The 2025 GSE credit box: FICO, LTV, and DTI across Fannie and Freddie
A distributional read of 1.6M 2025 Fannie and Freddie acquisitions: two nearly identical conforming credit boxes — median FICO 763-767, LTV 80, DTI 39 — with a third above the old QM line.
For capital-markets and credit-risk desks, the conventional-conforming credit box is the single most important boundary in U.S. housing finance: it defines what Fannie Mae and Freddie Mac will actually buy, and therefore what the securitization machine will fund at agency spreads. It isn't a published cutoff — it's a distribution, the observed shape of what the GSEs acquired. Pulling 2025 acquisitions from both enterprises' loan-level files — 967,607 Fannie loans and 642,976 Freddie loans, over 1.6 million in all — lets us read that shape directly.
The two boxes are nearly identical
Start with the headline: the conforming box is essentially one box operated by two enterprises. On the three variables that drive agency credit risk, Fannie and Freddie land almost on top of each other.
Credit. Median FICO is 763 at Fannie and 767 at Freddie. Both books are heavily prime-weighted: 67.0% of Fannie acquisitions and 69.7% of Freddie's carry a FICO of 740 or higher, while the sub-680 tail is thin on both sides — 8.6% at Fannie versus 6.8% at Freddie. Freddie skews a few points higher, but neither enterprise is reaching materially down the credit spectrum. This is a prime, super-prime-tilted acquisition profile at both.
Leverage. Median LTV is 80 at both — the canonical conforming down-payment line. And both carry an almost identical mortgage-insurance load: 35.2% of Fannie loans and 35.6% of Freddie's are originated above 80 LTV, meaning roughly a third of each book requires borrower- or lender-paid MI to reach the GSEs' 80-LTV credit-enhancement threshold. For anyone modeling MI counterparty exposure or credit-risk-transfer economics, that ~35% high-LTV share is the number to anchor on.
Capacity. Median DTI is 39 at both enterprises — and here sits the most analytically interesting signal in the data.
A third of the book sits above the old QM line
The 43% DTI threshold that anchored the legacy Qualified Mortgage rule remains the market's mental reference point for "stretched" affordability, even after the pricing-based QM rewrite. On that yardstick, both GSE books carry meaningful capacity risk: 31.7% of Fannie acquisitions and 33.4% of Freddie's fall in the 43-64 DTI range. Roughly one in three conforming loans is being originated above the old QM line.
That is not a defect — the GSEs' automated underwriting engines have long approved above 43 DTI when compensating factors (reserves, credit depth, equity) support it, and the AUS-driven QM safe harbor permits it. But it is an affordability-stress indicator worth watching. When a third of new conforming production leans on the upper DTI band, the book's sensitivity to payment shock, income disruption, and rising insurance-and-tax escrows is structurally higher than the prime FICO and 80 median LTV alone would suggest. High FICO and moderate LTV are doing the work of offsetting elevated DTI — a barbell the desk should price and stress accordingly.
Where the two books diverge
The differences are small but directionally consistent. Freddie runs a slightly cleaner credit profile (higher median FICO, thinner sub-680 tail) and is more purchase-weighted: 78.3% of Freddie's 2025 acquisitions are purchase loans versus 71.5% at Fannie. A more purchase-heavy book typically carries a marginally higher LTV and DTI mix — first-time and move-up buyers stretch more than rate-and-term refinancers — which is consistent with Freddie's slightly higher high-DTI share despite its stronger FICO distribution.
Caveats — read before you cite
- Distributional only. No shared loan key between the two files, so every Fannie-vs-Freddie comparison here is book-level shape, not a loan-to-loan match.
- Non-comparable windows. Fannie is full-2025; Freddie is Q1-Q3 only (Q3 appears partial), so raw counts and purchase-share aren't directly comparable until the windows are aligned.
- GSE-eligible slice only. No FHA/VA (that's Ginnie Mae) and no application universe — denials, withdrawals, and non-conforming loans live in HMDA. These files show what the GSEs bought.
Read within those bounds, the 2025 data delivers a clean verdict: the conforming credit box is prime, 80-LTV-centered, and — on capacity — carrying about a third of its production above the old 43 DTI line at both enterprises. The box hasn't loosened on credit; the stretch is in affordability.