The path of long-end rates determines MBS coupon selection, while extension risk in newly issued RMBS is concentrated in specific tranches

Overview

The report analyzes coupon selection in U.S. agency MBS, the composition of overseas demand, and extension risk in non-agency RMBS through the lens of interest-rate sensitivity. Its core conclusion is that whether Treasury buybacks can sustainably improve MBS performance depends on whether they lower long-end yields; some of the relative value in lower coupons has already been realized, while certain tranches of newly issued non-QM and CES transactions face clear maturity-extension risk in a high-rate environment.

Core views

After the Treasury unexpectedly announced that it would double the size of its long-duration Treasury buybacks, the MBS basis tightened modestly on Wednesday, with belly and lower coupons outperforming higher coupons; however, some of that performance was retraced after rates rose again on Thursday and Friday. The report argues that simply increasing buybacks is insufficient to address the underlying causes of rising long-end yields. The recent increase in long-end rates has been driven simultaneously by higher term premiums arising from changes in the Federal Reserve's communication approach and policy uncertainty, market concerns about the Fed's credibility, widening fiscal deficits, rising global long-end yields, and increased investment-grade corporate bond supply related to artificial intelligence investment. Consequently, the high sensitivity of MBS to rate volatility may persist. Coupon selection depends on the precise path of long-end rates. If proactive Treasury measures can constrain long-end yields in the near term, production coupons and higher-coupon FN 5.5s and 6.0s may outperform; if further measures cause long-end yields to decline decisively, belly coupon FN 5.0s are expected to have an advantage. Conversely, if the 10-year Treasury yield rises further and exceeds 4.8%—5.0%, extension concerns and outflows from fixed-income funds could cause higher coupons to underperform. The report compares the average monthly excess returns of the 30-year conventional MBS index across different rate environments: August through November 2025 represents the declining-rate period, while May through July 2026 represents the rate-rebound period. The results show that belly and lower coupons are more favorable when rates decline, while production and higher coupons are more favorable when rates are stable or rise modestly. Lower-coupon FN 2.0s—2.5s have significantly outperformed since the report's July 24 recommendation and rank among the stronger performers in the coupon stack as measured by excess returns on the 30-year coupon index. Because their spreads have already tightened substantially, reducing their valuation appeal, Goldman Sachs recommends returning to neutral from its previous stance rather than continuing to chase relative performance that has already materialized. Overseas demand has improved from 2025. U.S. Treasury International Capital data show that overseas investors purchased a net $51 billion of agency securities at face value in 1H 2026, the best first-half performance since 2023. Most major holding regions were net buyers except Canada, mainland China, and South Korea; however, the vast majority of net demand came from common hedge-fund domiciles such as Luxembourg, Bermuda, the British Virgin Islands, the Bahamas, Ireland, and the Cayman Islands, and therefore may not represent genuine end-investor overseas demand. Excluding those regions, Hong Kong, Taiwan, and Singapore were the principal buyers. Japan purchased a net $1.4 billion in 1H 2026, significantly below the $9.6 billion recorded for full-year 2025. To understand the sources of Japanese demand, the report uses Japanese Ministry of Finance data on net purchases of overseas long-term bonds by investor type as a proxy. These data cover all non-Japanese long-term bonds rather than agency MBS alone, so they provide only directional indications. The data show that Japanese trust banks, which primarily reflect pension-fund activity, have consistently been net buyers of overseas fixed income in recent years. The report believes that rising equity markets have rapidly increased the value of pension funds' domestic and overseas equity allocations, requiring periodic rebalancing into domestic and overseas bonds to maintain target weights, with some of those funds potentially flowing into agency MBS. By contrast, Japanese insurers have consistently been net sellers of overseas long-term fixed income since 2020, mainly because currency-hedged agency MBS yields have become less attractive relative to domestic bonds; after modest net purchases in 2025, Japanese banks shifted to net selling in 2026 year to date. Bank demand was stronger in mid-2025, when MBS spreads were approximately 20 basis points wider than