mortgagetape^™

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

📌 Takeaways by role

• Secondary marketing / rate-sheet operators: Insurer bids pay a durable +130bp acquisition concession for Rocket-dominant GNMA pools vs same-vintage, same-coupon peer paper — and that premium is still visible on today’s marks (+101bp FV spread). The pay-up is small vs the broad investor base but non-zero, and it’s lender-specific.
• Capital-markets / MBS investors: Freedom extracted the biggest acquisition premium (+148bp) and gave it all back on today’s marks (−84bp FV spread). The acquisition premium amortized into a loss, consistent with Freedom’s reputation as the fastest-prepay book on the FHA tape. PennyMac is the one top-5 issuer whose pools insurers bought at a discount (−74bp vs peers).
• Originator strategy: Brand-based pay-ups on agency paper are supposed to be commoditized away. They aren’t — not uniformly. Rocket’s name earns a real, sustained concession; Freedom’s doesn’t; PennyMac’s is negative. The size of the concession and whether it survives seasoning are both lender-specific.

The question

Agency MBS is supposed to be a commodity. Same coupon, same vintage, same agency wrap — same price. The TBA market exists precisely because sophisticated buyers treat Fannie 5% 2024 as fungible with any other Fannie 5% 2024.

But insurers don’t buy TBA. They buy specified pools and hold them. So insurer-held CUSIPs let us test whether pool-level pricing actually clears at parity — or whether the buy-side is quietly paying a concession for some issuer brands over others.

We ran that test on GNMA paper, using NAIC Schedule D Part 1 holdings (year-end 2025) joined to Ginnie pool CUSIPs we could attribute to specific issuers via loan-level issuer_name data.

How the comparison works

For every CUSIP held by at least one of 13 flagship life-insurer filers at year-end 2025, we:

1. Identified the pool’s dominant issuer (highest UPB share by issuer at Sep-2026 llmon2 snapshot).
2. Classified each pool as Rocket-dominant, UWM-dominant, PennyMac-dominant, Freedom-dominant, Lakeview-dominant, Carrington-dominant, or Other.
3. For each pool, we have the insurer’s actual_cost (price paid at acquisition), book_adjusted_carrying_value (today’s amortized basis), and fair_value (today’s market mark).
4. Within each 100bp coupon × vintage-year cell, we computed the par-weighted average acquisition price for the “Other” cohort — the peer benchmark.
5. For each top-5 issuer, we compared their cohort’s par-weighted price against that peer benchmark in the same cell (requiring at least 3 non-top-6 peer CUSIPs so the benchmark isn’t noise).

The resulting spread in points is the “pay-up” (or discount): how much more or less insurers paid for a given issuer’s paper vs structurally similar paper from a smaller issuer.

The result

Issuer	Par matched	N CUSIPs	Acquisition spread	Today’s FV spread	Prepay & value outcome
Freedom	$5.4M	84	+1.48 pts	−0.84 pts	Premium burned by refi churn
Rocket	$9.4M	51	+1.30 pts	+1.01 pts	Durable premium, slow prepay
Carrington	$0.7M	19	+0.87	+0.88	Execution-neutral stability
Lakeview	$9.1M	44	+0.30	+0.27	Par-equivalent commodity
PennyMac	$1.1M	10	−0.74 pts	−0.60 pts	Collateral discount (LTV / DPA)

Three distinct patterns across the top-5 issuers. UWM’s insurer-held paper is too thin ($0.57M par at year-end 2025) for a meaningful benchmark match at any sensible bucket size.

What’s happening

Rocket — brand pay-up that holds. Insurers paid 130bp over peer benchmark for Rocket-dominant pools at acquisition. The premium is still worth 101bp on today’s fair-value marks. In other words: insurers bought it at a premium, held it, and the premium didn’t decay. Most likely mechanism — Rocket’s pools prepaid slower than the peer cohort (higher credit, lower LTV, less streamline-refi churn). When convexity-drag doesn’t show up, the acquisition premium amortizes slowly and survives to the current mark.

Freedom — biggest premium, biggest fade. Freedom extracted the largest acquisition premium of any top-5 issuer (+148bp), but it has turned into an −84bp FV loss. The premium did not survive seasoning. This is consistent with the market read that Freedom runs the most aggressive FHA streamline-refi book in the top-5: fast prepay burns a purchase premium rapidly, because every paid-off dollar of par returns at 100 against a cost basis above 100. The acquisition premium is realized as negative carry over time. Investors who bought the Freedom premium got a lender brand that under-delivered on behavior.

