Mortgage Rates in 2027: A Quarterly Outlook

10-year = expected policy path + term premium

The New York Fed’s ACM model is our decomposition, not our forecast. On November 10 (ACM data through that day; NY Fed file updated November 12, 2026, reported 5:50 p.m. ET on Macrodatahub and consistent across Macronomy and Hidden Tax vintages):

A second official decomposition is easy to confuse with ACM. FRED’s THREEFYTP10 is the Kim-Wright (2005) three-factor term premium on a 10-year zero, not ACM. Its last observation is 0.8738% on November 6, updated November 10 at 2:04 p.m. CT — 22 basis points above the ACM print, on a slightly earlier day, from a different model. The 2026 year-to-date average is 0.76%, the highest annual average since 2009; October’s monthly average was 1.01%, with a peak of 1.10% on October 5. That is a mid-2000s premium (2005–07 annual averages: 0.75%, 0.70%, 0.68%), not a 2010s premium, when the same series spent years near zero or below. We keep ACM as the identity we can pair with a risk-neutral yield. We use Kim-Wright for the official history.

Friday’s 5.152% 10-year has no official ACM split. Our mid-2027 judgment — labeled as such — is that the expected-rate component settles near 4.05%–4.15%, as December’s hike proves the last and 2027 cuts come into view, and that the term premium settles near 55–65 basis points, a partial mean reversion from Kim-Wright’s 87 and from October’s 110. Add them and you are near 4.70%, not at 5.20%. The last time Kim-Wright averaged near 50 basis points for a full month was September 2024 (0.10%). The 2026 low, on the last day of February, was 0.46%, next to the 10-year’s 52-week low of 3.961% on February 27. We are not assuming a return to those lows. We are assuming a retracement of an elevated premium, not its disappearance.

30-year = 10-year + the mortgage spread

As of 2:18 p.m. ET on November 13, Mortgage News Daily had the average top-tier 30-year fixed at 7.50% and the 10-year at 5.151%. The spread is 235 basis points. Using the 3:00 p.m. ET Tradeweb close of 5.152% gives the same 235. Freddie Mac’s PMMS, a Thursday-to-Wednesday average released at noon ET on November 12, is 7.11% — stale relative to this week’s backup, and the basis on which MBA and Fannie forecast.

A 235-basis-point primary spread is wide against history, and history is the heart of the easing case. The annual average of weekly FRED MORTGAGE30US minus DGS10 was 169 basis points in 2010–19 and 161 in 2000–07. The post-2022 climate has been 242 basis points; 2023 averaged 285, with a weekly peak of 310 in the week of June 1. Mean reversion is already underway on a yearly basis: 252 in 2024, 231 in 2025, and 200 year-to-date through November 12. January 2026 already printed weekly spreads of 182 to 189 basis points. Friday’s 235 is a week’s sell-off, not the year’s climate. Our year-end assumption of 180 is a return to the top of the pre-2022 norm, not a heroic tight. Optimal Blue, at the end of October (Street Insider, November 10, 6:00 a.m. ET), had a 221- basis-point primary spread and a 108-basis-point primary-secondary gap. We did not obtain a clean, dated current-coupon option-adjusted spread we would defend, so we will not invent one. The ICE BofA MOVE index closed at 98.43 on November 12, up 6.49% (Investing.com), a 52-week range of 55.77 to 115.02. A MOVE nearer the middle of that range is one of the ways the option premium in the spread comes in.

What they assumed, and what happened

Three primary January documents are on the table: Fannie Mae’s Economic Forecast and Housing Forecast, both dated January 13, 2026, with interest-rate forecasts based on December 31, 2025; Wells Fargo’s U.S. Economic Outlook of January 14, 2026, 6:00 a.m. ET; and MBA’s Mortgage Finance Forecast of January 21, 2026. MBA’s January PDF publishes a 10-year and a 30-year path and does not publish a funds-rate, inflation, or term-premium path; we will not invent ones. A January 28 commentary by MBA’s chief economist, Mike Fratantoni, after that month’s FOMC, is cited separately and is not the January 21 forecast itself.

