Did HUD, FHFA push homebuying into a new era with credit score advancements?

This week, federal housing leaders Scott Turner and Bill Pulte held a joint press conference where they announced the rollout of new credit score models designed to lower costs for mortgage borrowers while continuing to mitigate risk for lenders.

Turner was clear that in the next few months, the U.S. Department of Housing and Urban Development (HUD) will start the process to accept both FICO Score 10T and VantageScore 4.0. “America was built on competition” he said, indicating that the decision would allow “more data to be considered, as new models include borrower utility data and rental history that were previously not included.”

This simultaneous approval and rollout also allows the validated FICO and VantageScore models to compete in a fair and equitable manner for Federal Housing Administration (FHA) loans. In addition, it ensures that the most robust and up-to-date models are being used, which will lead to transparency and market stability for FHA and Ginnie Mae products.

Pulte, the director of the Federal Housing Finance Agency (FHFA), announced that there would be a pilot program for Fannie Mae and Freddie Mac to “operationally test” the acquisition of loans underwritten with VantageScore 4.0. He indicated that up to 21 of the nation’s largest lenders could take part in the pilot and that the selling guides for the government-sponsored enterprises (GSEs) were being updated.

Pulte noted that Freddie Mac has already purchased $10 million in loans tied to VantageScore 4.0, that Fannie Mae has not yet acquired any of these loans, and that no securitizations to date have included loans using VantageScore 4.0. Additionally, he said that the FICO 10T data analysis would be released in the coming months, while confirming that the GSEs must implement separate loan-level pricing adjustment (LLPA) grids for VantageScore and FICO to “accurately reflect the risk weighting of the different models.”

Turner and Pulte say the changes are “advantageous, especially for first-time homebuyers.” Both officials focused on the theme of “new modern credit scoring models” and the value-add proposition for renters who want to become homeowners. “If you paid your rent for 10 years, it should be factored into your credit score,” Pulte said.  

The resounding theme of the press conference was about making credit scoring “cheaper” for borrowers at the time of loan origination. Today, both FICO 10T and Vantage Score 4.0 are being offered for 99 cents — and each are offered at no charge when FICO Classic is purchased in tandem.

What does this mean for homebuyers?

The question now is, does this truly advance a new era for homebuyers? The simple answer is no, since there is more work to be done, but the market should react positively to the announcement since it’s the first time in a while we’ve heard from Pulte and Turner. The fact that they’re communicating about long-standing initiatives is positive.

But let’s dig deeper. The reason that the announcement is not as exciting as it could be is that lenders are not the only stakeholders in the mortgage process. The ultimate “takeout” for Fannie and Freddie is the bond buyer, not them.

To achieve liquidity in the mortgage market, there are a minimum of 10 touch points, and each point represents a party that needs to make risk-based decisions on whether to accept VantageScore 4.0 and what their requirements may be to allow it. The visual below represents the high-level life cycle of a mortgage and gives an overview of the entities that will have to get on board for there to be any material traction in the mortgage industry.

Fannie Mae and Freddie Mac also need many of these counterparties to maintain their businesses. While they are not warehouse lenders themselves, they are linked to the warehouse lending ecosystem, as they provide the liquidity for agency loans originated, while the aggregators are the same liquidity providers for nonagency loans.

Fannie and Freddie, like any other aggregator or balance-sheet asset holder, must hedge the assets on their balance sheet and employ leverage, in the form of bond repurchase agreements and other strategies, to manage their portfolio.

How will the models be evaluated and implemented?

When each of these parties assesses the differentiated risk profiles of the historic and current credit models, they will look at multiple metrics and conduct comprehensive due diligence. This assessment will include, without limitation:

• Financial stability
• Operational resilience
• Historical experience
• Performance-based data
• Regulatory compliance
• Security
• Saturation risk

Each party that approves the use of either vendor or model will require their clients to have certain contractual protections. This involves not only their service-level agreements with fulfillment entities — such as right to audit, breach notifications, etc. — but also disclosure from the model provider and vendors themselves.

Most of the parties shown in the chart above will have to take this detailed analysis to their credit committees, where the organization will then have to make an informed decision about whether they will proceed to allow a new vendor or new models. For those where the score is used for transaction-based execution, a deep dive analysis will need to be conducted into the models, as well as refinements made to their own pricing and/or trading models, to determine and refine pricing context and controls.

To begin this analysis, LLPA grids for both VantageScore 4.0 and FICO 10T need to be released by Fannie and Freddie for the market to review. Generally, if the relevant party believes they will lose money, they are not going to change their current course, since the unknown risks may be too much for their current financing or investment strategies without additional data to analyze.

