How GSE risk-based mortgage pricing expanded.2007-2023

What changed structurally

2007

The initial broad LLPA layer

Fannie announced new charges only for credit scores below 680 at LTVs above 70%, ranging from 0.75% to 2.00%. The announcement also revised subordinate-financing and two-unit pricing.

2008-11

Rapid expansion into a full credit-score/LTV surface

Successive revisions widened the grid. By the 2011 matrix, pricing varied across nearly every FICO tier and multiple LTV bands, including meaningful charges for borrowers well above 680.

2015-18

Mature matrix plus cumulative attribute fees

The core score/LTV grid reached 3.00%-3.25% in several lower-score, higher-LTV cells, before adding separate fees for cash-out, investment property, condos, subordinate financing, units, and other attributes.

2022

Targeted overlay increases, not a base-grid redesign

Two 2022 Lender Letters previewed the coming shift: LL-2022-01 (Jan. 2022) raised LLPAs for second-home and high-balance loans; LL-2022-05 (Nov. 2022) waived LLPAs for HomeReady, Duty to Serve, and low-income first-time-buyer loans while raising cash-out refinance pricing—the direct precursor to the 2023 loan-purpose redesign.

2023

Loan-purpose segmentation and finer score bands

Fannie split purchase, limited cash-out, and cash-out into separate grids, added 760 and 780 score tiers, and retained cumulative property/product adjustments. Limited cash-out refinance pricing is generally higher than purchase pricing at the same score and LTV. The values shown here come from the Jan. 28, 2026 version of that 2023-announced framework.

At-a-glance: 80% LTV

Illustrative comparison using the nearest available 75.01%-80.00% column. Focused on representative score buckets around 700, 720, and 740.

700

Representative 700-719 bucket

2007Zero

20111.000%

20181.250%

2023 purchase1.375%

2023 limited cash-out1.875%

720

Representative 720-739 bucket

2007Zero

20110.500%

20180.750%

2023 purchase1.250%

2023 limited cash-out1.625%

740

Representative 740+ / 740-759 bucket

2007Zero

20110.250%

20180.500%

2023 purchase0.875%

2023 limited cash-out1.125%

Core takeawayThe framework evolved from a narrow surcharge aimed mostly at sub-680 borrowers into a comprehensive, cumulative pricing system that affects many prime borrowers and differentiates sharply by loan purpose and property attributes.

Interactive matrix comparison

Choose a vintage. Darker cells indicate larger upfront price adjustments. Values are percentages of the original principal balance.

Lower fee

Higher fee

Where the 2023 pricing changes met actual Fannie Mae loan volume

A cell-for-cell comparison using Fannie Mae's 2022 Single-Family Historical Loan Performance Primary Dataset. The sample contains 1,074,155 purchase loans with original terms of 30 years or longer and $356.6 billion of original UPB. Only 711 loans lacked a representative origination FICO score.

2023 change in borrower cost vs. the 2018 grid

Positive numbers mean the borrower paid more; negative numbers mean the borrower paid less. Red = higher borrower cost in 2023. Green = lower borrower cost. Hover any cell to link the same FICO/LTV cell on the volume map.

Why the top three FICO rows differThe 2018 grid had one broad 740+ tier. The 2023 framework split it into 740-759, 760-779, and 780+. At 75.01%-80% LTV, the same 2018 starting point of 0.500% became 0.875% for 740-759 (+0.375 borrower cost), 0.625% for 760-779 (+0.125), and 0.375% for 780+ (-0.125). So the top row is not the old “740+” bucket; it is the new 780+ tier.

Where Fannie actually bought 2022 purchase loans

Exact FICO × original-LTV distribution from the 2022 Fannie Mae Primary Dataset. Toggle between each cell's share of all loans and actual loan count.

Combined view: where pricing changes overlapped with loan concentration

Each cell shows its contribution to the sample-wide average borrower-cost change: the cell's share of 2022 Fannie purchase loans multiplied by the 2023 change in borrower cost. Red cells pushed the weighted average fee higher; green cells pulled it lower.

Weighted fee relief

Weighted fee increase

Source and scope: Fannie Mae Single-Family Historical Loan Performance Primary Dataset, 2022 Q1-Q4, purchase loans with original terms of 360 months or longer. The combined acquisition/performance files were reduced to one record per loan and bucketed by representative origination FICO and original LTV. This is Fannie Mae only, not Freddie Mac or the combined GSE market.

The three-bucket squeeze

The three largest LTV buckets in the GSE purchase market were 90.01%-95%, 75.01%-80%, and 85.01%-90%. Together, they represented 60.85% of 2022 purchase originations—and all three became more expensive for a 738-score conventional borrower under the 2023 grid. Experian reports 738 as the average credit score at origination for conventional/conforming mortgages, based on 2022 HMDA data.