Term premium belongs in the diagnosis even though they did not write a number for it. The official FRED Kim-Wright 10-year premium was 0.571% on December 31, 2025, and 0.5855% on January 13. It peaked at 1.0965% on October 5 and last printed 0.8738% on November 6 — a rise of 30 basis points from the January 13 print, and of 51 at the peak (FRED THREEFYTP10, updated November 10 at 2:04 p.m. CT). The fiscal 2026 deficit was $1.992 trillion (final Monthly Treasury Statement, through September 30). Coupon auction sizes remain on hold at high levels (Treasury, November 4). Those are the ingredients of a higher term premium, and they were not in the January 4.2% 10-year.

Which assumption did the most damage

Split the 30-year miss into the 10-year and the spread, on each house’s own basis, against the fourth-quarter-to-date FRED averages of 5.09% and 7.28% and a 218-basis-point average weekly spread, through November 12.

The 10-year did most of the work. The spread’s contribution is real — Fannie’s 170-basis-point implied Q4 spread was the most optimistic, and it accounts for a larger share of Fannie’s miss — but in every case the Treasury yield is the majority. That 10-year miss has two sourced pieces, which we will not pretend to add to the last basis point: the policy path flipped from cuts to a hike, and the Kim-Wright term premium rose 30 basis points from mid-January, 51 at the October peak. Headline inflation and oil, running above the January paths, are why the Committee could hike. Core inflation, through the third quarter, is not a sufficient explanation on its own. The first-quarter 30-year calls were essentially exact. The error was assuming that a 4.2% 10-year and a 170-to-190-basis-point spread were a regime, not a January weather report.

Why the October consensus repeats the error, the other way

Having been too low, the same houses anchored to the autumn backup and wrote it forward. Fannie Mae’s October 12 Economic Forecast — interest rates based on September 30 — puts the 10-year at 5.3% in the fourth quarter of 2026 and in every quarter of 2027, the federal funds rate at 4.0% in the fourth quarter of 2026 and at a 4.4% average in 2027, and, in the companion Housing Forecast, the 30-year at 7.5% in every quarter of 2027. The implied spread is 220 basis points, all year. MBA’s October 11 forecast has the 10-year at 5.0% then 4.9%, the 30-year at 7.2% then 7.1%: an implied spread of 220 basis points as well. Wells Fargo’s October path, as reported by NerdWallet on November 12 — the primary October document not obtained — is 7.05%, 7.00%, and 6.95% through the third quarter of 2027. Goldman Sachs’s 6.4%, which we have only second-hand from an October 30 outlook, is the exception, and we have already said we will not claim originality against it.

That is the January mistake with the sign flipped. In January they treated a 4.2% 10-year and a tight spread as durable. In October they treat a 5.0% to 5.3% 10-year and a 220-basis-point spread as durable. Fannie’s 5.3% 10-year for all of 2027 is a straight-line extension of a September 30 tape that sat next to a 52-week high of 5.310% on October 5. Its 4.4% 2027 funds-rate average implies further hikes from today’s 3.75%–4.00% and then a hold, the opposite of what twelve months after a final hike has usually looked like on the 10-year: declines of 19, 85, 95, and 203 basis points in 2006, 2018, 2000, and 1995, on FRED daily data, with 2023 the exception at plus 41. The September hike, and a December hike already in the price, will still be working through the economy in 2027. Existing sales at 3.75 million, SLOOS GSE-eligible demand at −32.1, and 65.6% of mortgages below 5% are the lag, already visible. The 220-basis-point spread those 2027 paths embed ignores a 2010–19 average of 169, a 2000–07 average of 161, a 2026 year-to-date average of 200, and January weeks that already printed 182 to 189. Kim-Wright at 87 basis points is a mid-2000s high, not a floor; it has already come in 22 basis points from the October 5 peak.

Our base case — a 4.70% 10-year and a 180-basis-point spread — is the opposite choice: that December is the last hike, that term premium and the mortgage basis retrace rather than plateau at a stressed print, and that 45 basis points on the 10-year is a conservative reading of the post-hike record. The sticky 30% case exists because the October consensus could still be right, as 2023 showed. We give that view three chances in ten, not five, and we will be graded on the difference.

