mortgagetape^™

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

Our first HR/HP piece documented that HomeReady + Home Possible held their share of the GSE conforming purchase market through the rate shock while the borrower profile shifted toward more equity and less first-time-buyer concentration. Our second piece showed a 62% take-up gap — most income-eligible borrowers land in standard conventional financing rather than HR/HP. This third piece asks a different question: who actually originates these loans? The answer surprised us. Between 2018 and 2024, Wells Fargo went from 18.5% of all HR + HP purchase share to 1.2% — a 17-point collapse tied to Wells’s post-2022 retail-mortgage exit. Rocket Mortgage rose to become the #1 named seller. UWM emerged from essentially nothing to 9.2%. And the specialization index — a seller’s share of HR/HP divided by their share of the overall conforming market — spans 13x, from Freedom Mortgage at 0.14 (basically no HR/HP participation) to Rocket at 1.80 (nearly twice the HR/HP share their overall size would predict). The seller map of affordable lending has been quietly redrawn.

📌 Executive takeaways by role

• Originators and channel-strategy teams: Concentration is diverging. Overall conforming HHI dropped from 1504 (2020) to 966 (2024) as UWM ate share and the market fragmented. But HR + HP HHI held at 1150–1320 across the same period. HR + HP is now 353 HHI points more concentrated than the overall market — a specialty product delivered by a narrower set of sellers. IMB retail dominates (Rocket, CrossCountry, Guaranteed Rate); correspondent aggregators (PennyMac, AmeriHome, NewRez) systematically under-index. Jump to originator takeaway.
• Capital markets and MBS investors: The compositional shift inside HR + HP pools matters for pool selection. Rocket-serviced HR + HP CUSIPs will look different from Wells-legacy HR + HP CUSIPs (which are shrinking in the seasoned book but still material in the older vintages). Between 2018 and 2024 the top-15 seller list turned over meaningfully — five names are new to the list, four are gone. Vintage-specific seller composition is now a first-order pool-selection variable, not a background detail. Jump to capital-markets takeaway.
• Program stakeholders and policy analysts: The CRA hypothesis — that HR + HP is a bank product used for CRA credit — is not confirmed by the data. Depositories under-index HR + HP as a share of their overall lending (JPMorgan Chase is the exception; Wells, U.S. Bank, and Truist all under-index). The programs are increasingly IMB-retail-delivered. That has implications for community-development policy and for CRA enforcement — if the programs designed to serve LMI borrowers are being delivered mostly by non-banks that aren’t subject to CRA, the regulatory model may need to update to match the delivery pattern. Jump to policy takeaway.

Concentration held while the market fragmented

The mechanical starting point is the Herfindahl-Hirschman Index — the sum of squared market shares, scaled to 10,000 points. HHI under 1,500 is unconcentrated. 1,500 to 2,500 is moderately concentrated. Above 2,500 is highly concentrated. We compute it separately for the HR + HP subset and for the overall conforming purchase market, per year.

📈 HHI trajectory: HR + HP vs overall conforming, 2018–2024

The chart below tracks two HHI series year-by-year. HR + HP is the solid rust line. Overall conforming is the dashed slate line. Both sit in the “unconcentrated” band, but HR + HP has diverged upward from overall conforming since 2022.

Key finding: HR + HP concentration held or rose while overall conforming concentration fell. In 2024, HR + HP is 353 HHI points more concentrated than the market it sits inside — the biggest gap in the seven-year series.

Two patterns matter:

First, the overall conforming market has fragmented since 2020. HHI dropped from 1504 (2020, when the top-3 sellers plus the OTHER bucket dominated) to 966 in 2024. That decline reflects the UWM-led broker-channel expansion, the rise of a broader set of mid-tier IMBs, and the deliberate retreat of some large depositories from mortgage retail. A market with an HHI of 966 is genuinely competitive by DOJ standards.

Second, the HR + HP subset did not follow that fragmentation path. HR + HP HHI has actually risen since the 2022 trough — from 1109 to 1319 in 2024. That means the same market forces that fragmented the broader conforming market did not fragment HR + HP delivery. If anything, they concentrated it. A smaller set of sellers is doing the HR + HP work than would be predicted by market-wide dynamics.

The mechanism is composition. When Wells Fargo exited most of its mortgage retail business post-2022 (following the CFPB consent order, product-line reductions, and the “we’re focused on relationship-based lending” pivot), a large chunk of its HR + HP volume didn’t reallocate proportionally to hundreds of smaller sellers. It reallocated to the sellers specifically positioned to serve income-eligible borrowers — namely Rocket, CrossCountry, and JPMorgan Chase. That’s a re-concentration into a different set of sellers, not a broad diffusion.