PennyMac — the discount. Insurers paid 74bp below peer benchmark for PennyMac-dominant pools. The sample is small (10 CUSIPs, $1.1M par) but the direction has held across every bucket-size sensitivity we ran. The most likely explanation: PennyMac’s pool composition skews toward collateral characteristics insurers systematically discount — higher LTV, more DPA (down-payment-assistance) loans, more lower-FICO FHA — even within the same coupon/vintage window. The brand isn’t the signal; the embedded collateral is.

Lakeview and Carrington — commodity paper. Essentially no pay-up or discount. These are execution-neutral in the insurer bid.

Mechanism — why Rocket’s premium survives

Three forces drive whether a purchase premium holds or decays:

1. Prepay behavior. Pools that pay faster than the cohort amortize their premium down to par faster, realizing the premium as negative carry. Pools that pay slower let the holder keep collecting the above-market coupon on an above-par book. Rocket’s FHA/VA book is known for slower-than-peer prepay — likely a mix of borrower credit quality, lower share of streamline eligibles, and tighter servicing recapture.

2. Default roll-off. GNMA pools don’t take credit losses because the agency wrap covers principal. But a pool with higher defaults experiences buyouts, which prepay-equivalent-drag the premium. Rocket’s lower FHA default rate vs the Ginnie cohort helps here too.

3. Pool-level collateral gaming. Rocket’s pools may systematically select for characteristics insurers like (higher FICO within FHA, lower DPA share, lower-LTV manufactured, specific geography). Freedom’s pools, in the other direction, over-concentrate the streamline-refi collateral that drives fast prepay.

For an insurer paying up 130bp at acquisition, the question is whether the lender’s pool behavior will give them 130bp of extra running yield + mark-to-market stability over the holding period. Rocket delivers; Freedom doesn’t.

Caveats

• Panel size. Our parser covers 13 flagship insurer filers. Top-tier filers we don’t yet have: MML, JHUSA, LLIC, Principal, Guardian. Their buying behavior could shift the aggregate picture.
• Small matched universe per issuer. Matched par ranges from $0.7M to $11.5M across the 13-insurer panel — the finding is directional, not census-scale, and should be read as a signal about GNMA specified-pool trade pricing rather than a precise estimate of every lender’s clearing level.
• Point-in-time snapshot. Insurer holdings are disclosed year-end. The acquisition price is the insurer’s own cost basis, not necessarily the pool’s issuance price. A pool bought seasoned in 2023 at 94c reflects the 2023 secondary-market level, not the issuance clearing level.
• Dominant-issuer attribution. Many Ginnie pools are multi-issuer. We attributed each pool to its top-UPB-share issuer, which maps ~85% of par cleanly but blurs multi-issuer pool economics.
• Vintage mix. The matched universe is seasoned paper (mostly pre-2015 vintages). For newer vintages the top-6 issuers dominate so hard that peer-benchmark cells thin out.

What this changes

For originators: the brand-pay-up signal is real but highly lender-specific. A smaller originator trying to extract Rocket-level pay-ups needs the collateral and prepay behavior that justifies it, not just the brand.

For MBS investors: the message is to think about where the acquisition premium is going. A premium paid for a lender with fast prepay is a premium that won’t survive. A premium paid for a lender with slower, stickier collateral is a premium that collects running yield on top of face coupon.

For the market: this is one more place where “all agency paper is commoditized” turns out not to be literally true. The insurer bid — small as a share of the TBA market — is quietly assigning different concessions to different brands, and the concessions are working through to realized P&L in different directions.

---

Methodology note. Source: NAIC Schedule D Part 1 insurer holdings at December 31, 2025, joined to Ginnie Mae loan-level issuer attribution from llmon2 (September 2026 snapshot) via pool_id → CUSIP. Dominant-issuer classification by par-weighted UPB share. Benchmarks computed within each 100bp coupon × vintage-year cell, excluding all top-6 issuer-dominant CUSIPs from the benchmark universe. Par-weighted spread aggregation across matched cells. The matched universe is 48-111 CUSIPs per issuer and $0.7M-$11.5M of par — a sample, not a census.