Given this 25 bp hike flips the directional logic in our "Macro Implications Update." That note leaned on imminent cuts → faster CPR and a sell bias in high coupons. A hike does the opposite near-term:   • Prepay/convexity: CPR drifts down (rough guide: −1 to −3 CPR over the next couple prints, coupon-dependent). High-coupon convexity risk eases a bit; extension risk rises in 3.5–4.5s. Macro_Implications_Update_Revis…   • Positioning shift vs the doc: Trim overweight in 4.0–4.5 carry; move from "sell/underweight" toward neutral-to-buy in 5.0–6.0 cohorts (the refi surge the doc worried about is less urgent after a hike). Macro_Implications_Update_Revis… Buy vs. sell (TBA bias and price impact) Here’s a clean trading take you can act on today:   • 5.5–6.0s (FN/FR/G2): Tilt buy on weakness. Less near-term refi pressure, slightly friendlier convexity. Favor slower specs (low WALA, low balance, high geo dispersion) if payups don’t overreact.   • 5.0s: Selective buy/neutral. Good cushion, manageable convexity.   • 4.0–4.5s: Lighten/hedge into any bounce. These extend the most; carry stays fine, but price beta to rates dominates near-term.   • GNMA vs Conv: GNMA high-coupon less at risk than the note assumed; relative value improves if servicer/streamline fears fade. Rough dollar-price move from the 25 bp hike Rule of thumb (ignoring second-order convexity): ΔP ≈ −Duration × 0.25%. Typical effective durations right now:   • High coupons (5.5–6.0): Dur ~4.0–4.5 → −1.0 to −1.125 pts   • Middle (5.0): Dur ~4.5–5.0 → −1.125 to −1.25 pts   • Lower (4.0–4.5): Dur ~5.0–5.5 → −1.25 to −1.375 pts Convexity will claw back a few ticks for high coupons (less negative gamma) and add a few ticks of downside for low coupons (extension). Net: the sell-off is heaviest in 4.0–4.5s, lightest in 5.5–6.0s. What changed vs your attachment (net summary)   • The attachment’s call to underweight 5.5–6.0 due to refi/convexity is downgraded; hike tempers that risk.   • The attachment’s overweight 4.0–4.5 for defensive carry is less compelling right after a hike due to extension and worse price beta. AI Prepay Implications of +25 bp The hike flips the "cuts-soon" premise in the note: refi impulse recedes, CPR glides lower, low-coupon extension rises, high-coupon convexity improves. Given this 25 bp hike AI Term Sheet (Base / Upside / Downside)   • Base path (most likely, 55–60%)     • Rates corridor: +25 bp sticks; term premium a touch wider.     • CPR corridor: −1 to −2 CPR over 1–2 prints (coupon-dependent).     • MBS mechanics: Extension in 4.0–4.5; convexity relief in 5.5–6.0; payups stabilize.     • Desk bias: Hold/hedge 4.0–4.5; add 5.0–6.0 on weakness; GNMA high-coupon RV improves.   • Upside path (bullish MBS, 20–25%)     • Rates corridor: Hike seen as "one-and-done"; curve bull-flattens.     • CPR corridor: Drifts only ~−0.5 to −1 CPR; spec payups re-bid on carry.     • MBS mechanics: 5.0–5.5 lead; 6.0 workable but watch premium decay if rally extends.     • Desk bias: Overweight 5.0–5.5, neutral 6.0; re-enter clean 4.5 specifieds if payups gap wider then fade.   • Downside path (bearish MBS beta, 20–25%)     • Rates corridor: Another hike priced or long-end cheapens further.     • CPR corridor: −2 to −3 CPR; seasoning + burner screens stall.     • MBS mechanics: 4.0–4.5 extend hardest; dollar-prices gap lower; basis softens.     • Desk bias: Reduce 4.0–4.5 beta, keep only special stories; stay long 5.5–6.0 but trim if premium/rally risk re-emerges. NSA / MLA / Routing For AI Prepayment Model:   • NSA gating (native sparse attention): Narrow borrower/state buckets to the most rate-sensitive cohorts → cleaner CPR nowcasts after a regime flip; reduces false burner signals that your pre-hike logic emphasized.   • MLA capacity (piecewise-linear): Swap heavy GEMM segments for CSD-2 + MLA stencils to track kinked S-curve shifts around payment shocks (e.g., streamline/IRR thresholds).   • Corridor routing: Treat coupon×WALA×geo as nodes; route flows along "long-range corridors" (spec/servicer/burner corridors) with dynamic weights so basis/CPR shocks propagate realistically instead of uniformly. Trading Grid Adjust (overlay on your one-pager)   • 4.0–4.5 (FN/G2): Sell / Hedge. Extension > carry; only keep cheap special specs (low WALA, story pools).   • 5.0: Neutral → Buy. Best balance of carry vs beta; add on weakness with disciplined screens.   • 5.5–6.0: Buy / Buy on weakness. Near-term convexity tailwind; prefer slower specs (low balance, broad geo dispersion).   • GNMA 5.5–6.0: Buy / RV long. Streamline/recast fears fade with hike; watch servicer headlines. Given this 25 bp hike Hedge & Risk Notes (desk-practical)   • DV01 stub (per $100mm): approximate with effective duration × 10,000; hedge with ~$40–63mm 10y per $100mm across 15yr–30yr buckets (see grid).   • What to monitor this week:     1.Daily rate-lock/primary-secondary spread (refi impulse proxy),     2.Spec payups vs CPR nowcasts (NSA-screened cohorts),     3.Dollar-rolls on 5.0–6.0 (tight → lighten, soft → add),     4.GNMA streamline chatter vs TBA basis.

