6.95% on an 18bp Move (17 September 2026): The Fed Hiked, the Ten-Year Fell — and the Mortgage Spread Absorbed Nothing

The Federal Open Market Committee raised the target range for the federal funds rate by a quarter point to 3-3/4 to 4 percent on 16 September 2026, its first increase since 2023, on a 12-0 vote. The next morning, the day the new range took effect, the 10-year Treasury yield fell seven basis points to 4.94% and snapped an eight-session climb. Freddie Mac's survey, released the same day, printed the 30-year fixed at 6.95% — up nineteen basis points. This piece argued on 12 September that the Fed does not set the mortgage rate and that the spread which absorbed three years of Treasury moves had nothing left to give. Both halves held. The mortgage rose almost exactly as much as the 10-year had risen inside the survey's collection window, and the spread, measured properly, moved a single basis point.

What actually happened

Three releases landed inside thirty hours, and together they form about as controlled an experiment as this question ever gets.

On Wednesday 16 September at 2:00 p.m. Eastern the Committee raised the target range by a quarter point to 3-3/4 to 4 percent, by a 12-0 vote. The statement was unusually short. It described economic activity as "expanding at a solid pace", said "Inflation remains elevated", and stated that the action "will support a timelier return to the Committee's 2 percent goal."

On Thursday 17 September at 8:30 a.m. the Census Bureau and HUD published New Residential Construction for August. At noon, Freddie Mac's Primary Mortgage Market Survey put the 30-year fixed at 6.95%, up from 6.76%, with the 15-year at 6.26% from 6.09%. A year earlier the 30-year had averaged 6.26%.

And across the same session the 10-year Treasury yield fell.

US Treasury daily par yield curve closes.

The Fed raised the overnight rate on Wednesday. On Thursday the rate that actually prices a mortgage was seven basis points lower than it had been on Tuesday.

The 5% close the series had not seen since 2007

The round number was finally taken, and on official closes rather than an intraday print. On the Treasury's daily par yield curve the 10-year closed at 5.00% on 15 September and 5.01% on 16 September. Checked across every close in the intervening years, those are the first at or above 5.00% since 19 July 2007, when it printed 5.04% — the 2023 selloff that everyone remembers as the 5% scare stopped at 4.98%.

That makes the comparison exact rather than rhetorical, and the realized numbers sharpen the original point rather than softening it.

Treasury yields from the daily par yield curve; mortgage rates from Freddie Mac's PMMS (19 July 2007 quoted with 0.4 points); 2007 effective fed funds from the Federal Reserve Bank of New York, whose effective rate printed 3.63% through 16 September 2026 before the new range took effect.

Read the third row against the second. Even after raising rates, the overnight rate the Federal Reserve controls sits roughly 137 basis points below where it stood in July 2007 at the midpoint of the new range. The rate an American household is quoted on a thirty-year mortgage is 22 basis points above it. The Fed has just tightened, and the gap between the two dates is wider in the borrower's disfavour than it was before the tightening.

The curves also remain different animals. In 2007 the 10-year sat almost exactly between the 2-year and the 30-year — a flat curve, a level story about policy. Today it sits 27 basis points above the 2-year and 34 below the 30-year: a slope story about duration and issuance, examined at length in the August long-end selloff. A slope story reaches a mortgage far more directly than a level story about overnight money.

Why a hike can lower the ten-year

This is the part that looks like a paradox and is not one.

A 30-year fixed mortgage is built in three layers. The base is the 10-year Treasury yield, used as a proxy because a 30-year loan is never held for 30 years — it is refinanced, or the house is sold. The second is the spread investors demand for holding mortgage credit rather than government credit, which mostly prices one feature: the borrower owns a free option to prepay. When rates fall the loan disappears into a refinancing and the investor is repaid at the worst moment; when rates rise the borrower keeps the cheap loan. That asymmetry has a price, and the price rises with interest-rate volatility. The third is the originator's own margin.

None of the three is the federal funds rate. The funds rate reaches the 10-year only through the expected average of overnight rates across a decade, plus a term premium — and that expected average can fall on the day of a hike if the Committee signals the tightening is nearly over.

The spread held, and the stopwatch flipped direction

The earlier version of this piece corrected its own reading of a 181 basis point print, and the correction is what made this week's number predictable.

Since November 2022 Freddie Mac has built the survey from loan applications submitted to its Loan Product Advisor system, aggregating applications from the Thursday before the publication week through the following Wednesday and publishing Thursdays at noon. The mortgage number released on 17 September therefore describes rates locked between 10 and 16 September. The Treasury yield it is conventionally set against describes 17 September.

