← All articles

🏙️ Urban

151,100 Fewer Homes Finished Through October. The 9.9% Start Rise Includes Zero.

January through October completions are 151,100 homes behind 2025, and that gap clears its error bar. The 9.9% rise in starts does not.

Late-afternoon aerial of a half-built suburban tract, with wood frames, lumber stacks, concrete slabs, and a distant crane.

Nine point nine percent.

Privately owned housing starts in October ran at a seasonally adjusted annual rate of 1,309,000, which the Census Bureau and HUD release this morning put 9.9 percent above September's revised 1,191,000, and the parenthesis on that line is ±16.8 percent, with an asterisk.

Subtract 16.8 from 9.9.

Doing so puts the lower bound at −6.9 percent and the upper bound at 26.7 percent, so the 90 percent interval includes a decline as easily as a boom, and 16.8 divided by 9.9 says the margin is 1.70 times the point estimate. The PDF says a range that contains zero is not statistically significant, meaning there is not enough evidence to conclude the actual change differs from zero. Single-family starts, 864,000 against September's 831,000, carry the same mark: 4.0 percent (±9.0 percent).

Permits fell, and that line has no asterisk.

Authorizations were 1,295,000, 7.4 percent under September's revised 1,398,000 and 8.7 percent under the October 2025 rate of 1,418,000, while single-family permits were 772,000, down 8.7 percent from 846,000 and 12.9 percent from 886,000 a year earlier. FRED's PERMIT prints 1,295 for October and HOUST prints 1,309, matching the release, and PERMIT1 prints 772. On that single-family series January 2023 was 755, and every month from February 2023 through September 2026 printed higher, so October is the low since that January.

No asterisk, because explanatory notes in the PDF say Table 1 is not subject to sampling error: Census builds the report from two surveys, a FAQ on the 2027 redesign says the Building Permits Survey has collected from the full universe of known permit-issuing jurisdictions since January 2022, and starts come from the Survey of Construction, a sample, which is why Tables 3 and 5 carry intervals. The permit quantity response rate is 76.6 percent, so nonsampling error remains.

One hundred three thousand.

September's permit rate minus October's is 1,398,000 minus 1,295,000, and the South went from 756,000 to 675,000, a drop of 81,000, which divided by 103 is 78.6 percent of the national decline. West dropped 13,000, the Northeast dropped 10,000, and the Midwest rose 1,000, and those four pieces sum to 103,000. Single-family permits fell 74,000, from 846,000 to 772,000, of which the South was 38,000 (51.4 percent) and the West was 25,000 (33.8 percent), with western single-family permits falling from 189,000 a year ago to 148,000, the table's −21.7 percent, and the single-family permit rate fell in all four regions.

Rates are seasonally adjusted annual rates except the year-to-date rows, which are not seasonally adjusted thousands of homes, January through October. Percents and margins are the printed figures in the November 18, 2026 Census release, CB26-187. Census computes percent changes on unrounded data. An interval excludes zero when the printed margin is smaller than the absolute value of the printed change.
ChangePrinted move90 percent interval
Total starts, October vs September+9.9% (±16.8)−6.9% to +26.7%, includes zero
Single-family starts, month+4.0% (±9.0)−5.0% to +13.0%, includes zero
Five-or-more-unit starts, month+25.5% (±50.6)−25.1% to +76.1%, includes zero
Total permits, month−7.4%, no sampling intervalTable 1, nonsampling error remains
Single-family permits, month−8.7%, no sampling interval772,000, low on PERMIT1 since January 2023
Single-family starts, year to date−4.6% (±3.2)−7.8% to −1.4%, excludes zero
Completions, year to date−12.3% (±4.4)−16.7% to −7.9%, excludes zero
Single-family completions, year to date−10.7% (±4.3)−15.0% to −6.4%, excludes zero
Five-or-more completions, year to date−15.6% (±9.6)−25.2% to −6.0%, excludes zero
Single-family completions, October vs year ago−10.7% (±10.4)−21.1% to −0.3%, excludes zero by 0.3 points

Most of that bounce is the noisy slice. Five-or-more-unit starts went from 345,000 to 433,000, up 88,000 of the 118,000 increase in total starts, which is 74.6 percent, and that component's own change is 25.5 percent (±50.6 percent). Single-family starts rose 33,000, and 88,000 plus 33,000 overshoots the 118,000 by 3,000 because two-to-four-unit starts are suppressed, and HOUST1F prints 864 for the single-family rate.

