themortgagellm™

‹ All insights

HomeReady and Home Possible through the rate shock: stable share, changed borrowers

 ·  The Mortgage LLM Team

Written by The Mortgage LLM Team — a group of industry analysts leveraging our proprietary mortgage-domain language models to synthesize and decode housing data.

The affordability crisis started in 2022. Rates doubled from 3% to over 7% in eighteen months. Home prices held. Purchase-origination volume collapsed to less than a third of its 2021 peak. You might have expected the affordability programs — HomeReady on the Fannie side, Home Possible on Freddie’s — to either boom (they exist to help borrowers stretch through exactly this environment) or bust (their income and LTV design points look like they belong to the pre-COVID market, not a $500K-median-home one). Neither happened. HR + HP’s share of GSE conforming purchase originations is now within 2 percentage points of its 2019 baseline. The composition of who uses them, and where they use them, has changed dramatically.

📌 Executive takeaways by role

  • Originators & marketing / LO teams: Don’t assume the HR/HP applicant is still the 97 LTV, 3%-down first-time buyer with a 720 FICO. Post-2023 mean LTV on these programs has dropped from 92% to 85%, FTHB share has fallen from 79% to 74%, and more of the volume is now repeat buyers using the LLPA advantage rather than borrowers who need max leverage. Adjust your marketing personas. Jump to originator takeaway.
  • Capital markets & MBS investors: HR + HP pools originated 2023+ have a materially different composition than pre-COVID pools — lower LTV, higher DTI, more repeat-buyer. The recent-vintage prepay and default profile is not comparable to 2018–2021. Model these vintages separately, and expect the geographic mix to skew heavily toward heartland states rather than the coastal/mountain-west boom markets that once dominated. Jump to capital-markets takeaway.
  • Program stakeholders & policy analysts: The programs’ income limits didn’t need adjustment — the market adjusted itself geographically. HR + HP has quietly rotated from boom markets (ID, UT, AZ, WA, MA, NV all lost 9–13 pp of share) to the midwest heartland (IN, IA, SD, OH, WI gained 4–10 pp). The design worked. But the coastal exclusion — driven by home-price growth pushing incomes out of program limits — is worth understanding before any conversation about high-cost-area income-limit expansion. Jump to policy takeaway.

What the raw data shows

Fannie and Freddie both flag their signature affordability programs directly at the loan level. On the Fannie side, fnm_sfp_raw.special_eligibility_program = 'H' identifies HomeReady originations. On the Freddie side, fre_slld_origination_raw.program_indicator = 'H' identifies Home Possible. Both flags are populated with 100% coverage across every vintage from 2018 forward — no coverage-hole problem. We deduplicate Fannie’s monthly servicing file to one row per loan at first appearance; Freddie’s Standard Dataset is already loan-level origination.

Restricting to primary-residence purchase-money originations gives a clean numerator (the HR or HP loan) and a clean denominator (the conventional conforming purchase universe that these programs compete inside). The quarterly trajectory from 2018 Q1 through 2025 Q3 tells the story.

📈 Quarterly share of GSE conforming purchase, 2018–2025

The interactive chart below tracks HR + HP’s share of the conventional conforming purchase market on the primary axis, with the mean origination note rate overlaid on the secondary axis. The U-shape is the entire story.

Hover any quarter for the HR + HP loan count, total conforming purchase volume, and mean origination note rate.

Three anchors define the shape:

  • 2019 Q2 baseline: 17.77% — the peak of the pre-COVID affordable-lending share. HR + HP accounted for nearly one in five conforming purchase originations.
  • 2022 Q1 trough: 7.86% — the deepest point during the peak refi wave. Not because HR + HP was shrinking — HR + HP absolute volume actually grew through 2021 — but because the denominator (total conforming purchase originations) had ballooned as rate-and-term refis pulled non-affordable-program borrowers into GSE acquisitions in record numbers. It’s the classic denominator-dilution shape.
  • 2025 Q3 recovery: 15.53% — within 2.3 pp of the pre-COVID baseline. The rate shock cleared out the refi wave and restored HR + HP’s structural share of the market.

