The mortgage penalty that defined 2023 has quietly unwound. Rising Treasury yields took every cent of it.

The explanation everybody reached for in 2023 has quietly stopped being true.

The story went like this. Mortgage rates were brutal not because the government's borrowing costs were high, but because the gap between the two had blown out to something almost never seen. Lenders and bond investors were demanding an extraordinary premium to put a 30-year loan on an American house, and until that premium came down, no amount of good news on inflation would reach a borrower.

Every part of that was correct at the time. The premium has since closed by about three quarters. And the 30-year mortgage rate today, 6.55 percent, sits 26 basis points below the 2023 average of 6.81 percent.

Almost nothing reached the borrower. Here is where it went.

What is the mortgage-Treasury spread, and what is it now?

The spread is the difference between the average 30-year fixed mortgage rate and the yield on the 10-year Treasury note. In the week of July 16, 2026, it was 1.98 percentage points: a 6.55 percent mortgage against a 4.57 percent 10-year. The long-run average from 1990 through 2019 was 1.66 points.

The 10-year is the reference because it approximates how long a 30-year mortgage actually lives. Almost nobody keeps one for thirty years. People sell, and they refinance, and the practical life of the loan has historically clustered closer to a decade, which is why the mortgage market prices off the 10-year rather than the 30-year bond.

What sits in that gap is not one thing and not one party's profit. It compensates the investor who buys the mortgage-backed security for the risk that the borrower repays early, it covers the lender's cost of originating and servicing the loan, and it absorbs whatever the market currently charges for holding housing credit rather than government credit. When commentators say the spread is wide, they are describing a total, not an accusation against a particular actor.

How wide did the spread get, and when did it peak?

It peaked at 3.26 percentage points in the week of November 10, 2022, when Freddie Mac's survey put the 30-year at 7.08 percent while the 10-year Treasury yielded 3.82 percent. That is the widest weekly reading in the Treasury Department's daily par yield series going back to 1990, wider than any week of the 2008 financial crisis.

The 2008 comparison is the one worth holding onto. At the depth of the crisis, in the week of December 18, 2008, the spread reached 3.11 points, with a 5.19 percent mortgage against a 2.08 percent 10-year. The 2022 blowout beat it. A housing market with no bank failures and no credit event priced mortgage risk more expensively than a housing market that was actively collapsing.

For scale in the other direction: the narrowest week since 1990 was March 8, 1996, at 0.97 points. In 2021 the spread averaged 1.51. The whole distribution of the last thirty-six years runs from roughly one point to roughly three, and 2022 and 2023 sat at the ceiling of it.

How much of the excess has actually closed?

About three quarters. Measured against the 1990-to-2019 average of 1.66 points, the 2023 spread carried 1.19 points of excess. The 2026 average through July carries 0.30 points. The annual series shows a clean four-year descent rather than a sudden repair.

Each row is the mean of that year's weekly readings. The 2026 row covers 28 weeks.

The descent is steady and it is not finished. The spread has not printed below 1.80 since the week of February 10, 2022, and the narrowest week of 2026 so far is 1.83. Calling this normalized would be premature. Calling it the defining problem of the mortgage market, as was fair in 2023, is no longer supportable.

If the spread fell more than a full point, why is the mortgage rate almost unchanged?

Because the 10-year Treasury rose by most of what the spread gave back. From the November 2022 peak to July 2026, the spread narrowed by 1.28 percentage points while the 10-year yield rose by 0.75 points, from 3.82 to 4.57 percent. The mortgage rate fell by the difference, 0.53 points, from 7.08 to 6.55 percent.

That arithmetic is the whole article. A borrower experiences one number, and that number is the sum of two independent series that have spent four years moving in opposite directions.

It also means the composition of the rate has changed even though its level has not. In 2023, the average 6.81 percent mortgage was 3.96 points of Treasury yield and 2.85 points of spread: 58 percent government borrowing cost, 42 percent everything else. Today's 6.55 percent is 4.57 points of Treasury and 1.98 of spread, or 70 percent and 30 percent. The mortgage market has gone back to looking roughly the way it did before 2022. The government's cost of money has not.

This is worth being precise about, because the two facts point at different futures. A wide spread is a market dislocation, and dislocations tend to mean revert. A high 10-year yield reflects the price of government debt, and nothing about it is required to revert on any schedule.

What has the spread cost a borrower in dollars?

At the peak it cost about 417 dollars a month on a 400,000 dollar loan. Today the remaining excess costs about 84 dollars a month on the same loan. Both figures compare the actual rate to what the rate would have been at the 1990-to-2019 average spread, holding the Treasury yield fixed at its level on the day.

The arithmetic is a standard amortization on a 400,000 dollar 30-year fixed loan, and anyone can reproduce it. In November 2022, a 7.08 percent rate produced a payment of 2,683 dollars a month. Had the spread been 1.66 points instead of 3.26, the rate would have been 5.48 percent and the payment 2,266 dollars, a difference of 417 dollars a month and roughly 150,000 dollars across 360 payments. Today the same comparison runs 6.55 percent against 6.23 percent: 2,541 dollars a month against 2,458, a difference of 84 dollars and about 30,000 dollars over the life of the loan.

The 400,000 dollar figure is an illustration chosen for round arithmetic, not a claim about what any particular household borrows. Scale it as needed. As a rule of thumb on that principal, a full percentage point of rate is worth roughly 258 dollars a month.

What actually moves the spread?

Several things at once, and disentangling them properly requires data this piece does not have. What can be documented is that the largest single buyer of American mortgage debt has been shrinking its position throughout the period the spread has been elevated.

