Mortgage Rate Prediction 2026: Markets Say 7.5%

Every mortgage rate prediction on the first page of Google right now is built on a number that has already been overtaken. I checked them one by one this morning. One of the pages ranking for this exact query was published in September 2023 and still forecasts 5% by late 2024.

Here is the live number instead. The Freddie Mac 30-year fixed averaged 7.40% in the survey released October 8, 2026, up from 7.28% the week before and 6.76% four weeks ago. That is the seventh consecutive weekly increase and the highest print since November 2023. On a $400,000 loan the move since early September costs an extra $172 a month.

What nobody ranking for this term seems to know is that you can read the market's own forecast rather than an economist's. Polymarket operates a ladder of contracts that resolve directly off the Freddie Mac Primary Mortgage Market Survey, and a second ladder on the 10-year Treasury yield that drives it. I pulled both books, both venues' Fed pricing, the Treasury curve and Fannie Mae's September forecast PDF, and ran the spread arithmetic myself. The answer is that the published forecasts and the traded prices are not describing the same year.

Where the 30-Year Rate Actually Is Right Now

Mortgage rates have moved in a straight line for seven weeks. The table below is the Freddie Mac weekly survey next to the Treasury 10-year par yield from the same week, with the spread between them in the last column. I use the Treasury Department's own daily curve for the yield rather than a data vendor, because the Polymarket contracts resolve off it.

Two things jump out. The mortgage rate has risen 75 basis points in seven weeks while the 10-year has risen 53, so roughly a third of the move is the spread widening rather than the bond market selling off. And the 10-year is doing something historic: the Treasury par close hit 5.31% on October 5, its high for 2026, after starting the year at 4.19%. Intraday on October 7 it traded above 5.35%, the highest since 2002.

For context on how far this has travelled: the 2026 low was 5.98% in the week ending February 26. The year-to-date average is 6.46%. A borrower who locked at the February low and a borrower locking today are 142 basis points apart on the same house.

What Each Mortgage Rate Prediction Says, and What It Was Built On

Almost every mortgage interest rate prediction you will read is a repackaging of three or four institutional forecasts. So I went to the source documents rather than the aggregators, and the footnotes turned out to matter more than the numbers.

Fannie Mae's Housing Forecast is dated September 11, 2026 and projects the 30-year fixed averaging 6.8% in the fourth quarter. The first two weeks of that quarter printed 7.28% and 7.40%, an average of 7.34%. For the quarterly figure to land on 6.8%, the remaining weeks would have to average roughly 6.71%, which means an immediate 69 basis point collapse holding for the rest of December.

The reason the forecast is this far out is printed at the bottom of the page. The note reads: "Interest rate forecasts are based on rates from August 31, 2026." On August 31 the 30-year fixed was 6.66% and the 10-year was 4.67%. The Federal Open Market Committee then raised its target range on September 17 and the long end of the Treasury curve sold off through late September. The forecast published on September 11 could not see any of it, and nothing on the aggregator pages that quote it says so.

The companion Economic Forecast is the more revealing document. It has the 10-year Treasury averaging 4.8% in Q4 2026 and never exceeding 4.9% through the end of 2027. The 10-year closed at 5.22% yesterday. Meanwhile its federal funds projection of 4.0% for Q4 2026 is close to what markets price. Fannie Mae has the Fed roughly right and the long end badly wrong, which tells you where the forecast error in every mortgage prediction currently lives.

Prediction Markets Price the Mortgage Rate Directly

This is the part the SERP misses completely. Polymarket lists an event titled "Will the 30-year Mortgage Rate hit __ in 2026?" whose resolution source is, in its own words, the 30-year Fixed Rate Mortgage published through the weekly Primary Mortgage Market Survey. It is not a proxy. It settles on the same number your lender's pricing desk watches.

Read it as a forecast and it says this: the rate is 10 basis points below 7.50% and the market gives better than two-to-one odds that it gets there on some weekly print before New Year. The chance of returning to 6.50%, which is the figure the Mortgage Bankers Association has carried as its 2027 base case since May, is 18%. The chance of a 5-handle is under 3%.

One honesty note on that ladder, because it is the kind of thing that gets quoted carelessly. The 5.50% strike is quoted at a 3.3% mid while the 5.70% strike sits at 1.45%, and that is backwards: anything that reaches 5.50% has already passed through 5.70%, so the lower strike cannot be the likelier one. Both books are a couple of cents wide on tiny size, so this is untraded noise rather than an arbitrage. It is also a reminder to read the bid and the ask rather than the midpoint on any thin strike, which is how I handle every number in this piece.

The 10-year ladder is the deeper book, with about $956,000 of volume against $98,000 on the mortgage market, and it implies a full distribution for where the yield peaks this year.

That distribution is monotone and sums to 100%, which is more than I can say for most published rate forecasts. The downside ladder is the one that should worry anyone waiting for cheaper money: the market prices an 11.5% chance the 10-year even touches below 4.6% at any point before 2027, and 5.4% on below 4.5%. Fannie Mae's forecast needs it to average 4.8% for the quarter.

How the Market's Mortgage Forecast Moved in Four Weeks

A forecast is only interesting if you can see it change its mind. I pulled the daily order-book history on both mortgage contracts, and the swing is larger than anything in the published literature.

On September 10, one day before Fannie Mae published its forecast, the market put a 53.5% chance on mortgage rates getting back to 6.50% or lower this year. Today it is 18%. Over the same four weeks the odds of a 7.50% print went from 10.5% to 68.5%, a 58 point move.