Starts are outrunning new permits: 864,000 minus 772,000 is 92,000, or 11.9 percent of the permit rate, and authorized-but-not-started single-family units stood at 151,000, which is 2.10 months against a monthly start rate of 72,000, versus 1.89 months a year earlier when the stock was 141,000 and starts were 893,000. Hold the 92,000 gap and assume nothing is canceled, and 151,000 divided by about 7,667 a month is 19.7 months until that stock hits zero.

Finished homes are where the interval clears.

January through October completions, not seasonally adjusted, were 1,080,800 against 1,231,900 a year earlier, and the release prints −12.3 percent (±4.4 percent), an interval from −16.7 percent to −7.9 percent and a level gap of 151,100 homes. Apply those printed edges to the printed 1,231,900 base, an approximation because Census used unrounded inputs, and the shortfall runs from about 97,320 homes to about 205,727. January 1 through October 31 is 304 days, and 151,100 divided by 304 is 497 fewer finished homes a day. COMPUTSA prints the seasonally adjusted October rate at 1,272, and that month-to-month move of 6.5 percent (±13.2 percent) includes zero, while the ten-month decline does not.

Single-family year-to-date starts clear zero on the downside too: 765,900 against 802,600, printed −4.6 percent (±3.2 percent), a point gap of 36,700 homes, and applying the edges to 802,600 puts the shortfall between about 11,236 and 62,603. October single-family completions versus a year earlier print −10.7 percent (±10.4 percent), an interval from −21.1 percent to −0.3 percent. Clearing zero by 0.3 points on the rounded figures is too close to lean on. Units under construction match the completion rate at 1,272,000, so unfinished stock equals 12.0 months of October's pace, against 11.1 months a year earlier.

Divisions use the printed thousands in the November 18 release and, for the rate history, FRED series PERMIT1 and MORTGAGE30US as pulled November 18, 2026. Payment factors assume a 30-year fully amortizing loan with no points, taxes, or insurance.
WorksheetArithmeticResult
Starts interval9.9 − 16.8, and 9.9 + 16.8−6.9% to +26.7%
Margin versus the point estimate16.8 / 9.91.70 times
South's share of the permit drop81,000 / 103,00078.6%
Multifamily share of the start bounce88,000 / 118,00074.6%
Single-family starts over permits(864,000 − 772,000) / 772,00092,000, or 11.9%
Backlog if that gap held151,000 / (92,000 / 12)19.7 months
Completion shortfall per day151,100 / 304497 homes
Optimistic edge of that shortfall1,231,900 × 0.079about 97,320 homes
Unfinished stock, in months of completions1,272,000 / (1,272,000 / 12)12.0 months
Payment per $100,000 at 7.11% vs 6.24%standard 360-month factor$672.71 vs $615.07, a gap of $57.64, or 9.37%

After October, Freddie Mac put the 30-year fixed rate at 7.11 percent as of November 12, down from 7.18 percent and up from 6.24 percent a year earlier, and MORTGAGE30US matches that pair and the archive: 7.40 percent on October 8 and October 15, then four declines to 7.11, or 29 basis points under the October high, while 7.11 minus 6.24 is 87 basis points and 7.11 divided by 6.24 is 1.139, or 13.9 percent above the year-ago rate. September 24 printed 7.03 percent, so Freddie's "lowest in seven weeks" is not a low against late September.

