mortgagetape^™

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

This article was written in collaboration with Optimal Blue, an Official AI Partner of Mortgage Tape, using Optimal Blue’s Competitive Data License Plus (CDL+), which covers approximately 35% of locked-and-hedged U.S. residential production.

📌 Executive takeaways by role

• Capital markets & private-label MBS investors: The DSCR-over-GSE-investor note-rate spread peaked at 54 bp in Q3 2025 and sits at 35 bp quarter-to-date in Q3 2026, an ~18 bp move in four quarters (roughly 4–5 bp per quarter). If the trajectory continues, DSCR execution reaches parity with GSE-investor pricing by mid-2027. Either reading — market maturation (bullish for the private-label channel) or unsustainable competitive pricing (warning) — treats the trajectory as the leading indicator. Jump to capital-markets takeaways.
• Non-QM aggregators & DSCR originators: Moving beyond the “46 bp average” to matrix-based pricing gives originators a sharp competitive edge. The real picture is a matrix (26 bp to 109 bp across FICO × LTV cells) and it’s compressing. Route your originator conversations around the cells you compete in, not the cross-cell average. Watch which cells are compressing fastest to see where the next repricing wave hits. Jump to originator takeaways.
• MSR investors & secondary marketing: DSCR MSR valuations anchored to 2025 spread assumptions are already stale. The compressing premium implies tighter prepayment behavior converging with agency-investor patterns; strip pricing needs to move with it. Composition is not steady — Investor-DSCR grew from ~28% to ~34% of the non-QM total over the same window, so part of the observed compression may be a composition effect and part is a real price signal. The next 2–3 quarters will separate the two. Jump to MSR takeaways.

The premium is a matrix, and it’s shrinking

Ask a non-QM aggregator what a DSCR loan costs relative to an agency investor loan today and you’ll get one number: forty-six basis points. Here is exactly what that number is and how it was computed, so any reader can reproduce it:

• Field: note_rate (30-year fixed lock rate) from Optimal Blue Competitive Data License Plus lock records.
• Populations: DSCR = loan_type='NonConforming' AND dscr_ratio IS NOT NULL. GSE investor = loan_type='Conforming' AND occupancy='Investment Property'.
• Computation: For each FICO × LTV cell (5 × 5 grid) with n ≥ 500 DSCR locks in the 2026 window, take median DSCR note rate minus median GSE-investor note rate. The 46 bp headline is the simple average of those cell-level rate spreads.

This is a rate spread, not a price adjustment. It is expressed in basis points of note rate, not in bp of upfront price (LLPA). Rate spreads and price spreads are related — at par pricing, 1 bp of rate ≈ 4-5 bp of price for a 30-year fixed — so at 46 bp of rate the implied price spread is roughly 2 to 2.5 points. The whole article works in rate terms.

It’s a defensible number. It’s also a lazy one. Averaging across the FICO × LTV grid erases both the range that actually matters and the direction the market is moving. The real picture: the DSCR premium spans 26 bp to 109 bp across the cell grid, it peaked at 54 bp in the third quarter of 2025 on the workhorse cell, and it sits at 35 bp quarter-to-date in the third quarter of 2026. That is an 18-basis-point move over four quarters. Neither the range nor the compression shows up in a single average number, and both matter more than the average does to any desk that actually trades this paper.

The matrix

Slice the same lock population by FICO band and LTV band and the flat 46 bp number becomes a grid that behaves in a specific way. High-LTV cells run 55 to 109 bp regardless of credit tier. Low-LTV cells run 30 to 40 bp regardless of credit tier. The credit-score effect within a fixed LTV band is real but modest — worth roughly 10 to 15 bp between the strongest and weakest FICO cohorts at the same leverage. The LTV effect is dominant.

That pattern is not what you’d predict from an agency-investor mental model, where LLPA moves as much on FICO as on LTV — the classic 720 / 75 grid dependency the standard risk-based pricing framework was built around. On DSCR, the non-QM market is telling you that leverage matters much more than credit. The mechanism is intuitive on reflection: DSCR underwriting is fundamentally property cash-flow-based, and higher-leverage loans have thinner cash-flow coverage margins, tighter DSCR ratios (see the ratio distribution below), and greater sensitivity to rent-market shocks. Credit score is a secondary underwriting input at best — a filter for reserves and payment discipline, not the driver of loss content the way it is on owner-occupied.