The volume story is different

The share trajectory tells one story. The absolute-volume trajectory tells another. HR + HP purchase-primary originations peaked at ~333,000 loans in 2021 (year total across both GSEs). They crashed to ~160,000 in 2022 as the rate shock cut the entire purchase market roughly in half, and have recovered only partway — annualizing at about ~200,000 loans in 2024–25, still ~35% below the 2021 peak.

That decline tracks the total conforming purchase market closely. The affordability programs shrank by approximately the same proportion as the overall market. So the composition stayed steady even as absolute participation collapsed. Neither program grew counter-cyclically to compensate for the affordability crisis. Neither was squeezed out by it. Both simply moved with the market.

But the geography rotated

Aggregate stability hides a big compositional shift. Comparing state-level HR + HP + HFA share of purchase originations in 2019 vs 2024 reveals a coherent geographic rotation.

📊 Geographic rotation: 2019 vs 2024 state-level share shift

The interactive chart below tracks the shift in HomeReady and Home Possible market share across states. Notice the distinct migration out of high-cost mountain and coastal markets and into midwest heartland states.

Key finding: Program income caps (80–100% AMI) acted as an automatic geographic router — moving program participation away from markets where home-price inflation outpaced eligible incomes.

Hover any state bar for the 2019 share, 2024 share, and total 2024 conforming purchase volume.

The heartland gained affordable-lending share across the board:

  • Indiana +9.7 pp (19.1% → 28.8%)
  • Iowa +8.1 pp (23.4% → 31.5%)
  • South Dakota +7.1 pp · West Virginia +5.9 pp · Missouri +5.8 pp · Nebraska +5.6 pp · Connecticut +5.2 pp · Ohio +4.7 pp

Meanwhile the boom-market states — precisely where home prices climbed fastest after 2020 — saw dramatic affordable-share drops:

  • Idaho −12.9 pp (33.4% → 20.5%)
  • DC −12.3 pp · Utah −10.6 pp · Arizona −10.6 pp · Washington −10.2 pp · Massachusetts −10.0 pp · Nevada −9.4 pp

The 2019 → 2024 state-level correlation is 0.73 — meaningful reshuffling, not just proportional churn. The mechanism is straightforward: HR and HP have income limits (80–100% of area median income, roughly). When home prices climb 40–60% in a state but incomes don’t keep up, the price of an affordable home moves out of program eligibility even for the borrowers the programs were designed for. In Idaho, Boise’s median home price roughly doubled between 2019 and 2024. In Indiana, Indianapolis prices grew maybe 30%, and program-eligible incomes could still buy a house.

HR and HP are working as designed — they’re routing borrowers where affordability actually exists. But the geographic footprint they cover has narrowed materially. A national lender relying on 2019 marketing personas will have their HR/HP volume concentrated in the wrong places.

And the borrowers changed

The third layer of the story is compositional. We split HR + HP originations into three cohorts — pre-shock (2018–21), transition (2022), post-shock (2023–25) — and computed each cohort’s mean FICO, LTV, DTI, loan amount, and first-time-homebuyer share.

👥 Borrower profile drift: pre-shock, transition, post-shock cohorts

The three panels below track mean LTV, first-time-homebuyer share, and mean DTI across the three cohort windows, with HomeReady and Home Possible overlaid. All three metrics moved in directions that reshape the “who is this program for?” answer.

Key finding: The programs no longer serve primarily high-leverage first-time buyers. They increasingly serve income-eligible repeat buyers with meaningful equity who are using the LLPA pricing advantage to stretch on income in a high-rate environment.

Hover any point for the exact mean value per cohort.

Two shifts stand out as directionally consistent across both GSEs:

Mean LTV dropped from ~92% to 85%. These are still 97-LTV-eligible programs. Pre-COVID, most HR and HP loans came in at or near the LTV ceiling — that was the entire point. Post-shock, the mean LTV has fallen by seven full points. The distribution has moved from “borrowers using max leverage because they have to” toward “borrowers who could put more down and still are.” The 90th-percentile LTV is still at 97% (the ceiling holds), but the median and mean of the cohort have moved down.