Where the borrowers are

60.85%

The three largest LTV buckets accounted for more than three out of every five GSE purchase loans.

20% down bucket

+0.500

LLPA rose from 0.750% to 1.250%. On a $320,000 loan, that is about $1,600 more upfront.

5%-10% down bucket

+0.375

LLPA rose from 0.500% to 0.875%. On a $380,000 loan, that is about $1,425 more upfront.

10%-15% down bucket

+0.500

LLPA rose from 0.500% to 1.000%. On a $400,000 home, the added cost is roughly $1,700-$1,800, depending on the exact LTV.

Origination-share view: all nine Urban Institute LTV buckets are shown, with the three largest highlighted.

Fee evolution at 80% LTV

Representative score buckets centered on 700, 720, and 740, using purchase/no-cash-out pricing where the source provides it. The 2023 dashed comparison is limited cash-out refinance. For the 740 bucket, earlier vintages use 740+ while the current Fannie grid uses 740-759.

Interpretation

Prime borrowers entered the grid. The initial 2007 announcement did not impose the new base surcharge on 680+ borrowers. By 2011 and 2018, 680-739 borrowers generally faced non-zero LLPAs at 80% LTV.
The peak base-grid burden rose materially. In the 2018 grid, the 75.01%-80% column reached 3.00% for several sub-660 tiers. The current 2023 purchase grid peaks at 2.75%, while limited cash-out reaches 3.50% and cash-out reaches 5.125% at 75.01%-80%.
Refinances became a distinct pricing category. The 2023 redesign formally separated purchase and limited cash-out. At 80% LTV in the current matrix, the limited cash-out grid is 0.125%-0.625% worse than purchase across many score tiers.
Base-grid comparisons understate total borrower impact. Condo, investment-property, second-home, high-balance, subordinate-financing, and other LLPAs are cumulative.
Cross-GSE caution. The 2018 document supplied is Freddie Mac Exhibit 19; the 2007 and current 2023-framework documents are Fannie Mae. Their grids were broadly parallel but not legally interchangeable.

Fannie Mae annual profitability

Annual net income or loss from 2007 through 2025, followed by first-half 2026 net income. Values are rounded to the nearest tenth of a billion dollars.

Net income (loss), $ billions

The profitability shiftFannie reported losses from 2007 through 2011, returned to annual profitability in 2012, and has remained profitable annually since. The extraordinary 2013 result included large reserve releases, deferred-tax-asset effects, and repurchase settlements, so it should not be read as a normal operating run rate. “1H26” is a six-month amount and is not annualized.

Fannie Mae acquisition volume vs. net income

Annual mortgage-market liquidity provided compared with reported net income, 2015-2025. Values are in billions of dollars.

Mortgage-market liquidity / acquisition volume, $ billions

Net income, $ billions

Volume fell. Earnings did not fall with it.Annual volume dropped sharply after the 2020-2021 refinance boom, but net income remained comparatively resilient. One important reason is that Fannie earns guaranty-fee revenue from its entire multi-trillion-dollar guaranty book, not only from loans acquired in the current year. The large stock of low-rate mortgages is less likely to refinance or prepay, which can extend the period over which those loans generate guaranty-fee income. Net income also reflects credit-loss provisions, fair-value changes, taxes, and operating expenses, so the relationship is not one-for-one.

“Volume” follows Fannie Mae's reported annual liquidity provided to the mortgage market. The 2025 release reports $409 billion of liquidity provided and $14.4 billion of net income.

Fannie Mae net worth accumulation

Net worth, $ billions

$103 billion of net-worth growthFannie grew net worth from approximately $13.5 billion at the start of 2020 to $116.5 billion at June 30, 2026. The quarterly detail shows a steady retained-earnings pattern: $77.7 billion at year-end 2023, $94.5 billion at year-end 2024, $109.0 billion at year-end 2025, $112.7 billion after the first quarter of 2026, and $116.5 billion after the second quarter.

How Fannie recognizes LLPA revenue

An LLPA is collected upfront from the lender, but Fannie generally does not recognize the entire amount as income on day one.

Step 1

Cash is received at acquisition

Upfront fees primarily related to single-family LLPAs are received when Fannie acquires the loan.

Step 2

The fee becomes a cost-basis adjustment

The amount is deferred rather than immediately recognized as current-period profit.

Step 3

Income is recognized over time

Fannie amortizes the deferred amount into net interest income over the loan's contractual life, with actual recognition accelerated when loans prepay or otherwise liquidate.

This distinction matters when evaluating LLPAs: the borrower or lender pays the adjustment upfront, while Fannie's reported earnings benefit is spread across future periods. As of recent filings, Fannie separately tracks a substantial balance of “unamortized deferred guaranty fees” that will flow into income over the remaining lives of loans and related MBS debt.

