Case Study 1 — The 2023 Restructuring of the GSE Pricing Framework

A real, public case. In late 2022 and the first half of 2023, the Federal Housing Finance Agency (FHFA) directed a broad recalibration of the upfront fee matrices that Fannie Mae and Freddie Mac charge on the loans they buy — the grids this chapter has been calling loan-level price adjustments. The changes were technical. The public argument about them was not, and for several weeks in the spring of 2023 a pricing matrix became front-page news.

This case is here because it teaches three things at once: how the grids are actually structured and governed, what happens when a pricing input is chosen badly, and how a loan officer should talk about a pricing change while it is being argued about on television.

A note on numbers. This case study deliberately contains no grid values. The published matrices are revised periodically, the versions in force at any moment are the only ones that matter, and reproducing 2023 cells in a textbook would be exactly the habit §7.1 of this book's standards forbids. Everything below is structural. Verify the current framework and the current matrices at FHFA, in the Fannie Mae Selling Guide, and in the Freddie Mac Seller/Servicer Guide.

Background: what these grids are and who sets them

The GSE upfront fees are not set by lenders. Fannie Mae and Freddie Mac publish them, FHFA — their conservator since 2008 — supervises them, and every lender that delivers loans to the enterprises passes them through. That is why the same borrower characteristics move price in a broadly similar direction at almost every conventional lender in the country, and why a loan officer cannot negotiate them.

The fees have a stated purpose: to align the price of a loan with the risk it carries and with the capital the enterprises are required to hold against it. That last clause matters. FHFA finalized an Enterprise Regulatory Capital Framework, and once the enterprises had a capital rule with defined risk weights, their pricing grid and their capital requirement were two descriptions of the same thing that no longer matched. The 2022–2023 changes were, in FHFA's own framing, a re-alignment exercise.

The re-alignment came in stages:

• 2022. FHFA announced targeted fee changes, including increases on second-home loans and on high-balance loans, and — in the other direction — the elimination of upfront fees for several affordable categories: first-time homebuyers at or below area median income (with a higher threshold in high-cost areas), the enterprises' affordable products such as HomeReady and Home Possible, and state housing finance agency products.
• January 2023. FHFA announced a comprehensive recalibration of the upfront fee matrices across purchase, rate-and-term refinance, and cash-out refinance, effective for loans acquired on or after a May 2023 date.
• The same announcement introduced something new: an upfront fee keyed to the borrower's debt-to-income ratio, applying above a 40% DTI threshold.

Three of those four items were unremarkable inside the industry. The fourth was not.

The issue, part one: a pricing input that will not hold still

Every input to a rate sheet in this chapter has one property in common. The representative score is fixed when credit is pulled. The LTV is fixed when the loan amount and the value are known. Occupancy, property type, purpose, and product are fixed at application. The lock period is chosen once.

Debt-to-income is not like that, and every loan officer knows it.

DTI moves. It moves when an underwriter recalculates variable income on a different convention than you did. It moves when a paystub arrives showing a different year-to-date. It moves when a borrower pays off a car or, as on the Linden Street file, finances \$5,200 of furniture on day 41. On that file the back-end ratio was 42.66% at approval, 48.48% after the day-44 credit refresh, and 42.66% again after the account was paid. Under a DTI-based fee with a 40% threshold, that file would have crossed a pricing boundary twice in seven days — after the rate was locked, after the Loan Estimate was issued, and days before closing.

The industry's objection was operational rather than philosophical, and it was specific:

• Pricing would change after the lock. A locked rate is a promise, and a fee that can attach after the promise is made breaks the thing a lock is for.
• Disclosure would be disrupted. A change in points or credits late in the process interacts with the TILA-RESPA Integrated Disclosure rule's tolerance and redisclosure requirements (Chapter 22), and a fee driven by a number the underwriter controls is a poor fit for a system built on disclosing costs early and holding them.
• It would be gameable in the worst way. A borrower a few tenths of a point over a threshold has an incentive to move money around before the final recalculation, and a loan officer has an incentive to help. Chapter 27 explains why that is a category of pressure nobody in this business wants added to a file.

The Mortgage Bankers Association and other trade groups raised these points publicly. FHFA first delayed the DTI-based fee, moving its effective date out by three months, and then, in May 2023, announced it would rescind it entirely.

That is the outcome, and it is worth stating plainly because it is unusual: a proposed pricing input was withdrawn before it ever took effect, on operational grounds raised by the people who would have had to administer it. Not on grounds of fairness, and not because of the argument described below. Because it would not work.