currently. Broker-dealers remain steady net buyers, although some transactions may represent executions on behalf of other investors. For RMBS, the Freddie survey shows that the 30-year mortgage rate rose from 6.0% in February 2026 to nearly 6.7% in early August, while non-QM mortgage rates may rise with a lag. Higher rates reduce borrowers' willingness to refinance or move, slowing prepayments and extending the average life of securities, which forces investors to hold potentially lower-yielding assets for longer. RMBS are also affected by transaction call and coupon step-up provisions: issuers may call a transaction if the mortgage-pool balance falls to a specified threshold or the transaction reaches a minimum seasoning period, typically three years; if it remains uncalled four years after issuance, the coupon on senior tranches typically steps up by 100 basis points. New securities are therefore generally priced to a month-48 call, but issuers may decline to exercise the option if rates rise substantially, leaving investors exposed to significant extension risk. The ultimate impact also depends on tranche seniority and whether the principal waterfall is sequential or pro rata. The weighted-average coupons of newly issued mortgage pools have not risen in tandem with market mortgage rates, and because of the lag in pooling assets, recently issued transactions may contain loans with rates below prevailing market levels. They are therefore more susceptible to extension than transactions originated in 2024—2025. The current forward curve implies that the 10-year Treasury yield and mortgage rates may rise by 60 basis points over the next four years, suggesting that newly issued bonds face an extension risk of more than 40%. Using spot rates and defining extension as a 100-basis-point increase in mortgage rates over the next four years, the probability is approximately 25%. The report emphasizes that the rate level four years from now is what truly determines non-call risk. Unless mortgage rates rise extremely sharply in the near term, extension risk remains low for 2023 transactions that will become callable over the coming months. Tranche comparisons show that the impact of extension is most pronounced beginning with non-QM M1, CES A2, and their more subordinate tranches. Non-QM transactions distribute principal pro rata among senior tranches before allocating it to lower-rated tranches; at new-issue pricing, the weighted-average life is approximately two years for senior tranches and four years for subordinate tranches, primarily reflecting the assumption of a call in year four. Without the call, under current rate incentives, the fully extended weighted-average life of senior tranches is expected to jump to slightly more than 6.5 years, while that of lower-rated tranches would exceed 20 years. However, the spread between M1 and A1 remains only approximately 50—60 basis points, suggesting that M1 buyers may be more focused on maximizing yield than on extension risk. CES uses a sequential principal waterfall, making the impact of extension most pronounced in the A2 tranche. In typical new-issue pricing, A1 has a weighted-average life of approximately 2.5 years and lower-rated tranches approximately four years; under a fully extended scenario, A1 extends only modestly, while A2 nearly doubles to approximately eight years. Senior non-QM tranches issued one year ago currently trade from the high $98s to the low $99s, but the report believes the discount primarily reflects the sharp year-to-date increase in short-end yields rather than worsening extension risk. More seasoned transactions have higher weighted-average coupons, resulting in lower calculated extension risk than newly issued transactions. Although the lower weighted-average funding costs of transactions issued in late 2025 create uncertainty, the report judges that issuers may place greater importance on maintaining stable, programmatic issuance and call practices than on saving approximately 40 basis points in execution costs; the outcome will ultimately depend on the rate level approximately three years from now. The forecast table reflects a base case of relatively high rates and a wider agency MBS basis: the 10-year Treasury yield is projected at 4.40% at year-end 2026, above 4.17% in 2025; the 30-year fixed mortgage rate is projected at 6.45%, above 6.18%. The current-coupon nominal mortgage basis widens from 102 basis points to 110 basis points, Treasury OAS widens from 15 basis points to 25 basis points, while the non-QM AAA spread tightens from 125 basis points to 120 basis points. Total 1—4 family residential mortgage originations are projected to rise from $2.050 trillion to $2.156 trillion, including an increase in purchase mortgages from $1.356 trillion to $1.445 trillion and an increase in refinancing from $694 billion to $711 billion, although the refinancing share declines from 34% to 33%.

