Two Mortgage Rates, One Date: 6.76% and 7.07%

On Thursday, September tenth, 2026, two different 30-year fixed mortgage rates were published for the same day.

Freddie Mac's Primary Mortgage Market Survey put the 30-year fixed at 6.76% for the week ending September 10, up from 6.71% the week before. Mortgage News Daily put its top-tier 30-year fixed at 7.07% on September 10.

Thirty-one basis points apart. Same product, same date on the label.

Neither is a mistake. They are two honest measurements of two different things, and the difference between them is almost entirely explained by how each is built. One is drawn from loans borrowers were actually applying for across that week. The other is a same-day read on what a very specific kind of borrower was being quoted.

This matters because the survey number is the one that ends up in headlines, and the headline is what people carry into a lender conversation. If you walked in on September tenth expecting 6.76% and were quoted something starting with a seven, nobody lied to you. You were comparing a quote against a statistic that was never meant to function as a quote.

Start with the survey, because what it is built from is most of the answer.

What replaced it is not a poll at all. The survey is now derived from mortgage applications that lenders submitted to Freddie Mac during the survey week. Not quotes. Applications.

That distinction is the whole mechanic. An application sits downstream of the quote that produced it: somebody was shown a rate, thought about it, decided to go ahead, and a file was assembled and sent. The rate riding on that application is the rate that actually moved a borrower to act, which is a different object from the number printed on a lender's rate sheet this morning.

So the survey describes the rate environment borrowers were transacting in across that week. It is a picture of decisions already taken, drawn from a population of real loans. Freddie Mac's research note comparing the current series with the legacy one puts the average difference between them at under ten basis points for both the 30-year and the 15-year. The change did not move the level of the index. It changed what the index is made of.

Set that against what the survey is built from and the gap stops being mysterious. The 7.07% is a rate sheet on one day, and that day was the highest of the three daily readings we can see. The 6.76% is drawn from applications submitted across the week, carrying the rates borrowers were actually shown while they were deciding, and for much of that week those were lower than Thursday's sheet.

A second difference works in the same direction. The daily figure prices a single top-tier scenario. The survey is built from whatever loans lenders actually submitted, across whatever range of borrowers submitted them. One is a specification; the other is a population. They would not be expected to print the same number even on a flat day.

That is how you get 6.76% and 7.07% published with the same date attached. Not because either is wrong, but because each is answering a different question.

The practical translation: in a rising week, the survey tells you where rates have been. It is not designed to tell you where they are this afternoon, and reading it that way is the error.

Thirty-one basis points sounds small. Here is what it costs, so it stops sounding abstract.

Take a $400,000 loan on a 30-year fixed, principal and interest only.

At 6.76%, the monthly payment is about $2,597. At 7.07%, it is about $2,680. The difference is roughly $83 a month.

Run that out: $83 a month is just under $1,000 a year, and about $29,900 across all 360 payments if the loan is held to maturity and never refinanced.

You can recompute any of that with a standard amortization formula and the two rates. The figure moves with loan size and it is not a prediction about your loan. The point is that the distance between a headline and a lock is not a rounding difference. A household that penciled in the survey figure and signed at the daily figure did not make a small error.

There is another assumption hiding in the 7.07%.

A top-tier daily rate is not the average of what everyone got. It is the rate available to a borrower whose file has nothing in it that costs money. In practice that means close to all of the following at once.

• A high credit score, comfortably inside the best pricing band rather than just over a threshold.
• A conforming loan amount, not a jumbo and not a niche product.
• Twenty percent down or better, so loan-to-value is 80% or lower and there is no mortgage insurance.
• Owner-occupied as a primary residence, not a second home or a rental.
• A single-family property, full documentation, and a debt-to-income ratio well inside guidelines.

Miss one of those and you are not being quoted the top-tier rate. That is not a penalty in any moral sense; it is how risk-based pricing works. But it means the daily number is still a rate for a narrow borrower profile.

Most borrowers are not that profile. Someone putting 10% down, or buying a duplex, or self-employed with two years of returns to average, or sitting a few points below the top score band, is priced off a different sheet. Which is why the honest answer to what the 30-year rate is today is that there isn't one. There is a rate for a file, and the file is yours.

Four things move an actual borrower away from both published figures.

