Market Brief

The July new home sales report produced the worst monthly number since January, and the Census Bureau's own confidence interval says it cannot rule out that sales rose.

July new home sales came in at 607,000, down 10.5% from June. The margin of error spans zero, so the drop is not statistically significant.

The daily read on the numbers that set Florida housing costs: tax policy, insurance, mortgage rates and construction, and what they mean for a buyer or seller in Northeast Florida this week.

The number everyone quoted

The Census Bureau and HUD released July new residential sales on Tuesday, August 25. Sales of new single-family houses ran at a seasonally adjusted annual rate of 607,000, which is 10.5% below the revised June rate of 678,000 and 6.3% below the July 2025 rate of 648,000.

That is the figure that travelled. It is the lowest monthly reading since January and it landed in a week when the housing narrative was already leaning negative.

The number almost nobody quoted

Census publishes a 90% confidence interval next to every one of those percentages, and for the monthly change it is plus or minus 14.0 percentage points.

A 10.5% decline with a 14-point interval means the true change is somewhere between roughly a 24.5% decline and a 3.5% increase. The range contains zero. In the Bureau's own language, printed in the explanatory notes of the same release: if the range contains zero, "it is uncertain whether there was an increase or decrease."

The year-over-year comparison is in the same position, at 6.3% down with an interval of plus or minus 19.6 points. So is the year-to-date figure, 4.1% down with an interval of plus or minus 5.6. And so is the South region, where sales fell 13.0% with an interval of plus or minus 17.8 points. Florida sits in that South region, which is the number a Florida reader would most want to lean on, and it is the one with the widest relative margin of the four.

Census also states plainly that it takes four months to establish a trend in new home sales, and that the preliminary seasonally adjusted total is revised by about 5.0% on average.

None of that makes 607,000 a good number. It makes it an uncertain one, and the distinction matters if you are about to price a house on it.

What in the release is statistically solid

One headline figure in the report clears its own error bar, and it is the inventory line.

New houses for sale at the end of July stood at 488,000, up 1.9% from June's 479,000 against a confidence interval of plus or minus 1.2 points. The range excludes zero, so that increase is real in a way the sales decline is not.

Two supporting details in the same tables point the same direction and are worth more than the headline:

• Homes not yet started in the for-sale count reached 115,000 in July, against 96,000 a year earlier. Builders are carrying more unstarted permitted inventory than they were.
• The median time a completed new home sat unsold was 3.2 months in July, against 2.6 months in July 2025. That is a slower clock on finished product, measured directly rather than inferred from a rate of sale.

Inventory measurements are counts. Sales rates are estimates from a sample of building permits. When the two disagree in reliability, the count is the one to build on.

The price line is a mix shift, not a discount

The median price of a new house sold in July was $393,800, down 2.3% from June and down 0.9% from a year ago. The average price was $508,800, up 4.1% from June and up 5.4% from a year ago.

Median down, average up, in the same month, on the same sales. That combination describes a change in what sold rather than a change in what things cost. The price distribution in the release supports it: the $300,000 to $399,999 band took 34% of July sales against 28% in June, while the $600,000 to $799,999 band fell from 14% to 11%.

Both price changes carry intervals that contain zero as well, so neither is significant on its own. The mix shift is the more defensible reading, and it is the one that matters for a builder deciding what to start next.

Why this lands differently in Northeast Florida

New construction is a much larger share of this market than of the national one, so a national new-home number gets read here as a local signal more readily than it should be.

Two cautions. The first is the statistical one above: the South region figure is directional at best. The second is that a national sales rate says nothing about which subdivisions are absorbing and which are not, and that varies street by street in Duval, Clay and St. Johns.

What the inventory side of this release does support is an observation we have made from our own realMLS records: finished, unsold builder inventory moves on a materially slower clock than product sold before completion, and the discounting that clears it does not always show on the closing statement. Our work on shadow inventory and builder discounts covers the local version of that, and the Census median-months-for-sale line moving from 2.6 to 3.2 is the national echo of it.

Mortgage rates gave the report no help and no hindrance. Freddie Mac's survey for the week ending August 27 put the 30-year fixed at 6.66%, up from 6.65% the week before and 6.56% a year ago, with the 15-year at 5.98%. Rates have been close to still for weeks, which is worth remembering before attributing a one-month sales move to financing costs.

What we are not saying

We are not saying new home sales were fine in July. The point is narrower and duller than that: one month of this particular series cannot carry the weight the headline put on it, by the publisher's own standard, and a reader who saw "new home sales plunge 10.5%" was given a number without the range that belongs to it.

The next release, covering August, is scheduled for September 24. Two more prints after that and the four-month trend Census refers to will exist. Until then, the inventory count is the part of this report worth acting on.

People also ask

Did new home sales really fall 10.5% in July 2026?

That is the Census Bureau's point estimate, but the 90% confidence interval around it is plus or minus 14.0 percentage points, which means the true change could be anywhere from about a 24.5% decline to a 3.5% increase. Because the range includes zero, Census's own explanatory notes classify the change as not statistically significant. The sales rate of 607,000 is also preliminary and is revised by about 5.0% on average. Census states that it takes four months to establish a trend in this series.

What is the current supply of new homes for sale?

There were 488,000 new houses for sale at the end of July 2026, representing 9.6 months of supply at the July sales rate. The inventory increase of 1.9% from June was statistically significant, unlike the sales decline. The months' supply figure rose from 8.5 in June, but that change carries a confidence interval of plus or minus 21.3 points and is not significant, largely because the denominator is the uncertain sales rate.

Why did the median new home price fall while the average rose?

In July 2026 the median new home sale price was $393,800, down 2.3% from June, while the average was $508,800, up 4.1%. Median and average moving in opposite directions on the same set of sales points to a change in the mix of homes sold rather than a change in prices. Census data shows the $300,000 to $399,999 band rose from 28% to 34% of sales while the $600,000 to $799,999 band fell from 14% to 11%. Both price changes carry confidence intervals that include zero.

Sources

• U.S. Census Bureau and U.S. Department of Housing and Urban Development, *Monthly New Residential Sales, July 2026*, Release CB26-128, August 25, 2026.
• Freddie Mac Primary Mortgage Market Survey, week ending August 27, 2026.

Disclosure: Jon Brooks is a co-founder of Momentum Realty, a Florida real estate brokerage that owns and funds Florida Housing Intelligence. Momentum has a financial interest in Florida real estate transactions, including in markets named here. See our editorial standards.