Global Flow Shock Overlay

Recent financial headlines highlight heightened systemic risk from global capital flows. Reports suggest the Federal Reserve has been forced into emergency interventions to prevent USD assets from flowing back into China, with prominent voices such as Ray Dalio warning that two key asset classes (U.S. Treasuries and equities) face collapse risks under stress conditions. Implications for the MBS complex include: • Fed credibility risk: Abrupt policy shifts reduce market confidence in orderly rate-cut transmission, amplifying volatility in both USD and Treasury markets. • Capital flow sensitivity: High-coupon pools (5.5–6.0%) are disproportionately vulnerable to global demand shocks. Reduced foreign bid could exacerbate convexity and liquidity risks. • Treasury stress feedback: If Dalio’s collapse scenario materializes, higher hedging costs and tighter dealer liquidity deepen convexity hot-spots already highlighted in the Fed & labor macro update. Portfolio Impact: The global overlay reinforces the Sell bias in high-coupon pools while strengthening 4.0–4.5% conventionals as the defensive anchor. Hedging strategies must explicitly incorporate both ΔRate shocks and exogenous flow/credibility shocks. AI-driven attribution provides critical separation, ensuring stable convexity management even under capital flight scenarios.

Macro Implications Update (Fed & Labor Data)

Recent macro developments materially shift the prepayment outlook relative to the base case outlined in the August Flash. Two key updates drive this change: • Fed Rate Cuts: High-level economists and market consensus now expect three Fed cuts before year-end (September, October, December), with the possibility of a larger 50bp move or consecutive cuts if labor weakness persists. • Labor Market Revision: The U.S. Bureau of Labor Statistics is expected to revise payrolls down by 600k–900k jobs and monthly job growth by 50k–75k. This signals sharper cooling in the employment backdrop. Implications for MBS prepayments: • Lower mortgage rates are likely to arrive sooner and more decisively than assumed in the base case. This elevates CPR risk beyond a modest uptick, particularly in 5.0–6.0% coupon cohorts. • High-coupon GNMA and conventional pools (5.5–6.0%) are now acute convexity hot-spots, with elevated refinance surge risk. • Burnout dispersion narrows: even seasoned 2018–2020 cohorts regain prepay capacity, compressing relative differences. • Lower-coupon 4.0–4.5% remain the defensive carry anchor, with increased relative advantage versus high-coupons. Portfolio/Trade Framing Adjustments: • Overweight: 4.0–4.5% conventionals and seasoned burnout pools for carry and duration stability. • Underweight: FNMA/FHLMC and GNMA 5.5–6.0% pools, where convexity and optionality risks are most acute. • Hedging: Expect stronger dealer short-TBA positioning in higher coupons; real-money accounts will anchor lower-coupon spreads as defensive carry. AI Impact on Prepayment Modeling The recent Fed and labor market developments amplify the value of the AI operator framework. Whereas traditional models would broadly shift S-curves upward in response to multiple Fed cuts, AI’s recursive operator chain and sparse attention mechanisms attribute CPR