The third is points and credits. Almost any rate can be bought down with discount points or pushed up in exchange for a lender credit, so two quotes can show different rates and be the same deal, or show the same rate and cost thousands apart. A rate compared without its cost is not a comparison.

The fourth is the lock. A quote is good for a period of thirty days, forty-five or sixty, and longer locks cost more. A rate that looked available on a Thursday is not owed to you the following Monday unless it was locked.

It is worth saying what the weekly survey is good for, because lagging the dailies in a fast week is not an argument against it.

It is the long series, and it is the right tool for the questions people actually ask about the general level of rates: are rates higher than last year, what has the direction been over six months. For that, a weekly figure grounded in loans that were really applied for is exactly what you want, and a daily series would be noise.

What it cannot do is function as a quote. The mismatch is not in the measurement, it is in the use, where a statistic describing a week of transactions gets read as a price available today.

The daily series has the opposite profile. It is close to the market, so it tells you which way things moved this week, but it prices one top-tier scenario and aggregates nothing, so a two-day move can look dramatic in a way a month of data would not support.

The survey tells you about the year. The daily tells you about the week. Neither tells you about your loan, and a borrower within sixty days of needing a rate should stop reading both and start collecting quotes.

One more reading habit, because this pattern repeats every time rates move quickly.

When the market moves late in a week, most of the application flow feeding that Thursday's survey was already generated, off rates quoted before the move. The following week's survey picks it up, and the week-over-week change can look larger than anything that happened inside that second week, because part of it is the earlier move arriving. That is a property of a series built from transaction flow, not a second event.

So you can see a headline saying the survey rate fell in a week when quotes on your file went up, and neither statement is false. They cover different days. Before concluding your lender is out of line with the market, check what period your number covers.

Be careful in the other direction too. A borrower who reads a rising daily series and hurries into a lock has decided on two days of data.

Both numbers published for September tenth were correct. 6.76% came from applications lenders submitted to Freddie Mac across that week, carrying the rates borrowers were shown while they were deciding. 7.07% was a lender rate sheet on September tenth, priced for one top-tier scenario, after the daily series climbed from 6.89% on the eighth through 6.97% on the ninth. The 31-basis-point gap is what you get when you hold a week of transactions next to a single day's sheet.

Neither one is your rate. On a $400,000 loan the distance between them is about $83 a month, close to $1,000 a year, and roughly $29,900 over a full 360-payment term.

A survey rate is a fact about the mortgage market. A quote is a fact about you. Only one of them is for sale.

1. Why were two different 30-year mortgage rates published for September 10, 2026?

Freddie Mac's weekly survey reported 6.76% for the week ending September 10, 2026, while Mortgage News Daily reported a top-tier daily rate of 7.07% for September 10 itself. The survey is derived from mortgage applications lenders submitted to Freddie Mac during the week; the daily figure is a lender rate sheet for one top-tier scenario on that day. Both are accurate measurements of different things.

2. Is the Freddie Mac survey a poll of lenders?

3. What does 31 basis points cost on a mortgage?

On a $400,000 30-year fixed, principal and interest only, 6.76% is about $2,597 a month and 7.07% is about $2,680. That is roughly $83 a month, just under $1,000 a year, and about $29,900 over 360 payments if the loan is held to maturity.

4. What does top-tier mean in a daily mortgage rate?

It assumes a file with nothing in it that costs money: a high credit score well inside the best pricing band, a conforming loan amount, 20% or more down so loan-to-value is 80% or lower with no mortgage insurance, owner-occupied as a primary residence, a single-family property and full documentation. Most borrowers do not meet all of those.

5. Is the Freddie Mac survey rate wrong?

No. It is the long-run series for comparing rate levels across months and years, and it is grounded in loans borrowers actually applied for. It is not built to function as a quote available today, and reading it that way is the error.

6. Can the survey rate fall in a week when my quote goes up?

Yes, and neither statement is false. A late-week move hits lender rate sheets immediately, but it reaches the survey only through applications still being submitted, so it shows up more fully in the following week's number. That can make the following week's change look larger than what happened inside it.

7. What should I do instead of reading published rates?

Collect quotes on your own file from more than one lender on the same day, for the same loan amount, down payment and lock period, with points and lender credits shown beside each rate. That is the only number you can lock.