1. The spread has mean-reverted before, and has already started

From Friday’s 235 basis points to our 180 is 55 basis points — more than half of the 100-basis-point journey to 6.50%, and it does not require a recession. The 2010–19 average of 169 and the 2000–07 average of 161 are the norms we have in mind. The 2022–26 average of 242 is the climate we are leaving, not a law of nature. After the 310-basis-point weekly peak in June 2023, annual averages fell 54 basis points by 2025 and a further 31, on a year-to-date basis, in 2026. Ten weekly prints in 2022 and early 2026 already came in at or below 190, the last of them in the week of February 5, 2026, at 184. A 180-basis-point year-end spread is a completion of a reversion that is visible in the annual data, interrupted by this week’s volatility, not a guess from nowhere.

2. Term premium is high — and high premiums retrace

Kim-Wright at 0.87% on November 6, and at 1.10% on October 5, is elevated against the 2012–21 era and in line with the mid-2000s. Elevated premiums have retraced before. From late October 2023, when the monthly last print was 0.62%, the series was at 0.09% by the end of December — 53 basis points in two months (FRED THREEFYTP10). From this October’s 1.10% peak, it is already 22 basis points lower. Treasury’s November 4 refunding keeps coupon auction sizes unchanged for the next several quarters, which is a reason the premium should not collapse to zero. It is not a reason it cannot come in 20 to 30 basis points from a mid-2000s high. That is all our 55–65 basis-point 2027 assumption asks.

3. After hiking cycles end, the 10-year has usually fallen — and our 45 basis points are conservative

Monetary policy works with a lag, often cited in the 12-to-18-month range. The September 16 hike is already in that window for mid-2027; a December 9 hike, which we put in the base, would be as well. We are not pretending this hiking cycle has already ended. We are asking what twelve months after a last hike has looked like, and we computed it from FRED daily DGS10, retrieved November 13.

In the four completed cycles before 2023, the 10-year was lower a year later, by 19, 85, 95, and 203 basis points. The median of those four is a decline of 90 basis points; the mean is 101. Our base case asks for 45 basis points from Friday’s 5.15% — half the median, and less than every pre-2023 episode except 2006. The 2023 episode is the honest exception: the 10-year rose 41 basis points in the twelve months after the July 26, 2023 last hike. That is a reason our sticky case exists, not a reason to ignore the other four. A MUFG note on easing cycles since 1980 found a median 10-year rally of 91 basis points in the twelve months after the last hike (Chart of the Day, November 16, 2023). We prefer our own FRED arithmetic, and we flag that this cycle’s last hike is still ahead of us.

4. Housing demand is already weakening, and the lock-in wall is the mechanism

Existing-home sales ran at a 3.75 million seasonally adjusted annual rate in October, down 5.5% on the month and 8.8% from a year earlier, the slowest pace since August 2010 (NAR, November 12). Inventory was 1.59 million, 5.1 months; the median price was $416,700, up 0.4% year over year, the 40th consecutive monthly increase. The Housing Affordability Index stood at 100.7. This is what mid-7s rates do to a market in which most owners have no incentive to move.

Cumulative: 49.1% of outstanding first liens are at or below 4%; 65.6% are below 5%. Only 22.5% already sit at 6% or higher. Calculated Risk’s history of the same file: the under-4% share peaked at 65.2% in the first quarter of 2022. The average outstanding rate was 4.4% in the second quarter, against a Freddie Mac PMMS of 7.18% in the week of November 5 (Norada). The October SLOOS, released November 2 and covering the third quarter, 61 domestic banks, puts a number on demand: net demand for GSE-eligible purchase mortgages was −32.1 (53 banks), government −30.0, and every one of the seven purchase categories was negative (Federal Reserve, Table 1; FRED SUBLPDHMDENQ). Standards for GSE-eligible and government loans were unchanged, at 0.0 net tightening. Jumbo eased a little; subprime tightened. This is not a credit crunch for conforming loans. It is a rate-driven demand slump, milder than the freeze of late 2022 and early 2023 (−86.2 and −93.0 on the same GSE series) but clearly pointed the wrong way. A 6.5% 30-year does not unlock the 65.6% below 5%. It does start to unlock the 6% and higher cohort, and it takes the most painful edge off purchase affordability. That is how housing, with a lag, becomes an argument for easier policy rather than a sideshow.