The reality is, if lenders are unable to get the warehouse lenders, the aggregators, the bond buyers and bond repo providers on board to roll out these new models and allow additional vendors, they will not be able to use them. And if the ratings agencies take a much heavier hand to the VantageScore model when determining subordination levels for securitization, etc., issuers will sit in a similar position — that it does not make business sense to proceed with a change.

Now, let’s say that the market participant gets past the above steps. They move on to their next challenge, which is implementation for their organization. Their systems must be adjusted to accept multiple scoring models. The relevant systems impacted include, without limitation:

• Point of sale
• Loan origination and/or lock engines
• Pricing engine
• Hedging platform
• Automated underwriting systems
• Warehouse financing
• Closing and post-closing
• Due diligence
• Quality control
• Secondary market platform
• Asset management and servicing

Every aggregator must then update their guidelines to reflect the newly approved model or vendor, and every fair lending team must analyze, for example, disparate impact analysis across two models, not to mention the additional compliance requirements and other similar matters.

This is not easy. It is expensive, time-consuming and requires training throughout the organizations. If they are a hedge fund or a real estate investment trust, they are also going to have to get their fund administrator on board with this, as well as any independent directors and/or trustees for acquisition vehicles. This may require an adjustment to their investment strategy, which may need to be approved by investors.

If the company is a lender and 90% of its aggregator partners are not making the change to allow a new vendor or model — and their warehouse lenders were to, for example, increase the required “haircuts” for their financing if originating with VantageScore — it is highly unlikely they make the change for a minority percentage of opportunity. This could lead to confusion and potentially slow down production.  

A business moving from a single process flow to a multi-tier process flow is challenging. Organizations will then have to fully roll out and train the parties impacted. All of this costs a lot of money and takes significant time.

What are Turner and Pulte really saying?

The following is an analysis to identify if all this potential effort as referenced above is worth it, based upon what Pulte and Turner spoke about during the press conference.

Turner stated that “America was built on competition” and Pulte agreed.

I have raised this issue before, as I do not understand how approving and allowing another credit scoring model that is 100% owned via a joint venture by the “Big 3” credit bureaus (Experian, Equifax and TransUnion) equates to promoting competition.

In my humble opinion, it does not. If anything, the system is now creating an actual monopoly, which is defined as a market structure where a single entity controls prices and has the power to exclude competition.

Today, there are four separate legal entities involved in credit reporting and scoring: Experian, Equifax, TransUnion and FICO. They are at arm’s length to each other, and while FICO may have been the only scoring model historically, it did not preclude other companies from creating, attempting to create or marketing a score.

If VantageScore were to obtain material market share, this would provide enhanced power to the credit bureaus to fully control the pricing of all reporting and scoring. In addition, they could in fact get in the way of FICO’s operational execution and offered price to unjustly enrich themselves and promote dominance. This is a moral hazard we, as an industry, should not want to create.

If the intent of the new credit scoring model’s rollout is truly to promote actual competition, why is FHFA not rolling out FICO 10T simultaneously to VantageScore 4.0? HUD has chosen to launch these two updated models simultaneously, and I believe this is in the best interest of all market participants.

Both newer scoring models were approved at the same time in 2022, and the intent has always been for the models to “replace” FICO Classic. At no time was it contemplated that the VantageScore 4.0 model would go live side by side with FICO Classic for Fannie and Freddie usage.

By not rolling out FICO 10T simultaneously, FHFA is not providing the industry with the opportunity to assess the models at the same time. This will lead to a lack of transparency for market stakeholders and may even affect the market’s stability as it pertains to the predictability of default assumptions. This also creates extra expense and duplicative work for market stakeholders when FICO 10T goes live, so why not do it at the same time?

In addition, why isn’t the FICO 10T data analysis being released now? Why is it going to come out between now and the summer when Fannie and Freddie have completed their work? The answer is that they could release it now and have chosen not to — hence this is not allowing for fair competition.

It seems that the reason for the delay is that Fannie and Freddie are now awaiting additional data for VantageScore 4.0 between April 2023 and September 2025, as evidenced by a post on Fannie Mae’s website. The data that is completed should not be held up, as the VantageScore data that’s been released is through March 2023, so an “apples to apples” analysis can be completed now.

I urge market stakeholders to contact FHFA and the GSEs to seek release of the completed FICO 10T analysis.