changes specifically to ΔRate shocks, avoiding overreaction in lower-coupon carry pools. Key combined effects: • Faster attribution: AI isolates Fed-driven rate cuts as the dominant lever, enabling earlier recognition of risk. • Granular activation: Sparse attention ensures that only rate-sensitive cohorts (e.g., GNMA 5.5–6.0 VA streamline) are activated. • Hedge stability: Adaptive stencils recalibrate hedge ratios locally, stabilizing convexity management even as spreads widen. Portfolio Implication: The macro shock strengthens the Sell bias in high-coupon pools while reinforcing 4.0–4.5 conventionals as the defensive core. AI integration makes positioning timelier, more precise, and more stable under volatile autumn conditions. AI Macro Shock: AI replaces a static S-curve with a recursive latent operator network. Adaptive stencils (by incentive buckets) and sparse cohort attention yield timelier, more granular buy/sell signals and steadier convexity hedges. A Recursive Latent Operator is a framework for building models that learn efficient, abstract representations of sequential data by recursively updating a hidden state using a learned function. It's a foundational concept behind most modern state-of-the-art sequence models, especially those designed for long-range dependencies, offering a powerful blend of efficiency, compression, and predictive power.

Agency MBS Prepayment Flash — Aug'25 (Sep'25 prints)

Conventional 4.0–4.5 remain the carry-defensive core (speeds ~high-single CPR); GNMA high coupons are the optionality hot-spot with MoM variance. AI lets you target exact sub-cohorts driving changes, improving both selection and hedging versus a single S-curve view.

• The portfolio strategy is to treat 4.0–4.5 conventional MBS as the anchor holdings—providing stable income and relatively resilient performance when the market swings, while other buckets (low coupons, high coupons, Ginnie Maes) are treated more tactically.

Top Takeaways

Speeds broadly in high-single to low-teens CPR; Ginnies run a touch faster than conventionals.

MoM changes small (±1–2 CPR across large cohorts); limited new refi incentive without bigger rate moves.

Seasoning & burnout dispersion: older 2018–2020 cohorts run ~1–3 CPR faster at like coupon.

GNMA high coupons (5.5–6.0) show greater variance (VA streamline/issuer effects).

Program Commentary

FNMA & FHLMC: 4.5s around 8–10 CPR across vintages, orderly behavior. 5.0–5.5s show modest uptick but no wave; turnover dominates incremental changes.

GNMA I: 5.0–5.5s run low-teens CPR with VA streamline optionality. New 6.0s show volatile prints with thin seasoning.

GNMA II: 6.0s mid/high-single CPR with MoM choppiness. Legacy 5.0–5.5s low-teens CPR with wide issuer dispersion.

Risk Themes

Fed-cut narrative vs. mortgage rates: small cut nudges CPR higher but no 2020-style wave.

Burnout: seasoned 2019–2020 cohorts more prepay-capable than newer vintages.

GNMA policy sensitivity: streamline eligibility swings CPR several points in small cohorts.

Seasonality: post-summer turnover fade should dampen CPR into autumn absent a rate shock.

Buy / Sell Framing

Buy / Overweight:

• Conventional 4.0–4.5 ('22–'23) carry-defensive.

• Seasoned burnout pools ('18–'20) for duration needs.