The top-15 has been redrawn

Comparing the top-15 HR + HP sellers in 2018 vs 2024 makes the compositional shift concrete:

📊 Top-15 HR + HP sellers, 2018 vs 2024

The chart below shows the union of top-15 sellers from either year. Stone bars are 2018 share; rust bars are 2024 share. Sellers are sorted by their 2024 share. The aggregated “OTHER” and “OTHER SELLERS” buckets (~45% of the market combined) are excluded to keep the chart about identifiable individual sellers.

Key finding: Wells Fargo dropped 17.3 percentage points (18.5% → 1.2%). Rocket Mortgage (formerly Quicken Loans) grew from 8.5% to 10.7% and is now the #1 identifiable HR + HP seller. UWM went from ~2% (2018 legacy under United Shore) to 9.2% (2024) via broker-channel expansion. CrossCountry, NewRez, and PHH entered the top-15 in 2024 having been well below it in 2018.

Three specific sellers deserve callouts:

Wells Fargo — the 17-point exit. Wells was the largest single HR + HP seller in 2018 at 18.5% of the market. By 2024 it was down to 1.2%. This is the mortgage-retail retreat playing out at product-mix granularity — Wells’s post-2022 strategic pivot away from broad-market originations landed disproportionately on HR + HP because that’s a labor-intensive retail product with lower gain-on-sale than standard conforming. Wells’s HR + HP portfolio still exists in the seasoned book, but the origination flow has effectively stopped.

Rocket Mortgage — the direct-to-consumer capture. Rocket rose from 8.5% (2018, as Quicken Loans) to 10.7% (2024). This is not just Rocket taking share proportionally to their overall growth — their specialization index in 2024 is 1.80, meaning their HR + HP share is 80% higher than their overall conforming share would predict. Rocket’s direct-to-consumer marketing engine converts eligible borrowers into HR + HP applications at a much higher rate than typical retail LO channels do. This is the modern equivalent of what Wells did in 2018 — one seller building a national HR + HP business at scale, just through a different channel.

UWM — the broker-channel opening. UWM was essentially not present in HR + HP in 2018 (its United Shore legacy DBAs together were maybe 4% of the market). By 2024 UWM is at 9.2%. That share was built almost entirely through the broker channel, where UWM dominates wholesale conforming. UWM’s specialization index is 1.00 — dead-on proportional to their overall size — meaning UWM is treating HR + HP as a normal part of their product mix rather than a specialty focus. The volume comes from UWM’s overall broker-channel dominance, not from any preferential HR + HP focus.

The 13x specialization spread

Beyond the top-15, the more revealing view is the specialization index across the top-40 sellers by conforming volume — how much each seller over- or under-indexes on HR + HP relative to their overall size in conventional conforming purchase.

📊 Specialization index — over-indexers (teal/rust) and under-indexers (gray/bronze)

The chart below shows the top-10 over-indexers (specialization greater than 1.0) and top-10 under-indexers (less than 1.0). Colors mark lender type — teal for depositories, rust for IMBs (both retail and wholesale), warm gray for IMBs on the under-indexer side, bronze for homebuilder captives.

Key finding: Specialization spans 13x. Rocket, Wells Fargo, CrossCountry, Citizens Bank, JPMorgan Chase — all over-index. Freedom Mortgage (0.14), Nationstar/Mr. Cooper (0.27), Lennar (0.27), DHI Mortgage (0.30), U.S. Bank (0.32), and PennyMac (0.38) are the biggest under-indexers.

Three patterns emerge:

IMB retail over-indexes. Rocket, CrossCountry, Guaranteed Rate, Guild — all IMBs that deliver primarily through retail (direct-to-consumer or branch-network) channels — over-index on HR + HP. These sellers have built application-flow systems that route eligible borrowers into affordable programs, and they market the programs to consumers who then ask for them by name.

Correspondent aggregators systematically under-index. PennyMac (0.38), AmeriHome (0.62), NewRez (0.77), Freedom Mortgage (0.14), LoanDepot (0.60), Mr. Cooper (0.27) — all major correspondent aggregators that purchase closed loans from smaller originators — deliver HR + HP at less than half the rate you’d expect from their overall market presence. The mechanism is that the underlying correspondent originators (small IMBs, community banks) don’t specialize in HR + HP, so the aggregators end up with a low HR + HP mix even though they’re huge in overall conforming. This is a direct explanation for part of the take-up gap we documented in the previous piece.