For an originator quoting a specific file, this matters. A 780 FICO / 80% LTV DSCR loan doesn’t get the cross-cell average premium; it gets a cell-specific premium closer to 70–80 bp. A 720 FICO / 65% LTV loan doesn’t either; it gets 30–35 bp. Quoting the matrix instead of the average is the difference between winning the file and losing it.

The compression

The workhorse cell (720–759 FICO, 65–75% LTV, the deepest-liquidity slice of the DSCR grid) sat at a 54 bp premium in the third quarter of 2025 and sits at 35 bp quarter-to-date in the third quarter of 2026. That is roughly 4–5 basis points per quarter of compression over the past year, sustained, without a policy trigger or a rate-cycle inflection to point at.

Three distinct structural drivers are pressing down on the premium at once.

Capital deepening. The private-label DSCR securitization market has grown materially through 2025 and 2026 as more capital sources — insurance-owned issuers, credit-fund-sponsored aggregators, bank-backed platforms — entered the space. More competing bids at the pool level flow directly into originator rate sheets.

Underwriting maturation. The 2020–2022 DSCR programs paid a premium that partly reflected uncertainty about credit performance. Three years of vintage seasoning have delivered performance data credible enough for the market to price closer to actuarial expected loss rather than an uncertainty spread. The premium is compressing toward its true credit content, not disappearing.

Competitive pressure on aggregators. The DSCR-focused non-bank aggregators are competing hard for a well-defined originator base, and their pricing has moved. This is the compression path most likely to reverse if any single aggregator retrenches — the “unsustainable competitive pricing” case that a bearish reader should hold in mind.

Which of the three is dominant matters for whether the compression continues. If it’s the first two, DSCR execution converges further toward agency-investor pricing over the next 12–18 months and the private-label market becomes a much more competitive alternative to agency delivery. If it’s largely the third, expect a partial rebound as aggregators re-price for margin. The Optimal Blue quarterly panel over the next 2–3 quarters will tell us which — and we’ll publish the follow-up when the data supports it.

Composition is moving — and it matters for the compression story

Segmenting the non-QM lock population by Optimal Blue’s own income-verification taxonomy — the same buckets OB publishes in its Market Advantage report — the shares in Q2 2026 look like this: Investor-DSCR ~33%, Bank Statement ~15%, All Other ~52%. In OB’s headline Market Advantage segmentation (which folds some further categorization we don’t fully replicate here) the DSCR and Bank Statement buckets sit close to parity in the 30% range. Either framing produces the same headline: DSCR is a plurality of non-QM, not a supermajority.

The share of non-QM that is Investor-DSCR has also moved. From 28.5% in 2025-Q1 the DSCR share climbed to 33-34% by 2026-Q1 and has held there through Q2. That is a real 5-point rise in share over five quarters. That mix change matters for how the premium-compression story is read:

• If DSCR share had been flat, the 18 bp compression would be an unambiguous price signal on comparable populations.
• Since DSCR share grew, part of the compression could reflect composition — as the DSCR pool grew, the marginal loan entering may have been a stronger file than the average, mechanically pulling the median rate down.

The good news for the price-signal story: the compression shows up within matched FICO × LTV cells (Chart 2 uses the 720-759 / 65-75 cell specifically), so the cell-level compression is not a mix artifact within that cell. It could still be a mix artifact at the cell level if the FICO or LTV distribution within a cell shifted — but the effect from within-cell shifts is bounded (a 5-point FICO shift within a 40-point band moves the median rate by ~1-2 bp at most). Our read: the majority of the 18 bp is price, some is composition, and the exact split is what the next 2-3 quarters will resolve.

Sub-1.0 DSCR loans: 8-9% of the market trades cash flow for reserves

Contrary to the perception that DSCR is a strictly conservative credit product, close to one in ten loans closes with a ratio below 1.0. Across the 2025 and 2026 lock population, the median DSCR ratio hovers between 1.06 and 1.12, and 8.6% of DSCR loans lock with a DSCR ratio below 1.0 (populated ratios only). The property’s projected rental cash flow does not fully cover the mortgage payment on that slice. Not stressed, not deteriorating, and if anything drifting slightly down: the sub-1.0 share was 9.1% in 2025-Q1 and 7.9% in 2026-Q2.

This is not a market weakness signal, it is an accepted underwriting norm. Sub-1.0 DSCR loans compensate through other underwriting dimensions: larger reserves (typically 12+ months versus 6 months for stronger DSCR files), lower LTV, stronger FICO, or a mix. The GSE-investor equivalent would be a file that qualifies on borrower income and reserves rather than on the property’s cash flow alone; DSCR just makes the trade-off explicit in the ratio field.