Analysis framework

The report first identifies the sensitivity of MBS to changes in interest rates by examining market performance before and after the Treasury's long-duration Treasury buyback announcement, and then constructs a conditional coupon-selection framework by comparing coupon-index excess returns across different rate regimes. It subsequently combines U.S. Treasury International Capital data with Japanese Ministry of Finance investor-category data to break down the geographic and investor sources of overseas demand. The RMBS section starts with prepayment incentives, transaction call provisions, the 100-basis-point coupon step-up, tranche seniority, and principal waterfalls, using forward rates, spot rates, and fully extended weighted-average-life simulations to compare extension risk across newly issued and seasoned transactions, different products, and different tranches.

Methodology notes

• Event Strategy and Behavioral FinanceEvent-driven analysis

  Analysis of the Treasury's long-duration Treasury buyback event

  The report observes the immediate performance of the MBS basis and individual coupons after the announcement that buyback volumes would double, as well as the subsequent retracement, to determine how the policy event affects MBS through long-end yields.

• Fixed Income and Credit AnalysisDuration/Convexity Analysis

  Comparison of rate sensitivity across coupon tiers

  The report combines the effective duration and convexity of each coupon with excess returns across different rate periods to explain how the relative performance of different MBS coupons changes when rates decline, remain stable, or rise.

• Fixed Income and Credit AnalysisYield curve analysis

  Using the forward curve to estimate future rates and extension probability

  The report uses the change in rates over the next four years implied by the forward curve to assess the likelihood that transactions will not be called and compares this with probability estimates based on spot rates.

• Fixed Income and Credit Analysis

  Fully extended weighted-average-life simulation

  Using expected prepayment speeds under current implied rate incentives, the report simulates the fully extended WAL of each product and tranche when a transaction is not called and compares it with the WAL assumed at new-issue pricing.

• Fixed Income and Credit AnalysisSpread analysis

  Comparison of yield differentials across tranches and markets

  The report compares the spread between M1 and A1, the currency-hedged yield on agency MBS with the 10-year Japanese government bond yield, and current MBS spreads with those in mid-2025 to explain investor demand and risk pricing.

• Corporate Fundamentals and Financial Frameworks

  Proxy analysis of investor fund flows

  Because Japanese Ministry of Finance data do not separately identify agency MBS, the report uses net purchase trends in all overseas long-term bonds by different investor types as a directional proxy for Japanese demand for MBS.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Agency MBS FN 2.0s—2.5s

  They have significantly outperformed since the July 24 recommendation, but the report recommends returning to neutral after spreads tightened substantially.

  Strengths

  They generated some of the strongest excess returns in the coupon stack under the recent rate environment.

  Weaknesses

  Their valuation appeal is now lower than before.

  Comparison

  They have generally outperformed higher-coupon MBS since the July 24 close.

  Risks

  Return potential may be limited after further spread tightening.

• Agency MBS FN 5.0s

  The report expects relative outperformance if further policy measures drive long-end yields lower.

  Strengths

  They are belly coupons expected to benefit in a declining-rate environment.

  Weaknesses

  Performance is highly dependent on an actual decline in long-end rates.

  Comparison

  They are expected to outperform production and higher coupons in a declining-rate scenario.

  Risks

  If buybacks fail to lower yields, the expected advantage may not materialize.

• Agency MBS FN 5.5s and 6.0s

  They may relatively outperform if long-end yields remain constrained and rates are stable or rise modestly.

  Strengths

  They have a relative advantage in stable or moderately rising-rate scenarios.

  Weaknesses

  They are sensitive to extension concerns and outflows from fixed-income funds.

  Comparison

  They are expected to underperform FN 5.0s when rates decline decisively.

  Risks

  They may underperform if the 10-year Treasury yield exceeds 4.8%—5.0%.

• Newly issued non-QM M1 and subordinate tranches

  The pro rata principal waterfall and the assumption of a call in year four expose them to significant extension risk.

  Strengths

  M1 offers approximately 50—60 basis points of additional spread over A1.

  Weaknesses

  If not called, WAL can rise to slightly more than 6.5 years for senior tranches and exceed 20 years for lower-rated tranches.

  Comparison

  Extension risk is higher than for seasoned transactions with higher GWACs.

  Risks

  A significant future increase in rates or the issuer's failure to exercise the call option.

• Newly issued CES A2 and subordinate tranches

  The sequential principal waterfall causes the impact of extension to increase materially beginning with A2.

  Weaknesses

  Under full extension, A2 WAL rises from approximately four years to approximately eight years.

  Comparison

  A1 extends only modestly under a fully extended scenario, while A2 nearly doubles.

  Risks

  Persistently high rates causing the transaction not to be called.

• 2023 and other seasoned non-QM transactions

  Their higher GWACs result in lower current extension risk than recently issued transactions.

  Strengths

  They are approaching call eligibility and have relatively low calculated extension risk according to the report.

  Weaknesses

  Prices remain affected by the sharp year-to-date increase in short-end yields.