The inflection is September 23 to 25, when the 10-year broke out of a 4.95% range to 5.18% and both mortgage contracts repriced hard in a single session. No reporting I can find names a single trigger for that leg, and I would rather say so than invent one. What the market consensus was pointing at by the following week was a global long-end selloff, with Japanese, British and French yields dragging Treasuries along, plus heavy issuance and the capital demands of data-centre construction.

Why the Fed Is Not What Is Driving This

Here is the claim that separates this from every other mortgage interest rate prediction you will read today. The Fed path has been easing while mortgage rates rose. That is the opposite of the mechanism the aggregator pages describe, and both prediction-market venues agree on it.

Multiply each ladder out and add it to today's 3.875% midpoint and you get a year-end policy rate of 4.11% on Polymarket and 4.09% on Kalshi. Two unarbitraged books landing two basis points apart is the strongest result on this page, and it sits right on the Fed's own September projection of 4.1%. Kalshi's October ladder is carrying 1.4 million contracts of open interest on the no-change leg alone, so this is not a thin-market artefact.

Now put the two series side by side. On September 29 the October hike contract traded at 68.5% and the same arithmetic gave a year-end midpoint of 4.24%. Today the October hike is 15.5% and the midpoint is 4.10%. The market removed 14 basis points of 2026 tightening and priced out an entire near-term hike, a 53 point swing, and the 30-year mortgage rose from 7.03% to 7.40% across the same window.

The December contract is the one to watch, and it has barely moved: 77.0% on September 29, 74.5% today. The committee meets October 27 to 28 without a Summary of Economic Projections, then December 8 to 9 with one. September CPI lands October 14 and November CPI lands December 10, the day after the December vote, so the committee decides on a thinner information set than usual.

The Spread Arithmetic That Breaks Every 6.7% Forecast

A 30-year mortgage rate is the 10-year Treasury plus a spread that covers servicing, prepayment risk and whatever the secondary market demands that month. Forecasters lean on that spread compressing. I matched every Freddie Mac weekly print since 2019 against the Treasury close from the same week to see how much compression is actually left.

The spread has already done most of its healing. It went from a 285 basis point average in 2023 to 196 in 2026, and the 218 we are at today is the widest reading of this year rather than a starting point for further tightening. The compression tailwind that the published forecasts are implicitly leaning on is largely spent.

Which lets you ask the question directly: holding the 10-year at yesterday's 5.22%, what mortgage rate does each spread regime produce?

That is the whole argument in one table. To print Fannie Mae's 6.8% with the 10-year where it is, the spread has to go to 158 basis points, inside the range last seen when the Federal Reserve was buying mortgage bonds in size. To print the MBA's 6.5% you need 128 basis points, which is the 2021 floor. Neither is impossible. Both require the spread to do something it has not done since quantitative easing ended.

Run it the other way and the conclusion is the same. At the 2026 average spread of 196 basis points, a 6.8% mortgage needs the 10-year at 4.84% and a 6.5% mortgage needs 4.54%. The market prices an 11.5% chance the 10-year so much as touches below 4.6% before 2027. Every sub-7% mortgage rate prediction for this year is, underneath, a bet on a bond rally the bond market is not pricing.

What 7.40% Costs, and What Waiting Actually Buys

Percentages are abstract, so here is the monthly cheque. Principal and interest on a $400,000 loan over 30 years, at each rate that matters in this piece.

Waiting for Fannie Mae's 6.8% saves $162 a month and the market gives it well under even odds for this year. Waiting for the MBA's 6.5% saves $241 a month at 18% probability. Waiting for the February low saves $376 at under 3%. Those are the actual trade-offs, priced, rather than the "rates should come down next year" that the SERP offers.

The asymmetry is worth naming too. The market puts 6.3% on the 10-year reaching 6.00%, which on today's spread is an 8%-handle mortgage and $165 a month more than today. The distribution is not symmetric around the forecast consensus, and it has more weight above today's rate than below it.

What Would Change This Mortgage Rate Prediction

A mortgage rate prediction is only useful if it says what would falsify it. Three things would have to break in the right direction, and each has a date attached.

1. The long end has to stop selling off. This is the big one and it is not in the Fed's gift. The 10-year has added 103 basis points in 2026 and the market's modal outcome is that the 2026 peak is still ahead, with 69% on 5.4% printing. Watch the Treasury curve daily rather than FOMC statements.
2. Inflation has to cooperate on October 14. September CPI is the next hard print and the committee gets only two more before the December vote. Fannie Mae's own forecast has core PCE at 3.4% for 2026, which is not a number that lets the long end rally.
3. The spread has to compress from an already-tight level. At 218 basis points today against a 196 average for the year, the near-term direction of the spread is more likely to help than hurt. But it cannot deliver a 6-handle on its own, as the table above shows.
4. December 9 has to disappoint the hawks. The market has 74.5% on a December hike and that has held within three points for two weeks through a weak payrolls report. A no-change December would pull the front end down, though it would do less to the 10-year than most people assume.

My read, stated plainly so it can be checked later: the 7.50% print happens before year-end, and the Freddie Mac survey does not get back to 6.50% or lower in 2026. That is the market's view too, at 68.5% and 82% respectively, and I have not found an argument on page one of the SERP that engages with either number. If the October 14 CPI comes in soft and the long end rallies 30 basis points, I will be wrong about the first one, and the contract will tell you before any forecaster revises.