The distribution around the floor

The pooled DSCR ratio distribution across 2025-2026 tells a specific underwriting story:

Nearly half of all DSCR locks (48.4%) fall in the 1.00-1.15 band — right above the underwriting floor. That is not what a random draw of investor properties looks like. It is what a market being sized to the floor looks like: originators, brokers, and borrowers work the loan amount, rent assumption, or fee structure until the file just clears the required DSCR threshold. Layered on that: 1.0% of locks fall below 0.75, 7.6% in the 0.75-1.00 band (together the “sub-1.0” 8.6% share), 20.1% in 1.15-1.30, 11.4% in 1.30-1.50, and 11.6% at 1.50 or above.

Two things matter for a credit-risk-focused reader. First, the pile-up at 1.00-1.15 is the segment most sensitive to rent-market softening or vacancy shocks — a 5-10% dip in effective rent flips a meaningful share of that band below 1.0 mid-cycle. That said, the same 1.00-1.15 cohort tends to carry lower leverage on the property side: average LTV in this band sits in the 65-70% range, materially below the DSCR book’s overall LTV distribution, so there is a real equity cushion protecting against outright default even when debt-service coverage tightens. Second, the definitional caveat: DSCR calculations vary across originators (gross rents versus net rents, market rent estimates versus in-place leases, PITI-only versus fully-loaded HOA and reserves), so the exact band boundaries are somewhat program-dependent. But the shape — mode right above the floor, thin left tail — is real regardless of which convention is used.

For MSR investors and private-label MBS analysts, the 1.00-1.15 cohort is the segment worth watching for stress-cycle behavior. If the compression thesis holds and the DSCR premium reaches parity with agency-investor pricing, the marginal capital pushing the compression may be less discriminating in the thin margin above the floor than the current issuer base has been. That is where the next credit story lives.

The rate ladder

Grounding the “premium” abstraction in what borrowers actually see: on 2026 lock data, the median rate ladder runs GSE owner-occupied at 6.50%, GSE investor at 6.72%, DSCR across all credit tiers at 7.05%, and sub-680 FICO DSCR at 7.63%. That’s a 22 bp step from owner-occupied to agency investor, another 33 bp from agency investor to all-DSCR, and a further 58 bp from all-DSCR to sub-680 DSCR.

Two useful reference points for a reader benchmarking rates. First, the DSCR-to-GSE-investor step (33 bp on the all-DSCR median) is smaller than the cell-average premium on the matched-cell analysis above (46 bp) because the DSCR book on average carries lower LTV than the GSE-investor book. DSCR files sit heavier in the 60–75 LTV range; GSE-investor sits closer to 75–80. On matched cells the premium is materially wider than on the raw median comparison. Second, the sub-680 DSCR tier at 7.63% is the segment where the “premium” concept gets stress-tested. That cohort pays for both the DSCR product and the credit-tier premium, and whether the stack compresses at the same rate as the workhorse cell is the next question the quarterly data will answer.

Operational takeaways: pricing intelligence, watch signals, and what compression means for each desk

📊 For capital markets & private-label MBS investors: the compression trajectory is the leading indicator

The 18 bp compression over four quarters is either capital deepening (bullish for private-label pool pricing and issuance capacity), competition-driven (warning of a snap-back), or partly composition (as the DSCR pool grew, the marginal loan may have been a stronger file). Which mix dominates is knowable within 2–3 quarters. Watch whether the compression continues at 3–4 bp/quarter or whether any single quarter reverses. If it continues at trend and the composition drift stalls, DSCR pricing reaches parity with GSE-investor by mid-2027 and the case for private-label DSCR as a full alternative to agency-investor execution strengthens materially. Pool-level pricing models anchored to a fixed “DSCR premium” number should convert to a compressing curve, and any within-cell analysis of the compression should isolate the composition vs price split at the FICO/LTV grid level.

⚙️ For non-QM aggregators & DSCR originators: quote the matrix, not the average

The 46 bp average premium is what your competitors are quoting; the matrix from 26 to 109 bp is what actually determines whether you win the file. Route your rate-sheet conversation around the specific cell — a 780 FICO / 80% LTV file is a different price than a 720 FICO / 65% LTV file, and the difference is 40+ bp. Also worth naming to your originator base: the compression is real. Any file locked today is priced 15–20 bp tighter than the same file locked a year ago. That is a story your top originators can take to their referral partners, and a competitive-differentiation opportunity for aggregators that update their pricing intelligence quarterly. Finally, the pile-up at 1.00-1.15 DSCR ratio means most files are being sized to the floor — the next competitive wave may be programs offering a modest DSCR-ratio tolerance (accepting 0.95-1.00 with reserve offsets) rather than another 5 bp of rate.