Homebuilder captives under-index by construction. Lennar (0.27), DHI (0.30) — the captive mortgage arms of national homebuilders — under-index on HR + HP because their buyer profile (new-home purchasers) tends to skew higher-income than the resale purchase market. New-home construction is concentrated in higher-price-point markets, and the median new-home price runs 15-20% above the median existing-home price. That pushes the buyer-income distribution above the 80% AMI eligibility ceiling for many purchases.

Depositories are surprisingly mixed. JPMorgan Chase (1.30) is a clear over-indexer — chase.com/mortgage runs HR + HP as a promoted product line. Citizens Bank (1.41) and Truist (1.06) also over-index. But Wells Fargo’s 2024 spec-index is 1.76 based on a tiny 2,320-loan denominator — a residual-book number, not an active-origination number. U.S. Bank at 0.32 under-indexes materially. The depository picture is one of a few active participants (Chase, Citizens) plus a lot of banks that have quietly stepped back from this segment.

Channel mix — retail-heavy on HR + HP

Concentration among sellers is only half the story. The channel a seller uses to originate matters too. HR + HP has been consistently retail-heavier than the standard conforming market:

• 2018: HR + HP retail 58%, correspondent 31%, broker 11%. Standard conforming retail 49%, correspondent 42%, broker 9%. HR + HP was 9pp more retail-heavy.
• 2024: HR + HP retail 61%, correspondent 21%, broker 18%. Standard conforming retail 48%, correspondent 38%, broker 15%. HR + HP is now 13pp more retail-heavy than the broader market.

Correspondent channel share on HR + HP has DROPPED from 31% (2018) to 21% (2024) even as it rose on standard conforming (42% → 38% is a decline but a smaller one). Broker channel share on HR + HP grew from 11% to 18% — most of that is UWM’s broker-channel HP expansion.

The retail-heavy tilt is the same signal as the specialization index — HR + HP is delivered by sellers with direct-to-consumer or in-branch capability, not by wholesale-focused aggregators. That’s why the specialization spread is so wide: sellers whose model matches HR + HP’s requirements (retail application flow, LO training on affordable-program workflows, active marketing to eligible borrowers) capture disproportionate share, while sellers whose model doesn’t (correspondent aggregation, wholesale intake without underwriting training on HR + HP specifics) systematically miss it.

Cross-link to articles 1 + 2: the mechanism completes further

• Article 1 established that HR + HP share held ~15% of conforming purchase through the rate shock, with a compositional shift toward higher-equity repeat buyers.
• Article 2 showed a 62% take-up gap — most income-eligible conventional purchase borrowers don’t end up in HR + HP.
• Article 3 (this piece) adds the delivery-side mechanism: the take-up gap concentrates in the correspondent-aggregator and homebuilder-captive channels that don’t route eligible borrowers into HR + HP. The 350K missing loans per year are systematically originated at PennyMac, AmeriHome, NewRez, LoanDepot, Freedom, Lennar, DHI, and similar sellers where the correspondent originator or captive-lender pipeline doesn’t push HR + HP.

The full three-piece narrative: HR + HP is stable in share, geographically rotated to where eligible borrowers live, structurally under-reaches its eligible pool, and concentrates delivery at a narrowing set of retail-focused sellers.

Operational takeaways: channel strategy, capital markets, and policy

💼 For originators and channel-strategy teams: address channel-specific drops

HR + HP delivery is increasingly a retail-dominated game (61% retail in 2024 vs 48% in broader conforming).

• For retail IMBs and direct-to-consumer lenders: you are already over-indexing (Rocket, CrossCountry, Guaranteed Rate all maintain specialization indices above 1.0). You can widen this lead by automating AMI qualification at the pre-qual stage to capture the 62% take-up gap in your active footprints.
• For correspondent aggregators and wholesale lenders: your specialization indices sit between 0.14 and 0.77 because your underlying correspondent originators default to standard conventional financing. To recapture this volume, introduce HR + HP routing incentives, offer LO training playbooks to correspondent partners, or establish dedicated internal desks to review borderline AMI applicants.

📊 For capital markets and MBS investors: factor seller turnover into pool selection

The seller makeup of HR + HP pools underwent a complete transformation between 2018 and 2024. Five new names entered the top-15 seller list while legacy leaders like Wells Fargo saw their origination share collapse from 18.5% down to 1.2%.

