The Fed Hiked, and the Dots Matter More Than the Hike
The FOMC raised the target range 25 basis points to 3.75%–4.00% on September 16. It was the first hike since 2023 and it was unanimous.

The vote is less informative than the Summary of Economic Projections. Sixteen of eighteen participants now expect at least one more increase this year, which puts 4.00%–4.25% in play by December. Only four see cuts in 2027. Chair Warsh was direct about why, saying that “inflation is too high and has been too high for too long.” He also described financial conditions as still providing accommodation rather than restraint, which is the single most hawkish thing he said all week.

The market took the point. October hike odds sit near 60%.

What Twist Flattening Means, and Why It Is Not Good News
The curve did not shift in parallel last week. It twisted.

A parallel shift moves every maturity by roughly the same amount. A twist pivots the curve around a point, with yields on one side rising while yields on the other side fall or hold. Last week the pivot sat at the 20-year. The 2-year rose 13 basis points and the 3-year rose 14, while the 20-year was unchanged and the 30-year fell a basis point. The 10-year rose only 5 basis points to 5.01%. Short rates up and long rates flat to lower. The slope compressed, which is textbook twist flattening.

The economics behind it are straightforward. The front end prices monetary policy, so a hike plus a hawkish dot plot lifts it immediately. The long end prices terminal growth, long-run inflation and term premium. When the market believes the Fed will tighten enough to actually break inflation, the long end can hold or rally even as the front end sells off. The curve is telling us the Fed has become more credible on inflation and less optimistic on growth.

A flatter curve compresses carry for every leveraged buyer of agency MBS. Banks, mortgage REITs and money managers fund short and own long, so slope is the economics of the trade. When 2s/10s goes from 46 basis points in August to 25 basis points today, the marginal MBS buyer gets paid materially less to show up. That is a spread problem, not a rate problem, and it is the mechanism most likely to keep mortgage spreads from compressing even if Treasury yields cooperate.

The Bank of England's QT Rethink
The BoE held Bank Rate but announced it will stop selling long-dated gilts entirely and hold to maturity both the £221.7 billion maturing before 2035 and the £120 billion of longest-dated holdings that back banknote issuance.

This is worth watching for a reason that has nothing to do with the UK. A major central bank just conceded that long-end supply was doing damage it was not willing to keep inflicting. Our Treasury reached a version of the same conclusion in August when it doubled long-duration buybacks. Two developed-market debt managers independently deciding that the long end needs help is a data point about term premium globally, and agency MBS competes in that market.

The Bank of Japan
The BoJ raised its policy rate 25 basis points to 1.25%, the highest level since 1995, while signaling a slower pace ahead. Japanese institutions are among the largest foreign holders of Treasuries, and every increase in domestic JGB yields improves the case for repatriating that capital. The move was expected and the signal was dovish at the margin, so I would not overstate it.

Mortgage Rates at a Twenty-Month High
Freddie Mac's survey printed 6.95% on September 17, up 19 basis points on the week and 30 basis points since my August report. That is the highest 30-year reading since January 2025, and it is 105 basis points above the February low of 5.90%. The 15-year moved to 6.26% from 6.09%.

We are also now 69 basis points above the year-ago level of 6.26%. The year-over-year comparison turned against us in August and has gotten worse since.

Spreads, and Which One You Mean
The mortgage-to-Treasury spread has widened. The 6.95% survey rate against a 4.94% 10-year is 201 basis points, versus 196 in August, 195 in May and 193 in February. That is the widest of the reference points I have tracked this year and roughly 31 basis points above the 170 basis point historical norm.

The primary/secondary spread has not widened. I see it at 95.4 basis points on September 18, against a 99.1 basis point August average and a 102.9 basis point trailing twelve-month average. That is the 14th percentile of the last twelve months, closer to the May tight of 85.2 than to the February wide of 119.1.

Originator margin has compressed while the gap between mortgages and Treasuries has widened, which means the widening is happening in the secondary market rather than at the primary pricing level. The par coupon went from 5.532% in August to 6.000% on September 18, a 47 basis point move, while the weighted-average originator rate rose 35 basis points to 6.954%. We passed through less than the market gave us, and the spread compressed because of it.

Originators are absorbing spread to defend volume in a 6.95% environment, and a 95 basis point primary/secondary spread near the twelve-month tights is not much cushion if rate volatility picks up into the October meeting.

Non-QM
Non-QM deal spreads have held in, though I would not read that as a verdict on credit. Agency is where the damage is showing up, and the securitization market has not yet had to price the Fed's hawkish turn through a full new-issue calendar. With 2026 non-QM origination running toward record levels, the bid gets tested on supply well before it gets tested on performance.

Mortgage Insurance

FHFA Director Pulte's post on Tuesday put mortgage insurance cancellation policy on the industry's agenda for the first time in years:

“If your Home is worth more, or you have paid the loan down far enough, you should be able to drop EXTRA Mortgage Insurance. Right now Fannie Mae will not let your loan company call and tell you that you may qualify because your home has gone up in value! You have to know to ask. Freddie Mac will let them call. That is crazy. We are fixing that. Fannie Mae will follow Freddie's rule: if you may qualify to drop unnecessary extra insurance because your home has gained value, they can contact you and walk you through how to cancel your PMI. You can stop paying for coverage you do not need and keep the money.”