FTHB share dropped from ~79% to 74%. Pre-COVID, HR + HP was almost exclusively first-time-buyer territory. Post-shock, the FTHB share has fallen by 5–6 percentage points — meaning repeat-buyer use of these programs has grown from ~20% to ~26% of the cohort.

The combined shift explains what’s happening: HomeReady and Home Possible aren’t just affordability tools anymore, they’re pricing tools. Both programs carry LLPA advantages relative to the standard grid at 90/95/97 LTV. A repeat buyer with equity from a prior home who’s stretching to afford a purchase in a high-rate environment can now use HR or HP to get pricing they wouldn’t otherwise qualify for — even though they don’t need the high-LTV feature.

Mean DTI ticked up (39% → 42%) and mean loan amount rose ~10% (roughly tracking inflation). Mean FICO ticked up slightly on the HomeReady side (740 → 750) and stayed flat on Home Possible. The borrower is stretchier on the income side, lower-leverage on the equity side, less likely to be a first-time buyer, and more likely to be using the program for the pricing advantage.

HR vs HP: mostly parallel, with one interesting divergence

Both programs moved together through the cycle — the quarterly HR-share vs HP-share correlation across the 32 quarters is 0.85. That’s high enough to make “HR and HP behaved the same way” the honest read.

The one place they diverged is the recovery pace. Between 2022 Q4 and mid-2024, Home Possible recovered its share faster than HomeReady — a persistent 2–3 pp gap where HP was already back to 11–12% while HR was still at 7–11%. That’s a Freddie-specific timing effect, likely tied to Home Possible’s slightly broader eligibility on multi-unit and CLTV allowances, which gave the program a marginal advantage in the early-recovery cohort that was disproportionately repeat-buyers in higher-price markets. By 2025 the gap has narrowed, but HP still runs about 2 pp higher share than HR in the most recent quarters.

Operational takeaways: originators, capital markets, and policy

💼 For originators and LO marketing teams: update the applicant persona

Stop pitching HomeReady and Home Possible exclusively as max-leverage, 97% LTV first-time-homebuyer programs. The post-2023 mean LTV has dropped to 85%, and repeat buyers now represent 26% of program originations (up from 21% pre-shock). Position these programs as LLPA-advantaged conventional financing for income-eligible buyers who are stretching on income (mean DTI ~42%) and looking to minimize upfront loan-level pricing cost, not just leverage.

Geographically, reallocate outreach toward the midwest and lower-Northeast — IN, IA, MO, OH, WI, NE, CT are all up meaningfully; MD, MN, SD, WV remain top-share states. Pull budget away from coastal and mountain-west boom markets (ID, UT, AZ, WA, MA, NV, DC) where AMI caps no longer align with median home prices.

📊 For capital markets and MBS investors: model post-2023 vintages separately

Recent-vintage HR + HP pools carry a fundamentally different collateral risk profile than 2018–2021 vintages. With mean LTVs at 85% (down from 92%) and heavier concentration in midwest heartland states, these borrowers hold materially more equity at origination. Expect faster prepayment response during rate rallies — lower closing-cost drag and more extractable equity — alongside potentially lower default severity given the equity cushion, partially offset by higher DTI stretch.

Secondary desks should not apply pre-COVID baseline CPR/CDR curves to 2023–2025 production. MSR pricing on recent-vintage HR/HP is coming off a materially different cash-flow shape than legacy vintages, and needs vintage-specific assumptions rather than book-average ones. The aggregate ~35% volume decline hits MSR economics roughly proportionally with the overall GSE purchase market — the compositional shift is where the alpha is.

🏛️ For housing policy analysts and program stakeholders: evaluate high-cost-area income caps

The 10–13 pp share drop in coastal and mountain-west markets (ID −12.9, DC −12.3, UT −10.6, AZ −10.6, WA −10.2, MA −10.0, NV −9.4) demonstrates that the standard 80–100% AMI income limits have effectively priced qualifying buyers out of rapid-appreciation markets. The program’s automatic geographic self-adjustment shows the design is working as intended — but policymakers weighing high-cost-area income-limit expansion face a real trade-off. Expanding AMI ceilings restores coastal access, but risks diluting the program’s core low-to-moderate-income focus by shifting a share of program capacity to households well above traditional program cutoffs.