Collection-window yields are the simple average of Treasury par-yield 10-year closes across the survey's Thursday-to-Wednesday application window.

Three weeks, three same-day readings of 194, 181 and 201 basis points — a series that appears to compress violently and then blow out. Aligned to what the survey actually measures: 196, 196, 197. The spread has not moved in three weeks.

The bias has now shown itself in both directions, which is the cleanest possible evidence that it is measurement rather than market. In the week to 10 September yields were rising, the Treasury leg updated daily while the mortgage leg lagged, and the raw spread understated by 15 basis points. In the week to 17 September the 10-year fell seven basis points on publication day alone, and the raw spread overstated by four. Same artefact, opposite sign.

What that leaves is the arithmetic the previous version set out in advance: if the aligned spread simply held, a window averaging between 4.96% and 4.99% would map to a print in the low-to-mid 6.90s. The window averaged 4.978%. The print was 6.95%. The borrower absorbed all 18.3 basis points the bond market handed over, and the cushion between the two did no work at all.

The housing data flipped — and then revised away

August's construction report deserves reading backwards, starting with the revisions.

The July release this piece originally cited showed permits at 1,443,000 and starts at 1,239,000, a 12.4% monthly collapse, with single-family starts at 808,000. The August release revised July starts up to 1,309,000 and July single-family starts up to 853,000, while trimming permits to 1,433,000. The dramatic gap between permits pulled and ground broken — the split the earlier version called the clearest read available on builder intent — was roughly 70,000 units smaller than first reported.

Against those revised figures, August ran: permits 1,394,000, down 2.7%; starts 1,275,000, down 2.6%; single-family starts 918,000, up 7.6%; completions 1,128,000, down 11.9%.

Builder sentiment is the cleaner signal, and it moved the way the rate data implies. The NAHB/Wells Fargo index fell three points to 32 in September from 35 in August, its lowest in a year. Current sales conditions fell four points to 35, future sales expectations six points to 37, and prospective buyer traffic held at 23. The share of builders cutting prices rose to 38% from 35%, with the average cut at 6% for a sixth consecutive month, and 66% reported using sales incentives, up from 63%.

That last number is the spread story wearing a different hat. A mortgage-rate buydown is a builder paying cash to compress this very spread, one buyer at a time, and every basis point on the 10-year makes each one more expensive — the trade-off Lennar's third quarter priced out in gross margin this week.

The honest read-across to markets

This is a rates and housing story before it is anything else. The instruments it genuinely touches are Treasuries, mortgage-backed securities, and the homebuilders and building-products companies whose margins absorb those buydowns — which is arithmetic in an income statement, not a view on any security.

The connection to the currency board is slower and runs one way only. Shelter is the heaviest single component of US consumer inflation and is measured with a lag long enough that today's cooling appears in a print some quarters from now. Lower measured inflation eventually lowers the expected policy path, and the expected path relative to other countries is what the interest-rate factor on the meter scores for the US dollar. That is a chain with four links in it, and this week showed why the links matter: the dollar traded the projections, not the mortgage survey.

What would change the picture now

The spread is the variable to watch, because it has now shown that it will not cushion anything. At 197 basis points against a 2014-2021 norm of 175bp it retains some theoretical travel, but three consecutive weeks of refusing to move while the base rate ran 18 basis points is a strong statement about how much of that travel exists in practice.

Next is volatility. The hike arrived with projections that put the funds rate on a plateau rather than a climb, which is the configuration that keeps the prepayment option cheap and the spread stable. A path that reopens — a sustained inflation surprise of the kind August's core CPI at 0.3% against 0.2% expected delivered — would widen the plausible range of future rates, raise the price of that option, and push the spread and the base rate up together. That combination genuinely reprices a mortgage, and it did not happen this week.

Slowest is supply. Completions 11.9% below July and 27.1% below a year ago eventually tightens what is available to buy, but against 9.6 months of new-home inventory at the last reading that constraint changes the arithmetic in 2027, not this quarter.

What did not change the picture was the 25 basis points itself. Three central banks decided within three days, the Federal Reserve moved an overnight rate for the first time since 2023, and the 10-year Treasury ended the week seven basis points lower than it began the day of the decision. The mortgage rate went up anyway — by almost exactly the amount the bond had already moved the week before.

More on how this site approaches transmission mechanisms is on the about page, and the underlying reason long yields and currencies move together is covered in bond yields and currencies.

Educational macro context only — not investment advice.