Sell / Underweight:

• GNMA 5.5–6.0 thin-seasoning cohorts (variance hot-spot).

• High-coupon conventionals near incentive threshold.

AI Impact on Prepayment Modeling

AI replaces a static S-curve with a recursive operator network. Adaptive stencils (by incentive buckets) and sparse cohort attention yield timelier, more granular buy/sell signals and steadier convexity hedges.

Key benefits:

Operator chain identifies which lever (ΔRate / Turnover / Credit / Policy) moved CPR.

Adaptive stencils provide local fit for incentive buckets, stabilizing hedge ratios.

Sparse cohort attention activates only relevant sub-cohorts (issuer/geo/vintage).

Generates targeted specified-pool screens and optimal TBA hedge pairs.

Source: FN/FH/GNMA flash files. Month: Aug'25 | Prepared for Sep'25 prints.

Macro Policy & Market Dynamics

Rate Cuts and Prepayment Dynamics

September Fed cut could lower mortgage rates and increase refi incentives, especially in higher-coupon pools (5.5–6.0%). “buy-the-rumor, sell-the-news” effect: if markets sell off after the cut, long-end yields may stabilize or steepen, capping refinance waves.

Equity Correction → Flight to Quality

Equity sell-off drives flows into Treasuries and MBS. Treasuries rally → mortgage rates lower → CPR rises. But MBS spreads may widen if liquidity is hit, raising convexity risk.

Buy vs Sell Trade Implications

Buy side: lower-coupon 4.5–5.0% MBS offer carry and stability; real-money accounts can anchor spreads. Sell side: higher-coupon ≥5.5% MBS face refi surge risk; dealers may short TBAs to hedge convexity.

Trade Framing

Buy: 4.0–4.5% lower-coupon, seasoned pools with limited incentive. Sell: 5.5–6.0% TBAs, where refinance exposure is most acute.

Summary: If a Fed cut combines with equity correction, prepayment risk rises. Positioning should lean to buy lower-coupon carry pools and sell/hedge higher-coupon TBAs.

This import collapse is recessionary in signal. Likely downward pressure on rates → MBS price appreciation in near term. Prepayment risk may rise modestly (especially in higher-coupon pools) if rates fall, but the bigger driver is spread tightening from lower rate expectations.

Buyer's Market Signs Lower prepayment speeds make MBS cash flows more stable → better yield predictability for buyers. Wider spreads between MBS yields and Treasuries due to policy uncertainty → buyers can lock in attractive yields. Sellers may have to offer price concessions because fewer investors want to take on mortgage credit and duration risk in a high-rate environment.

Here's the CPR (Conditional Prepayment Rate) probability curve mapping: Blue (Before CPI Imputation Spike): Higher prepayment expectations due to greater probability of rate cuts and refinancing activity. Red (After Spike): Lower prepayment expectations because Fed rate cuts are less likely in the short term, keeping mortgage rates elevated. Gray Shaded Area: The reduction in prepayment risk — this stability in cash flows benefits MBS buyers, reinforcing a buyer's market environment.

This favors a short-term BUY stance for MBS, especially in lower-coupon or new-production pools where convexity risk is lower.

Here's the updated chart with a third "Post-Rate-Cut Scenario" curve (green dot-dash). The green shaded area shows the rebound in prepayment risk if the Fed cuts rates, relative to the post-CPI baseline. This curve sits midway between the red dashed line (low CPR after CPI spike) and the blue line (high CPR before CPI spike), reflecting a partial re-acceleration of refinancing activity.

Left (5.5% Premium): CPR is very sensitive to CPI housing shocks. A ±1% housing swing (≈ ±35 bps Δ) shifts prepayment speeds materially.

Premium TBAs = CPI-linked convexity risk. Investors must manage hedge/carry around Fed/CPI surprises.

Right (3.5% Discount): CPR is flat/insensitive; discount pools remain stable regardless of CPI-driven rate changes.

Discount TBAs = safe harbor. Stable CPR even under housing CPI volatility → useful for portfolios seeking duration stability.