A quieter but related observation: repeat-buyer growth (79% → 74% FTHB) is not necessarily a leakage of intent. Repeat buyers still meet income limits — they just aren’t first-time buyers. If the goal is affordable homeownership among LMI households broadly, they qualify; if the goal is specifically FTHB, the target-population share has drifted 5 pp. The data makes the trade-off visible; which framing is the “right” one is a design question.

Methodology and caveats

Data sources and cleanup: - Fannie side: mortgage.fnm_sfp_raw (Single-Family Loan Performance monthly file), deduped to one row per loan_id at first appearance. HR identified via special_eligibility_program = 'H'; HFA Preferred via = 'F'. - Freddie side: mortgage.fre_slld_origination_raw (Single-Family Loan-Level Dataset, origination file — the full 1999–2025 Freddie tape). HP identified via program_indicator = 'H'; HFA Advantage via = 'F'. Not STACR — STACR is the CRT-issued 10% subset, which biases toward high-LTV loans and would inflate HP share estimates. SLLD is the correct denominator for a share-of-market analysis. - Universe: primary-residence purchase-money originations, both GSEs. Refi (rate-and-term and cash-out), investment properties, and second-home loans are all excluded. - Origination timing: Fannie’s origination_date is in MMYYYY format (the prior in-house schema notes had this as CCYYMM — we corrected mid-preflight). Freddie SLLD lacks an explicit origination date, so we use first_payment_date (CCYYMM) as a 1–2 month-lagged proxy. Quarterly rollups are robust to that lag. - Fannie servicing survivorship: fnm_sfp_raw is a monthly active-book file, so the dedup captures loans active in at least one snapshot in our load window. Older-vintage counts are marginally understated by paid-off attrition; for 2020+ vintages this is negligible. - Flag coverage: 100% on both GSEs across every vintage 2018–2025. No missing-flag adjustment needed. - Rate overlay: mean quarterly note rate is computed directly from the same loan-level origination data (Fannie original_interest_rate + Freddie original_interest_rate). A public PMMS series would track this closely but is not currently loaded into the warehouse; the note-rate proxy is sufficient for the rate-environment overlay shown in Chart 1. - State geographic analysis: states with fewer than 500 conforming purchase originations in either 2019 or 2024 are excluded from the comparison to avoid small-cell noise. - Cohort composition analysis: FTHB share is computed as Y / (Y + N) — loans with the flag left blank are excluded from the denominator rather than being treated as “no.”

Not covered in this analysis, and possible follow-up work: - Post-shock HFA program breakout. HFA Preferred (Fannie) and HFA Advantage (Freddie) volumes and shares track the same U-shape as HR/HP but are much smaller (0.5–4% of the market). A follow-up piece could examine whether the HFA sub-cohort behaves the same way, or whether state-HFA structural differences produce a different post-shock trajectory. - Prepay-speed comparison of recent-vintage HR/HP pools vs 2018–2021 vintages. We have Freddie MBS pool-level data now (fre_mbs_pool_details, fnm_mbs_pool_details). A pool-level composition check joined against per-CUSIP prepay factors is a natural extension. - Default divergence by state. Heartland vs coastal HR/HP loans will have different loss profiles going forward. Worth revisiting in 12–18 months when the 2023–24 vintage HR/HP cohort has enough performance data.

Chart 1 is a Plotly line chart showing HR + HP share as a solid line with the mean origination note rate as a dotted overlay on a secondary axis. Chart 2 is a side-by-side horizontal bar comparison of the top 10 state gainers vs top 10 state losers. Chart 3 is a small-multiples line chart showing mean LTV, FTHB share, and mean DTI drift across three cohort windows.

Data pulled 2026-08-02. Methodology and codified rules are available on request; the underlying preflight is in 2026-07-10-hr-hp-growth-trajectory-preflight-spec.md.