5. Temporary help is still a fifth below its peak

Temporary-help payrolls were 2,526,000 in October, up 18,600 on the month and 3.0% from a year earlier — the highest such year-over-year gain since September 2022 (FRED TEMPHELPS, updated November 6 at 9:43 a.m. CT; Staffing Industry). We will not pretend the direction is down. The level, though, is 20.1% below the March 2022 peak of 3,161,400. That residual gap is a reminder that the staffing tail of the labor market never fully recovered. It is a supporting exhibit for an easing case, not a leading one, and it sits next to claims and JOLTS data that, as we say below, the bears have every right to cite.

Base — 50%. A glide to 6.5%.

December 9 is the last hike. January is a skip. The lagged effect of this year’s tightening, a still-weak housing market, and a mortgage spread that resumes its 2024–26 compression take the 10-year to 4.70% and the 30-year to 6.50% by December 2027, with the spread at 180 basis points. Inflation cools enough for the Fed to cut, not enough to look like 2020. Trigger: a year in which core CPI starts to print closer to 0.2%, housing stays soft, and term premium and the mortgage basis come in together.

Sticky — 30%. The bears’ world, and a mid-7s 30-year.

Shelter stays stuck, claims stay firm, and term premium holds in the 80–95 basis-point area or returns toward 100. The Fed hikes in January as well as December, then holds through 2027. The 10-year ends at 5.20%; the 30-year at 7.45%; the spread stays near 225. This is, in outline, the path a higher-for-longer house would write today. We give it three chances in ten, not five.

Deeper — 20%. A recession opens the door through 6%.

Unemployment moves through 4.5% with conviction, the Fed cuts more than a couple of times, and the 10-year falls toward 4.20%. The 30-year ends near 5.95%, with the spread at 175. Credit and volatility can widen the mortgage basis on the way down, so the 30-year may fall less than the 10-year. Trigger: a genuine labor-market break, not a soft housing market alone.

30-year fixed, by scenario

10-year Treasury, by scenario

History: 30-year mortgage and the 10-year, annual averages

The spread, after the quiet decade

Q1 2027 Ends March 31, 2027

Q2 2027 Ends June 30, 2027

Q3 2027 Ends September 30, 2027

Q4 2027 Ends December 31, 2027

Purchase comes first; refinance opens only a little

MBA’s October outlook had 2027 originations just under $2 trillion, down 5%, with purchase at $1.42 trillion, up 0.8%, and refinance at $571 billion, down 17%. Fannie Mae, on October 12, had $1.865 trillion total, $1.435 trillion purchase, and $430 billion refinance, a 23% refinance share. Those figures were written against mid-7s rate paths. A 6.50% year-end rate does not reopen a 2019-style refinance wave: 65.6% of the outstanding stock is still below 5%, and a 4.4% borrower remains about 200 basis points out of the money. It does give the 6% and higher cohort — 22.5% of the stock in the second quarter, and a share Realtor.com puts at 27–28.5% by the second quarter of 2027 if rates stay above 6% until then — a reason to refinance. Purchase is still the business. Affordability at 100.7, with existing sales at a 15-year low, is the starting point. A move from 7.50% to 6.50% is a meaningful payment cut for a new buyer. It is not, on its own, a boom.

Lock-in eases only at the margin

The wall is why sales have been so weak, and why a 6.5% rate improves turnover without normalizing it. Owners at 3% and 4% stay put. Some of those at 6% and 7% can move. Inventory at 5.1 months is already the most balanced it has been in this cycle; a modest thaw in listings is the housing-market counterpart of our rate path, not a surge in supply that breaks prices. The median existing-sale price is still up, 40 months running. We do not forecast a price collapse as the companion to 6.5%.