What the Rules Actually Say
Borrowers have three paths to cancellation today, and only one of them recognizes appreciation.

The policy gap Pulte identified was never about eligibility. It was about who is allowed to raise the subject. Freddie has always permitted servicers to solicit borrowers for cancellation and Fannie has not. Fannie updated its Servicing Guide the following day to allow proactive solicitation.

Why a Servicer Would Volunteer to Reduce a Borrower's Payment
The question I keep getting is why a servicer would go looking for reasons to cut its own borrower's payment. The answer is retention, and the logic is tighter than it first appears:

1. A borrower who has built equity has a real incentive to refinance, because lower LTV means lower LLPAs and a better rate than the one they have.
2. That borrower is a refinance target for every competitor with access to a credit trigger, and we lose the servicing if they leave.
3. Cancelling the MI delivers most of the payment relief a refinance would deliver, without a new loan and at almost no cost to the servicer.
4. Performance data supports the strategy, since borrowers whose MI has been cancelled refinance at lower rates than those who still carry it.

So the servicer spends nothing meaningful, the borrower saves real money every month, and the servicing asset stays put.

Does the Fannie Change Actually Move Cancellations?
Despite Freddie's more permissive stance, cancellation rates between the two GSEs have not diverged materially. Permission was available and the industry largely did not use it.

I expect that to change because FHFA is now publicly invested in the outcome and the Director has made it a personal talking point. Formal alignment plus visible regulatory attention is a different operating environment than quiet permission.

The Implementation Detail to Watch
A borrower seeking cancellation today needs either a BPO or a full appraisal, priced by the GSEs at $190 and $450 respectively. Those are marginal costs, since a borrower recoups them within a month or two of the cancelled premium, so they do little to discourage anyone who is already motivated.

If the GSEs move to their automated valuation models to establish property value, even that friction disappears.

This is good policy. It is genuine payment relief for borrowers who have earned it, and it hands servicers a retention tool that costs them almost nothing. I would rather compete in a market where the borrower gets the benefit of their own equity without having to know the right question to ask.

Federal Funds Rate

Next FOMC Meeting

Rate Change Probabilities

Context This is a complete reversal of the August setup. A month ago the market put September hike odds near one in three and the debate was whether the Fed would ease into year-end. Today the debate is how far the tightening runs. Note that the terminal rate implied by three more hikes is an inference from the strip and should be treated as the softest number in this table.

Fed Influencing Factors

Why the Fed Moved The August payroll rebound to +162,000 removed the labor-market excuse for patience that the July print had handed the doves. With employment re-accelerating and headline CPI stalled at 3.4%, the Committee had a clean case to tighten. The awkward detail is core CPI at 2.4%, which is the lowest since March 2021 and not obviously the profile of an economy that needs three more hikes. The gap between headline and core is almost entirely energy.

Freddie Mac PMMS & Mortgage Rates

Freddie Mac “The 30-year fixed-rate mortgage continues to fluctuate as markets assess economic data.” — Sam Khater, Chief Economist, Freddie Mac (September 17, 2026)

30-Year Mortgage Rate Trend

Source: Freddie Mac PMMS via FRED (series MORTGAGE30US), data through September 17, 2026

Mortgage-to-10-Year Treasury Spread

Spread Commentary

 The remarkable stability I described in August has broken. After holding in a 193–196bp band from February through August, the spread stepped out to 201bp. It is a 5bp move and I would not call it a regime change, but the direction is unhelpful and it arrived in the same week the curve flattened. Returning to the 170bp norm against today's 5.01% 10-year would imply a 30-year near 6.71%, so there is roughly 31bp of theoretical relief available with no help from Treasuries.

Primary/Secondary Mortgage Market Spread

Spread Commentary

 The August report flagged 107.0bp as the widest reading since July and expected compression toward the mid-90s over the following two weeks. That is exactly what happened, and it kept going. At 95.4bp we sit 11.6bp inside the August print and 7.5bp inside the trailing twelve-month average, in the 14th percentile of the last twelve months. The mechanism is the par coupon moving faster than primary pricing, with the coupon up 47bp since August 25 against 35bp on the weighted-average originator rate. Originators are eating spread to hold volume at 6.95%. That works until volatility forces the desk to widen, and 95bp is a thin buffer heading into October.

Treasury Yields and the Curve

Curve Shape

Read the WoW column top to bottom and the twist is unmistakable. The move decays monotonically from +14bp at the 3-year to -1bp at the 30-year, with the pivot at the 20-year. This is the mirror image of August, when the long end did all the work and I described continued steepening. 2s/10s has given back 21bp since that report and 10s/30s has given back 20bp. The 20-year at 5.38% now trades above the 30-year at 5.34%, a mild inversion at the very long end that usually reflects the 20-year's relative illiquidity rather than a signal.

Additional Housing Statistics

Data Note

 These remain July figures. NAR's August existing-home sales release falls after this report date, so the housing section has not been refreshed since August 26. Worth remembering that this data predates both the 6.95% survey rate and the Fed's hike, so it tells us nothing about how the market is absorbing either.

Upcoming Market-Moving Events