The Federal Reserve held 2,740,185 million dollars of mortgage-backed securities on April 13, 2022, according to the H.4.1 release. As of July 15, 2026, it held 1,948,337 million. That is a decline of about 792bn dollars, or 28.9 percent, achieved by letting the portfolio run off rather than by selling.

The mechanism behind the rest is structural rather than newsworthy. A mortgage-backed security gives the borrower a free option to prepay, which the borrower exercises precisely when it hurts the investor most, refinancing into lower rates and handing back principal that must then be reinvested at those lower rates. The value of that option rises with uncertainty about future interest rates. When the market is unsure where rates are going, the compensation demanded for writing that option goes up, and it shows up in the spread.

That mechanism explains why 2022 and 2023, two years of extraordinary uncertainty about the path of policy rates, produced the widest spreads on record. It does not by itself prove that the narrowing since is caused by falling uncertainty, and this piece does not have the volatility data to make that claim. The correlation between Federal Reserve runoff and the spread is likewise a correlation. The portfolio has been shrinking continuously since 2022, through both the widening and the narrowing, which is a fact that sits awkwardly with any simple story in which Fed selling is the main driver.

What would it take to get back to a 5 percent mortgage?

On today's spread of 1.98 points, a 5 percent mortgage requires a 10-year Treasury yield of roughly 3 percent, against 4.57 percent now. Alternatively, with the 10-year unchanged, it would require the spread to compress to about 0.43 points, which has never happened in the series since 1990.

Put plainly, the remaining path to materially cheaper mortgages runs through the Treasury market rather than through the mortgage market. Even a full return of the spread to its 1990-to-2019 average, an outcome that would represent complete normalization, gets the rate to about 6.23 percent on today's 10-year. The spread simply does not have another 150 basis points to give, because it never had them to begin with.

This is not a forecast about where Treasury yields go. It is a statement about which of the two series has room to move.

Where this analysis is weakest

Two problems, and the second one is genuine.

The first is small. The spread is computed here as a residual, the mortgage rate minus the Treasury yield, which lumps together the lender's origination margin and the bond investor's required yield into a single number. Those two components have moved differently and a decomposition into the primary-secondary spread and the option-adjusted spread on the securities themselves would be a better piece of work. It would require data that is not free and not public.

The second is a break in the series, and it lands almost exactly on the peak. Freddie Mac changed the Primary Mortgage Market Survey methodology effective the week of November 17, 2022, moving from a survey of lenders to data drawn from mortgage applications, and stopped publishing the points and fees that accompanied the rate. Before that change, the quoted rate came alongside an average of 0.95 points paid by the borrower, with a median of 0.70, across 1,715 weekly readings from 1990. Those points were a real cost that the quoted rate excluded.

The direction of that bias matters and it runs against the argument's convenience rather than for it. Because the historical rates understate the borrower's all-in cost by whatever those points were worth, the pre-2022 average spread of 1.66 points is, on a like-for-like basis, somewhat too low. The true historical baseline is a little higher, which means today's 1.98 is closer to normal than the comparison above suggests, and the remaining excess of 0.30 points is somewhat overstated. No adjustment has been applied here, because converting points to a rate equivalent requires an assumption about how long the loan is held, and inventing that assumption would be worse than naming the limitation.

The peak reading of 3.26 in the week of November 10, 2022 falls in the old methodology, and carried 0.9 points on top. The current reading of 1.98 falls in the new one. Any comparison across that boundary should be read as approximate.

Method, and the sources

The mortgage series is Freddie Mac's Primary Mortgage Market Survey, taken from the full historical file the company publishes. The Treasury series is the Department of the Treasury's Daily Treasury Par Yield Curve Rates, downloaded year by year from 1990 through 2026 and parsed by column header, because the number of maturities Treasury publishes has changed several times over the period.

Each weekly survey reading was matched to the 10-year constant maturity yield on the same date, falling back to the nearest prior business day where the survey date was a market holiday. That produced 1,907 matched weeks from January 1990 to July 2026. Averages are unweighted means of weekly readings.

The Federal Reserve holdings are from the H.4.1 statistical release dated July 16, 2026 and the archived release of April 14, 2022, reading the line for mortgage-backed securities held outright in the supplemental table.

If you find an arithmetic error, or you have the decomposition data this piece is missing, write to corrections@moneyandworld.com and we will correct it and say that we did.

Frequently asked questions

What is a normal mortgage-to-Treasury spread? About 1.66 percentage points, the average from 1990 through 2019. The median across that period was 1.62. Readings between roughly 1.5 and 2.0 points are ordinary, below 1.2 is unusually tight, and above 2.5 has only occurred around the 2008 crisis and in 2022 and 2023.

Why did mortgage spreads widen so much in 2022 and 2023? The spread reflects what investors charge for the prepayment option embedded in a mortgage, and that charge rises with uncertainty about future interest rates. Those two years carried extraordinary uncertainty about the path of policy. The Federal Reserve was also shrinking its mortgage-backed securities holdings throughout.

Is the mortgage spread still elevated in 2026? Slightly. It averaged 1.96 points through July 16, 2026, against a 1990-to-2019 average of 1.66, so roughly 0.30 points of excess remains. That is down from 1.19 points of excess in 2023, meaning about three quarters of the abnormal premium has closed.

Why have mortgage rates not fallen if the spread narrowed? Because the 10-year Treasury yield rose over the same period, from 3.82 percent in November 2022 to 4.57 percent in July 2026. The spread narrowed by 1.28 percentage points and the Treasury yield rose by 0.75, leaving a net decline in the mortgage rate of only 0.53 points.

What 10-year Treasury yield produces a 5 percent mortgage? Roughly 3 percent, if the spread stays near 1.98 points. At the historical average spread of 1.66, a 5 percent mortgage would require a 10-year yield of about 3.34 percent. The 10-year yielded 4.55 percent on July 17, 2026.