ACP Lending

Completions Fell 27% in a Year. Everything Else in Thursday's Release Was Noise.

By Steve Waller · September 18, 2026

Completions Fell 27% in a Year. Everything Else in Thursday's Release Was Noise.

Census reported Thursday that August housing completions ran at a 1,128,000 annual rate, down 27.1% on the year - the only movement in the release that clears its own margin of error. Permits held at 1,394,000, up 3.5% on the year, and the South took 745,000 of them. Freddie Mac's weekly survey jumped 19 basis points to 6.95% while the ten-year fell back to 4.94% from Wednesday's 5.01%.

What happened

Two things landed Thursday morning. Census published New Residential Construction for August at 7:30am Central, and Freddie Mac's weekly survey caught up with the last two weeks of the bond market.

Freddie Mac put the 30-year fixed at 6.95%, up from 6.76% a week earlier and 6.26% a year ago. That is a 19 basis point jump in a single weekly print and 69 basis points above the same week last year. The 15-year went to 6.26% from 6.09%. Sam Khater's comment on the release was that the 30-year "continues to fluctuate as markets assess economic data."

The ten-year Treasury went the other way. Tuesday closed at 5.00%, Wednesday at 5.01%, and Thursday at 4.94%, with the two-year down from 4.74% to 4.67% on the same session. The FOMC's quarter point took effect Thursday, putting the target range at 3.75% to 4.00% with the discount rate at 4.00%, and the long end finished the week below where it sat before the meeting. Mortgage News Daily's daily average, which moves before the weekly survey does, read 7.19% Thursday after touching 7.24% on Wednesday.

So the survey rose 19 basis points in a week when the index it follows fell. That is a lag catching up, not a new move. The number to carry into next week is 4.94%.

The one number in the Census release that is real

Most of Thursday's construction report cannot be told apart from no change at all, and Census says so in its own margins. This is worth spelling out, because the headlines written off this release will not.

Housing starts ran at a seasonally adjusted annual rate of 1,275,000, down 2.6% on the month - against a confidence interval of plus or minus 12.0%. Single-family starts were 918,000, up 7.6% on the month, plus or minus 14.0%. Year over year, total starts were down 1.2% against an interval of 10.8%. None of those tell you the direction of anything. A story this weekend saying builders pulled back in August, or that single-family rebounded, is reading noise out loud.

Completions are different. They came in at 1,128,000, down 11.9% on the month against a 9.7% interval, and down 27.1% on the year against an 8.9% interval. Single-family completions were 816,000, down 22.9% on the year. Those clear their own error bars with room to spare.

One statistically significant movement in the whole release, and it is that far fewer houses are being finished than a year ago.

Why a finished house matters more than a started one

A start is a hole in the ground. It competes with nobody. A completion is a brand new house with a builder incentive attached to it, standing between your renovated resale and the same buyer.

Builders have spent two years buying that buyer with rate buydowns, and a buydown can only be hung on a house that exists. Completions are deliveries happening now, not a forecast - so a quarter fewer of them is a quarter less of the inventory those incentives can be attached to, in the same months your current rehabs come to market.

That is not a reason to pay more for a house. Nothing that happens to a builder's delivery schedule improves a deal that did not pencil on Monday. It is a reason to be less pessimistic about the exit than the rate alone suggests, and to read a competing new-build listing carefully before assuming it sets your ceiling.

What is still in the pipeline

Permits did not collapse. 1,394,000, down 2.7% on the month but up 3.5% on the year, with single-family at 878,000, up 1.3% on the year. Permits are not a survey - Census counts them - so there is no sampling error to argue about here.

The South took 745,000 of those permits, 527,000 of them single-family, and 658,000 of the starts. Better than half the country's single-family housing production is happening in this region, which is the backdrop for every Texas number below.

Put the three series together and the shape is this: paper is still being pulled at a rate above last year, ground is being broken at roughly last year's pace, and the distance between breaking ground and handing over keys has stretched. Something between permit and certificate of occupancy is slower than it was. Cost is a reasonable suspect rather than a proven one - the Bureau of Labor Statistics had final demand prices up 0.4% on the month and 5.4% over twelve months in the August PPI, with diesel alone up 24.1% - but the release does not tell you which input is binding, and neither will anyone who claims otherwise this weekend.

Houston, against the national market

HAR's August report: 38,947 active single-family listings, 5.3 months of supply, a median of $330,000 - five thousand dollars below August 2025 - and 54 days on market against 52 a year ago. Closings came in around 7,100, down 11.5% on the year.

Now set that beside what NAR published on 10 September. Existing-home sales nationally at a 3.98 million annual rate, down 2.0% on the month and 1.2% on the year, a median of $429,100, 4.9 months of supply and 31 days on market. The South ran at 1.84 million with a median of $366,500.

Thirty-one days nationally. Fifty-four in Houston. Houston is not a softer version of the national market, it is a different one - more supply, a lower price, and a wait almost twice as long. A purchase test built on national holding-period assumptions will be wrong here, and wrong in the direction that costs money, because every extra month is carry you pay and the seller does not.

The margin this lands on

ATTOM's most recent flipping report, covering the first quarter, counted 64,348 flips nationally at 8% of all sales, with a typical gross profit of $66,000, a gross return of 25.4%, and 165 days from purchase to resale.

Texas gross margins in that same report: 7.2% in Houston, 5.1% in San Antonio, 4.3% in Dallas, 2.0% in Austin. Gross, before a dollar of rehab, carry or selling cost.

A 165-day hold started today ends in March. Between now and then two inputs moved this week and they moved against each other: the exit rate went up on the survey, which reaches every financed buyer, and new-home deliveries went down, which only reaches the buyers who were shopping new. They do not cancel. On a 7.2% gross margin they are also not the kind of thing you can wave through - that is the difference between this market and the one the national headline describes.

The week ahead

What we are financing

Nothing this week changes the arithmetic on a deal. It moved two of the inputs, and they are not the two most people are watching.

If you are underwriting a flip to a spring resale, the exit rate on your spreadsheet is probably August's, and the weekly survey has not finished catching up to the market even now. Rerun it at this week's number, then look at the competing new-build listings in the subdivision rather than assuming they will still be there.

If you are underwriting a ground-up file, Thursday's completions figure is the more useful half of the release. The distance between permit and delivery has stretched, and a construction budget built on last year's schedule runs short on interest reserve before it runs short on anything else. That is a structuring question, and it is easier to answer before the draw schedule is set than after.

And if you own a stabilized rental, this was a week when the long end actually fell. Both calculators on the site take the rate as an input - the deal calculator for the flip, the DSCR calculator for the rental takeout - so rerunning either is a two-minute job rather than a guess.

Sources

Market data is from public sources and is provided for information only. Nothing here is an offer of credit, a rate quote, or investment advice. Financing is subject to underwriting, lender approval, appraisal or valuation, title, insurance, property eligibility and program availability.

All briefs

Financing an investment property in this market? Start with a conditional pre-approval, or run the numbers yourself.

Start a Request Get Pre-Approved Deal Calculator

Every brief also posts on Facebook and LinkedIn — follow there for the short version.