been executed to date, both of these policy changes introduce new risk to the GSEs themselves. As taxpayers, we are the ultimate holders of this risk and we are quite literally on the hook should the GSEs incur more risk than they can manage.

It’s important to recall how and why the American taxpayer has become directly exposed to GSE financial risk. In late 2008, the GSEs approached insolvency necessitating the creation of an enhanced regulator for the enterprises and a rescue by the Treasury Department. The creation of the Federal Housing Finance Agency and the purchase of nearly 80% of the enterprises’ equity by the Treasury was necessary because the enterprises had increasingly used their funding advantage to subsidize hundreds of billions in purchases of private label mortgage securities, for which roughly a third were comprised of sub-prime mortgages. It is these investments that brought the GSEs to the brink of failure, not their core securitization function.

While the conservatorships of the enterprises are now more than eighteen years old, the taxpayer support for the enterprises remains ongoing because the enterprises continue to benefit from access to a backstop from the Treasury Department. This is why across changes in government, previous leaders at FHFA and the Treasury have worked to derisk the enterprises in order to minimize the chances that either GSE will need to draw on taxpayer funds again. One of the key policy changes that the two agencies have agreed to has been a cap on how large either GSE’s portfolio can become. This portfolio limit has ratcheted down and now sits at $225 billion for each enterprise, as dictated by the preferred stock purchase agreements (PSPAs) that govern Treasury’s investment in the enterprises.

Absent further changes to the PSPAs, the portfolio limit is now fixed at this amount, which provides ample room for each GSE to run its core operations. Last month, however, Director Pulte announced that the GSEs would resume buying back their own MBS. At the beginning of the year, this initiative was announced as a way to bring down mortgage rates and without question it had a discernible effect on mortgage rates which approached the high five percent mark before the conflict in Iran began.

This announcement drew positive reactions across industry participants and some advocacy groups. Significant MBS repurchase activity will have an effect on mortgage rates, though I do not expect that rate reduction would exceed twenty to thirty basis points. In this environment, however, that sort of rate mitigation will have a real impact on originations and refinancings. Given that the GSEs had largely ceased their repurchase activity in the middle of this year, there is ample portfolio capacity under the current limits to deploy roughly $100 billion towards this effort.

This activity, however, is not without risk. As analysts noted during the first round, both enterprises took on meaningful amounts of interest rate risk in the process, measured as “duration gap.” In 2025, the GSEs had no meaningful duration gap, with both enterprises taking on three months or less as indicated in their public filings. However, as of August of this year, both enterprises now each have in excess of twelve months of duration gap and this is before reporting on the new activity from September. As others have noted, with this level of duration gap, the enterprises could lose over a $1.2 billion each if interest rates rise fifty basis points or more. This risk is real as interest rates have only grown since Director Pulte made his announcement and the factors driving this increase show no signs of abating.

The enterprises could mitigate this risk through greater use of hedging. Hedging interest rate risk with derivates, however, is costly and it is unclear if the GSEs will try to more aggressively manage this risk given the new instructions to restart the repurchases. Unless economic factors fundamentally change, the GSEs will be buying MBS in a rising interest rate environment and, absent increased hedging, they will be increasing the risk in their portfolio along the way.

This risk alone is not yet be material, but coupled with how credit score implementation has unfolded, it raises questions about business and risk management priorities for the GSEs. Credit score modernization has been a long time coming, and after a methodical assessment period FHFA and the GSEs have been making policy changes to drive competition among scoring methodologies at lightning speed. While adoption of VantageScore 4.0 (VS4) was being tested in a pilot program with limited lender participation, the GSEs produced a rule-of-thumb that a VS4 score would be priced as a Classic FICO score after applying a 20 point haircut. For example, a 760 VS4 loan would be treated as a 740 Classic FICO score loan for purposes of LLPA adjustment. This haircut reflected a preliminary assessment that VS4 scores tended to be a bit higher than Classic FICO scores for the same borrower, though the interrelationship between scores on the same credit file or borrower can vary.