A posture, not a recommendation

This study is not advice to lock or float any mortgage, and not a recommendation to buy or sell any security. In our base case, a borrower who can wait into late 2027 is waiting for a lower rate, with no promise that the path is smooth — the first-quarter point is still 7.15%. In the sticky case, waiting is expensive. In the deeper case, more waiting is rewarded. Those are scenarios. They will be graded.

Shelter disinflation has stalled, and new leases are no longer the gift

In the October CPI, rent of primary residence rose 0.20% month over month, owners’ equivalent rent 0.32%, and shelter 0.27% — BLS rounds shelter to +0.3% (computed from FRED CUSR0000SEHA, CUSR0000SEHC, and CUSR0000SAS2RS; retrieved November 13). Zillow, updating after the 8:30 a.m. ET November 10 release, wrote that “housing disinflation remains stalled.” Its market-rent measure has accelerated every month since April, from 1.6% year over year in March to 3.0% in October, and is now growing faster than CPI rent. Zillow raised its forward path to 3.0% for single-family and 2.6% for apartments. There is no official 12-month CPI shelter rate (October 2025 is missing, shutdown); Zillow’s estimates are OER at +3.38% and rent at +2.80%. The April 2026 spike in the official indexes was mechanical — the missing October 2025 increment was added then — but September and October OER, at +0.35% and +0.32%, did not settle back at 0.2%.

If new leases no longer run meaningfully below official rent, the delayed CPI gift is largely spent. Three months of 0.3% core, about 3.6% annualized, and a 0.4% headline, with PPI at 6.5% year over year (CNBC, November 13), are the inflation facts a cutting cycle has to overcome. Our base does not deny them. It assumes they cool, with a lag, as prior tightening and a weak housing market work through. If they do not, the sticky 30% is the right scenario.

Labor flows are not, yet, breaking

October payrolls rose 165,000; unemployment was 4.1%; average hourly earnings rose 0.4% in the month and 3.1% over the year (BLS, 8:30 a.m. ET November 6). Underneath those stocks, the flows the easing case would like to see rolling over are not doing so. September JOLTS openings were 7.523 million, up 333,000, a 4.5% openings rate, about 1.05 openings per unemployed (BLS, November 3; FRED JTSJOL). Quits were 3.4 million, a 2.1% total-nonfarm rate, officially “little changed.” Initial claims for the week ended November 7 were 200,000, a four-week average of 195,250, against 228,000 a year earlier. Continuing claims for the week ended October 31 were 1.697 million, down 20,000, against 1.946 million a year earlier; the insured rate was 1.1% (Department of Labor). Temporary help, as we said, is up 3.0% year over year even as it sits 20% below its peak. A labor market this firm is the main reason the futures strip still sits near 4.50%–4.75% in late 2027, and the main reason our cutting cycle is a judgment, not a reading of the current flows.

Term premium can stay high, and 2023 already broke the post-hike pattern

Kim-Wright at 87 basis points is not an aberration of a single day. The 2026 year-to-date average is 76. The fiscal facts that helped put it there have not gone away: a $1.992 trillion fiscal-2026 deficit (final Monthly Treasury Statement, through September 30), coupon auction sizes held for several quarters (Treasury, November 4), and foreign official Treasury custody down $218 billion year over year. The Fed’s agency MBS book is still running off — $1,883.522 billion on November 11, down $186.441 billion year over year, and down $155.5 billion since December 31, 2025, even though “QT” as a slogan ended on December 1, 2025 (H.4.1, November 12; FRED WSHOMCB). Commercial banks held $2,730.5 billion of agency MBS in the week ended November 4, more than the Fed, but only about $15 billion more this year (FRED TMBACBW027SBOG, updated November 13 at 3:19 p.m. CT). And after the July 2023 last hike, the 10-year rose 41 basis points over twelve months. A cycle can refuse to follow the 1995 and 2018 script. That is the 30% case.

Sources, with as-of times