There is a deeper history behind the LLPA grid.

The pricing matrix is only one layer. A separate research inventory traces the additional fees, surcharges, capital rules, and policy-driven repricing that accumulated around conforming mortgages after 2008.

Research annex

The Full Stack

Government-imposed and government-directed cost layers on conforming loans since conservatorship

Keep the buckets separate. Treasury remittances, conservator-set pricing, structural cost drivers, withdrawn proposals, and federal regulations are not the same thing.

A. Remitted to Treasury

Money charged through the conforming channel that the Enterprises do not retain.

Item	Amount	Active	Research note
TCCA surcharge	10 bp ongoing	Apr 2012 – Oct 2032	Created in 2011 to offset a payroll-tax cut and later extended through 2032 by the Infrastructure Investment and Jobs Act. The research inventory treats this as a Treasury remittance rather than mortgage-risk pricing.

B. Pricing set by the conservator

Enterprise pricing and fee layers imposed or directed during conservatorship.

Date	Item	Amount	Research note
2008	LLPA matrix created	varies	The risk-based upfront grid itself enters the record in 2008.
2008–2015	Adverse Market Delivery Charge	25 bp upfront	Flat charge on single-family acquisitions for seven years, without credit/LTV differentiation.
Dec 2012	Base g-fee increase	+10 bp ongoing	DeMarco directive intended to move pricing toward private-capital economics and encourage private entry.
Sep 2015	AMDC replacement packagekey finding	+25 bp +25 bp +37.5 bp	The pivotal 2015 episode. The research inventory says FHFA removed the 25 bp AMDC but stated that the lost revenue had to be recovered. The replacement package added 25 bp to loans with LTV ≤80% and FICO ≥700, 25 bp to high-balance loans, and 37.5 bp to cash-out, investment-property, and secondary-financing loans. The stated rationale was to let borrowers unable to afford 20% down receive the full benefit of the AMDC removal.
2013–	Credit-risk-transfer costs	embedded	CAS/STACR investor premiums became a cost of the guarantee and are recovered through guaranty-fee economics.
Dec 2020 – Aug 2021	Adverse Market Refinance Fee	50 bp	Applied to refinances during COVID-loss projections; later repealed after eight months. On a $300,000 loan, 50 bp is $1,500 upfront.
Apr 2022	Second-home & high-balance LLPAs	up to 3.875%	Sharp repricing of second homes and high-balance loans before the broader 2023 grid redesign.
Feb 2023	Cash-out refinance LLPAs	increased	Separate action preceding the May 2023 purchase-grid restructuring.
May 2023	LLPA grid restructure	up to +75 bp	The research inventory characterizes the increases as concentrated in roughly 700–779 FICO and 70–95% LTV, with relief elsewhere in the grid.
2010 → 2024	Average g-fee	26 bp → 65 bp	The inventory cites FHFA's annual guarantee-fee reporting: roughly a 2.5× increase in the average g-fee over the period.

C. Structural cost drivers embedded in the price

Not line-item LLPAs, but policy structures that influence the economics the fees must cover.

Item	Research note
ERCF finalized Dec 2020	Raised required capital across the book. The research inventory notes that FHFA's 2022 guarantee-fee report attributed fee increases to ERCF components such as risk weights and buffers.
Net Worth Sweep effective Jan 2013	Swept essentially all Enterprise net worth to Treasury for years, limiting retained-capital accumulation during that period.
Senior preferred ratchet	The inventory highlights the interaction between retained earnings and Treasury's senior-preferred liquidation preference.
Conforming loan-limit ratchet	Not a fee, but a structural expansion mechanism for the conforming channel. The source inventory flags current limit figures for verification immediately before publication.

D. Announced, then withdrawn

Useful as evidence of how quickly nationwide pricing policy can change administratively.

Item	Outcome
Dec 2013 g-fee increase (+10 bp)	Announced under DeMarco and suspended by Watt in January 2014 before taking effect.
DTI-based LLPA	Announced in January 2023, delayed, and rescinded May 10, 2023 before implementation.
Adverse Market Refinance Fee	Actually collected for eight months, then repealed.

E. Federal regulation — different agency, same loan

Real conforming-loan cost channels, but analytically separate from Fannie/Freddie pricing.