💼 For MSR investors & secondary marketing: strip valuations need to move with the premium

DSCR MSR valuations anchored to 2025 spread assumptions are already stale. The compressing premium implies convergence toward GSE-investor prepayment behavior — as pricing tightens, refinance triggers activate at smaller rate moves, and prepay speeds converge with agency-investor curves. Composition is not steady: Investor-DSCR share of non-QM has grown from ~28% to ~34% over the same window, so any premium-tracking analysis should verify that composition drift isn’t inflating the observed compression. Servicers running DSCR strip books should mark to a compressing curve rather than a fixed spread, and the sub-1.0 DSCR ratio cohort (8-9% of the book) plus the 1.00-1.15 pile-up (48% of the book) are the two segments worth stress-testing for the marginal-credit tail that the next capital wave into DSCR may absorb.

What we know, what we don’t, and what’s next

The 18 bp compression in the DSCR premium over 2025 into 2026 is a durable finding across the four quarters where Optimal Blue’s DSCR-tagged data has held at scale, and it is real within the matched FICO × LTV cell that Chart 2 tracks. The matrix behavior (LTV-driven more than FICO-driven) is stable across the same window. The 8-9% sub-1.0 DSCR ratio cohort — and the 48% pile-up in the 1.00-1.15 band above the floor — are stable underwriting norms, not deterioration. Those three things we know.

What we don’t know: which of the compression mechanisms (capital deepening, underwriting maturation, competitive pressure) is dominant, and how much of the aggregate compression is composition-driven versus real price. That determines whether the trajectory holds or partially reverses, and it is the single most important question for anyone pricing DSCR paper going forward. The next 2–3 quarters of Optimal Blue data will separate the mechanisms; we’ll publish the follow-up when the compression trajectory has enough quarters to distinguish trend from cycle.

The pattern to watch: if the within-cell compression continues at 3–4 bp per quarter through the end of 2026 and into 2027, the DSCR-to-GSE-investor premium reaches parity by mid-2027 and the private-label investor market crosses a genuine structural threshold — pricing convergence that changes the case for delivery across a large swath of investor origination. If any single quarter reverses meaningfully, the aggregator-competition explanation gains weight and the trajectory looks more cyclical than structural.

Either way, the answer will show up first in the Optimal Blue quarterly panel. This is the piece that establishes the baseline; the next piece measures the trajectory against it.

---

Methodology. Lock-level pricing data sourced from Optimal Blue’s Competitive Data License Plus (CDL+) service via Mortgage Tape’s data lake, covering approximately 35% of locked-and-hedged U.S. residential production. All spreads and premiums in this article are computed on note_rate (30-year fixed lock rate) — they are rate spreads in basis points, not price adjustments in bp of upfront price. DSCR = lock records with loan_type='NonConforming' AND dscr_ratio IS NOT NULL. GSE investor = loan_type='Conforming' AND occupancy='Investment Property'. Coverage caveat: Optimal Blue’s DSCR-tagged coverage reached scale in the first quarter of 2025. The primary analysis window is Q1 2025 through Q2 2026 (six full quarters plus partial Q3). Prior-year quarters (2024) appear in Chart 2 for continuity but reflect thin DSCR samples (n = 3 to 445 per quarter) and should be read as directional only. Rate comparisons are median 30-year fixed lock rates within each product bucket. The matched-cell premium comparison controls for FICO band and LTV band, comparing DSCR to GSE-eligible investor locks within the same cell. DSCR ratio bands are computed on populated positive ratios only (dscr_ratio > 0 AND dscr_ratio < 5). Non-QM composition segmentation in Chart 3 uses OB’s income_verification_type field mapped to a three-bucket taxonomy (Investor-DSCR = Investor - DSCR; Bank Statement = any %Bank Stmt%; All Other = everything else including Full Doc, Asset-Related, alt-doc, no-doc) that approximates OB’s Market Advantage report. Cell counts in the matrix analysis range from ~500 to ~7,400 per cell in the 2025–2026 window; cells with n < 500 are excluded from the matrix. Rate ladder aggregate rates cited in Chart 5 reflect the full 2025–2026 lock window. Composition, ratio, and rate figures are computed from lock volume; loan-level performance data is not included in this analysis. Informational, not advice.