In early September, Director Pulte announced on X that the pilot program was over and that VS4 could be broadly used by any GSE lender. The GSEs soon followed that announcement with a new official LLPA grid that reiterated the 20 point haircut featured in the pilot. Importantly, with broad adoption, the GSEs left in place lender choice, meaning that for any given borrower a lender could choose which scores to pull and submit.

It’s important to make clear here that the 20 point haircut was a temporary rule-of-thumb approach. Everyone in the market, including the GSEs, is still learning how VS4 and Classic FICO compare with respect to any given borrower. Both methodologies are predictive of credit risk, this is not really a question of quality but rather of the challenges that come from translation. If, for the last forty years everyone spoke Spanish, learning and adopting French would take effort and developing a nuanced understanding of a new language takes time. The 20 point haircut was not unduly scientific, but it reflected a general propensity for VS4 scores to be a bit higher and it also included a small premium for the uncertainty that VS4 would bring to the GSEs over a similar loan scored with Classic FICO. Importantly, the differential likely included a nod to the potentially poorer performance of VS4 loans in the capital markets given the relative newness of the methodology.

Market participants would have expected the haircut to be replaced over time with more data-verified risk adjusted pricing and one would expect that pricing on VS4 would thus improve gradually over time. But this is not what happened, only a few short weeks later, FHFA and the GSEs changed pricing for VS4 loans, eliminating the 20 point haircut.

Pricing a 740 VS4 loan identically to a 740 Classic FICO loan does not eliminate the difference between those scores, it simply means that the GSEs are not compensated for the relative uncertainty around a methodology that they and we are still working to fully understand. A recent Mizuho analysis determined that the new LLPA grid resulted in approximately 4 basis points in revenue reduction for the GSEs, totally roughly $300 million is lost GSE revenue across each $100 billion in issuance. The GSEs can manage this potentially lost revenue, but the reasoning behind the change is not clear. It is unlikely the GSEs learned anything over the course of three weeks that would justify rendering the two scoring methodologies equivalent from a pricing perspective.

The GSEs now allow lenders to choose the methodology they prefer with each loan and they have removed a haircut that attempted to account for some early differences in the two methodologies. With the GSEs effectively subsidizing use of VS4, we can expect greater use of the methodology generally and potentially greater use of VS4 as the only methodology pulled. This means that the new disclosures that the GSEs have announced will be of limited utility to investors as they are unlikely to see many loans featuring both VS4 and Classic FICO.

It will take some time, perhaps months or longer, for VS4 loans put to the GSEs to reach a critical mass in enterprise securitizations. MBS investors that work to understand prepayment performance have built Classic FICO into their models. Once VS4 loans become a greater portion of the pools they buy, they are going to ask for a premium for absorbing collateral they are less familiar with. This means that at some point, likely in 2027, the GSEs will begin discovering their investors appetite for VS4 collateral through how they place and price VS4 heavy pools.

The key question for the GSEs and FHFA is what to do with this new information. In a business run by a management and board working on behalf of shareholders, the answer would be clear. Use capital markets feedback to retool pricing to account for risk and the incrementally poorer performance of VS4. Again, as VS4 proliferates, pricing for these loans should improve with increased MBS investor experience. But the GSEs are not traditional companies, they are run by their government conservator and they expose tax payers to risk. Will the government’s prioritization of the needs of borrowers and lenders lead to a further cementing of the current LLPA structure because it is, in effect, a modest g-fee reduction? I.e., will the GSEs continue to subsidize the use of a new methodology and continue to take on more risk in the process? Or will the need to risk-manage taxpayer provided resources win out and the GSEs will begin pricing for the risk they now better understand? I want to stress that for both GSE MBS repurchases and credit score modernization, we are not looking at material risks to the resilience of the GSEs, at least not yet. But both of these policy choices reflect how current executive branch leadership is grappling with fundamental questions about the nature and role of the GSEs. At present, policymakers are comfortable having the GSEs incrementally take on more risk if it supports important policy priorities. But the public and the industry have not been offered a take on the overall strategy we are watching unfold. Are we at the beginning, middle, or end of the GSEs risk-taking story? Does this risk story change if policymakers decide to undertake a new GSE equity offering? How do policymakers weigh using the GSEs to effect housing policy against both core safety and soundness considerations and the expectations of equity investors?