Because direct-to-consumer IMB paper (e.g., Rocket) exhibits different prepayment and servicing behavior than traditional bank or correspondent paper, secondary desks must treat seller composition as a primary pool-selection variable. Do not apply 2018–2021 pre-COVID prepayment/default models to 2023+ HR/HP CUSIPs without adjusting for this seller and channel shift.

🏛️ For housing policy analysts and program stakeholders: update policy beyond bank CRA

The traditional view that HR + HP functions primarily as a bank CRA compliance tool is contradicted by loan-level data. With depositories systematically under-indexing and IMBs now delivering the majority of HR + HP volume, policy efforts to expand LMI credit must adapt to current market structure.

Rather than relying solely on bank CRA enforcement, policymakers and GSEs should target the correspondent-aggregator delivery bottleneck by introducing reporting requirements or delivery incentives for aggregators. Furthermore, policy teams must recognize the concentration risk inherent in having a single seller (Rocket) account for over 10% of total national HR + HP originations.

Methodology and caveats

Data sources: - Fannie side: mortgage.fnm_sfp_raw, deduped to one row per loan_id at first appearance. seller_name_clean used (normalized upstream). HR flag = special_eligibility_program = 'H'. Origination year via SUBSTR(origination_date, 3, 4) (MMYYYY format). - Freddie side: mortgage.fre_slld_origination_raw (loan-level origination file). seller_name normalized with UPPER(TRIM()). HP flag = program_indicator = 'H'. Origination year via SUBSTR(first_payment_date, 1, 4) (CCYYMM). - Universe: primary-residence purchase-money originations, both GSEs. seller_name IS NOT NULL required (100% coverage on 2024 HR + HP; slight coverage loss on older Freddie vintages). - Channel field is R (retail) / C (correspondent) / B (broker), 100% coverage on both GSEs.

HHI computation: - Per-year, aggregate loan counts by seller across both GSEs. - Share = seller’s loan count / year total. - HHI = Σ (share × 100)² across all sellers. - Applied separately to (a) HR + HP subset and (b) overall conforming purchase.

Specialization index (Test 4): - Per seller per year: HR + HP loan count and overall conforming loan count. - Spec = (HR + HP share of program market) / (overall share of conforming market). - Restricted to sellers with 5,000 or more overall conforming purchase originations in the year to avoid noise from very small sellers.

Caveats: - Seller-name normalization is imperfect. UWM and its United Shore Financial DBAs appear as multiple string variants; Rocket and Quicken Loans span the 2018→2021 rebrand. We merged these on display but the underlying strings are separately reported by the GSEs and required manual mapping. - “OTHER” and “OTHER SELLERS” buckets (25% + 20% ≈ 45% of the 2024 HR + HP market) are pre-aggregated by Fannie and Freddie for sellers below their public-disclosure threshold. HHI computed on these opaque buckets treats them as two large “sellers” — which understates true concentration if the OTHER bucket contains many small independent sellers each with under 1% share. This is an inherent limitation of GSE public-file disclosure. - Bank/non-bank taxonomy is manual. We classified the top-15 sellers by well-known industry identity (bank subsidiaries vs IMBs vs homebuilder captives). Without a per-loan LEI or agency code, we cannot programmatically classify the OTHER buckets. HMDA panel data (agency_code by LEI) would give a proper join, but is not currently loaded. - Wells Fargo residual analytics. Wells’s 2024 HR + HP volume (2,320 loans) reflects the small remaining retail mortgage business and residual portfolio activity. The 1.76 specialization index is not representative of any active strategy — it’s a small-denominator artifact of a large exit.

Follow-up work: - Load HMDA panel for programmatic lender-type classification. Would enable a proper bank/non-bank share time series across all sellers rather than just the top-15. - Per-vintage prepay analysis by seller composition. MBS pool-level data (fnm_mbs_pool_details, fre_mbs_pool_details) exposes seller name at CUSIP level. Cross with prepay factors from fnm_mbs_bcpr for vintage-specific composition-adjusted CPR curves. - The correspondent-aggregator take-up gap by originator. If HMDA panel is loaded, we can identify which correspondent originators route eligible borrowers into HR + HP and which don’t — a natural fair-lending / CRA-relevant follow-up.

---

Chart 1 is a dual-line HHI trajectory with concentration-band shading. Chart 2 is a paired horizontal bar comparison of 2018 vs 2024 HR + HP share by seller. Chart 3 is a specialization-index bar chart with lender-type colored callouts.

Data pulled 2026-08-02. Preflight spec at 2026-07-10-hr-hp-channel-concentration-preflight-spec.md.