  Comparison

  Extension risk is lower than for recently issued transactions with lower GWACs.

  Risks

  An extremely sharp near-term rise in mortgage rates or rate levels approximately three years from now that are significantly higher than expected.

Key data

• Overseas net purchases of agency securities in 1H 2026$51 billionAt face value, the best first-half performance since 2023
• Japanese net purchases of agency securities$1.4 billion in 1H 2026; $9.6 billion in full-year 20251H 2026 was significantly below full-year 2025
• 30-year mortgage rateNearly 6.7% in early August6.0% in February
• Four-year rate change implied by the forward curve60-basis-point increase in the 10-year Treasury yieldThe report uses this to assess extension risk in newly issued RMBS
• Extension risk for newly issued bondsMore than 40%Based on the current forward curve
• Extension probability based on spot ratesApproximately 25%Using a 100-basis-point increase in mortgage rates over the next four years as the extension condition
• Fully extended WAL for non-QMSlightly more than 6.5 years for senior tranches; more than 20 years for lower-rated tranchesAt new-issue pricing, approximately two years for senior tranches and four years for subordinate tranches
• Spread between non-QM M1 and A1Approximately 50—60 basis pointsThe report believes the spread remains tight and that investors may be more focused on maximizing yield
• Fully extended WAL for CES A2Approximately eight yearsAt new-issue pricing, lower-rated tranches have a WAL of approximately four years, nearly doubling under full extension
• Prices of seasoned senior non-QM tranchesHigh $98s to low $99sThe report primarily attributes this to the year-to-date rise in short-end yields rather than extension risk
• Potential execution-cost savings for issuersApproximately 40 basis pointsRelated to uncertainty over whether transactions issued in late 2025 will be called as scheduled
• Forecast for the 10-year Treasury yield at year-end 20264.40%4.17% in 2025
• Forecast for the 30-year fixed mortgage rate in 20266.45%6.18% in 2025
• Forecast for the current-coupon nominal mortgage basis110 basis points102 basis points in 2025
• Forecast for current-coupon Treasury OAS25 basis points15 basis points in 2025
• Forecast for the non-QM AAA spread120 basis points125 basis points in 2025
• Forecast for 1—4 family residential mortgage originations in 2026$2.156 trillion$2.050 trillion in 2025
• Forecast for the refinancing share in 202633%34% in 2025

Impact & implications

The report argues that relative MBS performance cannot be judged solely by the size of Treasury buybacks; the key is whether policy genuinely lowers long-end yields. Falling rates are more favorable for FN 5.0s, while a stable or modestly rising-rate environment may be more favorable for FN 5.5s and 6.0s. However, a rise in the 10-year Treasury yield above 4.8%—5.0% would amplify extension and fund-outflow pressures on higher coupons. Although overseas demand has improved, its ultimate sources are not fully transparent; within RMBS, newly issued low-GWAC mortgage pools, non-QM M1, CES A2, and subordinate tranches are the most sensitive to future rates and whether transactions are called.

Risks

• If the 10-year Treasury yield rises further and exceeds 4.8%—5.0%, extension concerns and outflows from fixed-income funds could cause higher-coupon MBS to underperform.
• Increased Treasury buybacks alone do not address the drivers of rising yields, including term premiums, fiscal deficits, rising global long-end rates, and investment-grade corporate bond supply.
• A significant rise in rates over the next four years or issuers' failure to exercise call options could substantially extend the average lives of newly issued non-QM M1, CES A2, and subordinate tranches.
• The tight spread of approximately 50—60 basis points between M1 and A1 may not fully reflect extension risk.
• The lower funding costs of transactions issued in late 2025 may weaken issuers' incentives to call them, with the ultimate outcome depending on rate levels approximately three years from now.

What to watch

• Watch Chair Warsh's speech at the Jackson Hole symposium next week.
• Watch whether Bessent and the U.S. Treasury announce further Treasury buybacks or fiscal-consolidation measures and whether those measures can lower long-end yields.
• Watch whether the 10-year Treasury yield approaches or exceeds 4.8%—5.0%.
• Track the path of mortgage rates over the next three to four years, changes in the forward curve, and call probabilities for newly issued transactions.
• Monitor whether the spread between non-QM M1 and A1 remains at approximately 50—60 basis points.
• Track changes in overseas fixed-income flows from Japanese trust banks, insurers, and banks.