Pulte and Turner indicated “more data to be considered, as new models include borrower utility data and rental history, which were previously not included.”

While it’s true that newer models do include rental data, only about 13% of all renters have some rental payments reported. And only about half of this number (approximately 7% of renter households) are actively participating in ongoing monthly reporting.

While this could be beneficial to a borrower who consistently makes their payments on time, it can also be a risk. If their rental payments are reported late, this could negatively impact their credit score and future rental lease approvals as well.

As it pertains to utility data, only 2.4% to 3% of consumers are estimated to have any utility tradelines reflected in a routine and ongoing manner. Three in four electric utilities do not report any payment data at all, and most utility data reported comes when an account becomes severely delinquent or is sold to a third party for collections.

There are ways to opt into self-reporting services, where a consumer may attempt to require their utility providers to report, but the consumer is often charged a fee to do so.

What was also surprising to me is that Turner and Pulte focused their discussion points on rental and utility data, when in my opinion, the most significant impact of the new credit scoring methodologies is the trended data component. This now provides for scores to take into consideration performance over time and not just single point in time.  

In addition, there’s a bifurcation of certain categories to get better credit risk context. A good example is how FICO segments unpaid medical collections and non-medical collections into separate buckets. It has also adjusted its model for new personal loan credit risk trends, since personal loans have become a bigger part of consumer debt profiles.

Note, however, that VantageScore prematurely removed all medical collections from its model. A proposed rule may have been forthcoming but never came to fruition, so this is why Turner and Pulte may not have wanted to speak on the topic.  

Additionally, the treatment of “authorized user” accounts is also a beneficial addition. This was addressed beginning with FICO Model 8. As a professional who regularly deals with repurchases and breach-related issues, this can be an indicator of possible fraud or identity concerns, so the extra focus is beneficial.

Finally, the most important thing to keep in mind here is that it’s not the data that differentiates the VantageScore and FICO models, as they both use the same data from the Big 3 bureaus. What differentiates them is how the data is used.

FICO will not use data that does not have six months of history, while VantageScore will use a single month. Can a pattern of utilization or credit behavior be derived from one month of information? My opinion is no. FICO does not score inactive files or those that only have bankruptcy information reported, as this could be detrimental to the consumer.

Finally, just because more scores can be created doesn’t mean more mortgages can be originated.

Turner indicated this will be “advantageous, especially for first-time homebuyers.”

This statement was a bit surprising. Turner spoke about first-time homebuyers and the VantageScore 4.0 model in this context, and it is a well-known fact that the model includes a negative adjustment if a borrower does not have a mortgage. This could actually be detrimental to a first-time homebuyer in certain instances.

Pulte said that “maybe FICO scores going forward will include rental and utility data, like Vantage.”

This is an incorrect statement. FICO was the first to have utility and rental data in its models. These metrics were released by FICO in 2014, as part of FICO Model 9, while VantageScore did not introduce these concepts until 2017.

Pulte said that “we are in business, as Freddie Mac acquired $10 million in loans with VantageScores.”

Depending on the balance of the loans in question, $10 million equates to roughly 20 individual mortgages. The mortgage business produces $1.5 trillion to $2.5 trillion annually. At best, this is pricing the equivalent of a mini-bulk of whole loans.

If this is what the FHFA based its decision on, I believe this would be a first, as I have never seen a counterparty make a material investment or risk-based decision based upon 0.000588% of a single year’s production. Clearly, additional “operational testing” will need to be completed.

Finally, Pulte confirmed that the GSEs will have to implement separate LLPA grids for VantageScore and FICO to “accurately reflect the risk weighting of the different models” and alluded to this being complete.

But the Fannie Mae and Freddie Mac  announcements include no effective dates and provides only general comments: “If approved for use of Vantage 4.0, lenders will be provided further instructions.” This sounds like they’re going to begin the process of lenders applying to be selected for the pilot, but not that they’re ready with the new LLPA grids.

In summary, this week’s press conference, while a positive step for the market by virtue of Pulte and Turner engaging with the industry and providing updates on long-standing matters, has also left us with many unanswered questions.

At this time, based on the analysis herein, I believe the next best step is to make additional inquiries to FHFA, Fannie and Freddie to seek the release of the FICO 10T data through March 2023. The markets should also be provided with guidance on the timing of the actual execution and release of the material information that is still outstanding.

Without limitation, this should include the LLPA grids for VantageScore 4.0 and FICO 10T, which would allow for an assessment of all referenced components. We should make clear, as an industry, that we should proceed with one implementation process simultaneously and not two.