Item	Cost channel
HVCC → AIR 2009, GSE-originated	Appraisal-independence rules and the growth of the AMC layer. The research inventory treats this as unusually relevant because the original framework arose from Fannie/Freddie agreements.
Ability-to-Repay / QM 2014	Documentation and underwriting burden; the former GSE patch gave conforming loans a distinct safe-harbor advantage.
TRID / Know Before You Owe 2015	Systems, disclosure-timing, tolerance, and ongoing compliance costs.
Expanded HMDA 2018	Substantial expansion in reportable data fields and compliance workload.
Risk retention, Reg AB II, bank-capital treatment of PLS	Not borrower fees; relevant to the competitiveness and scale of the private-label channel.
Cost to originate	The inventory recommends pulling the current MBA Quarterly Mortgage Bankers Performance Report figure immediately before publication rather than relying on an estimate.

The 2015 ↔ 2023 pattern worth investigating

2015: after removing a flat 25 bp adverse-market charge, FHFA's replacement package put 25 bp specifically on LTV ≤80% / FICO ≥700 loans, while also repricing high-balance and other attributes.

2023: the purchase grid was restructured again, with many increases concentrated in mainstream prime-credit/LTV combinations and relief in other cells.

The research question is therefore broader than whether the 2023 matrix was controversial: is there a recurring policy method in which relief for one cohort is funded by repricing another high-volume, higher-credit cohort? The 2015 record is potentially important because that revenue-recovery logic was described contemporaneously rather than inferred after the fact.

Research inventory source list and verification notes

FHFA annual single-family guarantee-fee reports; FHFA guarantee-fee history; FHFA April 2015 guarantee-fee review results; FHFA releases from 2012, 2015, 2021, and 2023; Fannie Mae lender letters and LLPA matrices; TCCA/Infrastructure Investment and Jobs Act; Collins v. Yellen; Urban Institute analyses. The supplied research inventory flags current conforming-loan limits and current MBA cost-to-originate figures for fresh verification before publication.

Source notes

2007: Fannie Mae Announcement 07-16, dated Nov. 6, 2007, effective Mar. 1, 2008. It lists new LLPAs for LTVs above 70%: below 620 = 2.00%; 620-639 = 1.75%; 640-659 = 1.25%; 660-679 = 0.75%.
2011: Stephanie H. Lo's 2017 Harvard JCHS paper reproduces the shape of the Fannie matrix effective Apr. 1, 2011 through Aug. 31, 2015. Values here are a careful transcription/reconstruction of Figure 11 and the paper's numeric discussion. Treat this vintage as research reconstruction, not an original agency rate sheet.
2015-2018: Freddie Mac Exhibit 19 dated Mar. 28, 2018. Its change summary states that the score/LTV grids were increased and restructured effective Sept. 1, 2015; later 2016-2018 changes shown in the summary were primarily other grids or product features. The comparison therefore labels this core grid “2015-2018.”
2023 current framework: Fannie Mae's Jan. 28, 2026 matrix. The document's own change log says the redesigned loan-purpose-specific framework began Jan. 19, 2023 and became effective May 1, 2023. This presentation therefore labels those grids as 2023 purchase, 2023 limited cash-out, and 2023 cash-out refinance, even though the displayed values are taken from the Jan. 28, 2026 version of the matrix.
2022 bridge letters: Lender Letter LL-2022-01 (Jan. 5, 2022, effective Apr. 1, 2022) raised LLPAs on second-home and high-balance loans. Lender Letter LL-2022-05 (Nov. 4, 2022) waived LLPAs for HomeReady, Duty to Serve, and qualifying low-income first-time-buyer loans effective Dec. 1, 2022, and separately raised cash-out refinance LLPAs effective Feb. 1, 2023—pricing that was then superseded by the permanent cash-out refinance grid introduced under LL-2023-01 on May 1, 2023.
Representative credit score: Experian reports that the average credit score at origination for conventional/conforming mortgages was 738, using 2022 Home Mortgage Disclosure Act data obtained through the CFPB. The figure was also cited in the Yahoo Finance article supplied for this project. Experian source | Yahoo Finance article.
LLPA income recognition: Fannie Mae's SEC filings describe upfront single-family LLPAs as deferred cost-basis adjustments amortized into net interest income over the contractual life of the loan. Its average charged guaranty-fee metric annualizes upfront cash payments using an estimated average life at acquisition; prepayment speeds affect the timing of recognition.
Volume versus income: Annual volume figures for 2015-2024 follow the historical series used in the supplied comparison graphic. Fannie Mae's fourth-quarter 2025 release reports $409 billion of liquidity provided to the mortgage market and $14.4 billion of full-year net income. It also states that fourth-quarter net revenues were primarily driven by guaranty fees on a $4.1 trillion guaranty book of business.
Financial history: Annual net-income figures are rounded from Fannie Mae annual reports and results releases. The 2007-2011 loss period and unusual drivers of the 2013 profit are also documented by the Congressional Research Service. The latest point uses Fannie's second-quarter 2026 release: $4.0 billion quarterly net income, approximately $7.7 billion for the first half, and $116.5 billion of net worth at June 30, 2026.