The issue, part two: the public argument

In April and May of 2023, a different story about the same announcement reached a general audience. The claim, in its most compressed form, was that borrowers with high credit scores would now pay more in order to subsidize borrowers with low credit scores.

The claim rested on a real feature of the recalibration. When a grid is re-gridded, some cells go up and some go down, and it was true that certain higher-score cells rose while certain lower-score cells fell. Compare a single cell before and after and you can produce an accurate sentence that supports the story.

FHFA responded publicly, in a statement issued in the spring of 2023, calling the characterization a fundamental misunderstanding. Its position had two parts, and both are checkable against any version of the matrices:

1. The ordering never inverted. At any given LTV, a borrower with a higher credit score continued to pay a lower fee than a borrower with a lower score. No cell arrangement made strong credit more expensive than weak credit at the same LTV.
2. The changes were a capital-alignment exercise, not a transfer program. Cells moved because the risk weights they were being matched to had moved.

FHFA also opened a formal Request for Input on the single-family pricing framework, inviting comment on the goals of the framework, how the fees should relate to the capital rule, and how much cross-subsidy is appropriate in a system operating under conservatorship.

Both things can be true, and both are: individual cells rose for some higher-score borrowers, and the grid never stopped charging less for stronger credit. A loan officer who understands §29.4 can say that sentence in ten seconds and be right, which in the spring of 2023 was worth a great deal.

What this shows

One. The grid is not your lender's opinion, and saying so is a service. When a borrower with a 760 score arrives angry about something they read, the useful move is not to defend or attack the policy. It is to open the current matrix, find their cell, find the cell of the borrower they think is being favored, and let the two numbers speak. That is a thirty-second conversation and it is almost always more persuasive than an argument.

Two. A pricing input has to be knowable early and stable late. Score, LTV, occupancy, property, purpose, and lock period all are. DTI is not, and the 2023 episode is the clearest available demonstration of why the list looks the way it does. This is a genuinely useful test to carry: when you hear about a proposed new pricing factor, ask when it is fixed and whether it can move after the lock.

Three. Perishability is a property of the material, not a disclaimer. The matrices changed in 2022, changed again in 2023, had a component added and then removed inside a single year, and were the subject of a formal request for input on the whole framework. Any loan officer who memorized a cell in 2021 was quoting a fiction by 2023. Learn the structure. Look up the values. Every time.

Four. The affordable-housing waivers are real and are frequently missed. The elimination of upfront fees for certain first-time buyers at or below area median income, and for the enterprises' affordable products, is money that only reaches a borrower if somebody at the desk knows the program exists and checks eligibility. On the Harlow Street file this is not academic; on any first-time buyer file with income near the local median it is worth a specific check. The thresholds and the qualifying products change — verify them for the current year before you rely on them.

Discussion questions

1. Six pricing inputs in this chapter are fixed early and one proposed input was not. Write the general test in one sentence, then apply it to two hypothetical new inputs: (a) the number of months of reserves after closing, and (b) whether the property is in a designated flood zone.

2. On the Linden Street file, back-end DTI was 42.66%, then 48.48%, then 42.66% again, all after the rate was locked. Trace what would have had to happen operationally at each of those three moments under a DTI-based upfront fee. Name every document that would have been affected.

3. FHFA's response and the news coverage were both, in a narrow sense, accurate. Explain how, and then write the two-sentence version you would give to a 762-score borrower who calls you about it.

4. Cross-subsidy — charging some borrowers more than their risk alone would justify in order to charge others less — exists in almost every insurance-like system, including this one. Argue both sides for an entity operating in conservatorship with a statutory mission that includes access to credit. What would you need to know to have a real opinion?

5. This chapter's grids are labeled constructed and illustrative on every appearance. After reading this case, write the sentence you would say to a borrower who asks you to email them "the LLPA chart" so they can check your quote.

Sources and verification. FHFA announcements and press statements on single-family pricing (2022 and 2023), including the announcement, delay, and rescission of the DTI-based upfront fee and the Request for Input on the Enterprise single-family pricing framework; the Fannie Mae Loan-Level Price Adjustment Matrix and the Freddie Mac Loan-Level Price Adjustment schedules as published in the respective Selling and Seller/Servicer Guides; Mortgage Bankers Association public comment. All dates, thresholds, and values must be verified at the source — the framework has been revised more than once and will be again.