A payment is not the rate. At a monthly rate of 0.0711 divided by 12, the 360-month factor times $100,000 is $672.71, against $615.07 at 6.24 percent, a gap of $57.64, or 9.37 percent, smaller than the 13.9 percent gap in the rate because amortization does not scale one-for-one. October's Thursday prints were 7.28, 7.40, 7.40, 7.32, and 7.24, so November's 7.11 cannot explain an October permit.

Builders had already marked the month down. The NAHB/Wells Fargo index, out November 17, fell four points to 25, which NAHB calls the lowest since April 2012, with sales at 28, expectations at 30, and traffic at 16. Weights on NAHB's methodology page are 0.5920, 0.1358, and 0.2722, and 0.5920 times 28 plus 0.1358 times 30 plus 0.2722 times 16 equals 25.005, which rounds to 25. Traffic at 16 instead of 50 removes 9.25 points. NAHB says 46 percent of builders cut price, a post-Covid high, the average cut held at 6 percent for an eighth month, and 68 percent offered incentives for a 20th straight month at or above 60 percent.

What You Can Do

Buying, lending, or waiting on a project, ignore the 9.9 percent headline. Watch single-family permits at 772,000 and the year-to-date completion gap of 151,100 homes, whose nearer edge is still about 97,320, and ask whether a promised start is actually in the October permit count, because single-family permits fell in all four regions and 78.6 percent of the national permit drop was the South. Next construction release is December 17, on the economic briefing room calendar, and Freddie's next rate print is Thursday. A start rate whose interval runs to −6.9 percent is noise. 497 fewer finished homes a day is the tape.

Limitations

Nothing here is a price forecast or a shortage count. The 151,100 compares January through October with the same months of 2025, Census computes percents on unrounded data, and applying a printed margin to a printed base only approximates the home-count interval. October's sample figures are preliminary, September is revised, and the release says those preliminary estimates are revised 3.8 percent or less on average.

Table 1's missing asterisk is not zero error: the quantity response rate is 76.6 percent, the 19.7-month drawdown assumes the 92,000 gap persists with nothing canceled, and two-to-four-unit starts are suppressed. Payment math is a 30-year fully amortizing loan with no points, taxes, or insurance, at Freddie's conforming, 20-percent-down profile, not a quote on a house. NAHB's "lowest since April 2012" is NAHB's claim. Per the FAQ, this October print predates the January 2027 redesign, non-permit areas were about 0.5 percent of 2026 starts and 0.4 percent of completions, and the next release can still revise October.

The Strongest Case Against This Analysis

Grant the timing argument its full strength. Census says a permit trend can take three months and a start or completion trend six, so one ugly permit month can snap back, and a 9.9 percent start increase can be real even when the sample cannot prove it. Builders are drawing 151,000 authorized-but-not-started single-family units, up from 141,000, so starts can exceed new permits without a collapse, completions describe projects begun earlier, rates have fallen for four weeks, 68 percent of builders are offering incentives, and the multifamily slice that is 74.6 percent of the bounce swings hard. If November single-family permits jump back over 850,000, October's 772,000 will look like January 2023's 755, a low that did not stick, and December 17 is the fair test.

Measured permits do not need that recovery to show up later. They fell 7.4 percent with no sampling asterisk, single-family permits fell in all four regions, and 78.6 percent of the national drop was the South, while year-to-date completions sit below last year by a gap whose nearer edge is still about 97,320 homes. November can still rebound, but October's forward count and ten months of finished homes did not print one.

The Bottom Line

October starts rose 9.9 percent inside an interval from −6.9 percent to +26.7 percent, so the headline bounce includes zero, and 74.6 percent of the level increase was five-or-more-unit buildings. Permits fell 7.4 percent and single-family permits fell to 772,000, the lowest PERMIT1 month since January 2023. Completions through October are 151,100 homes behind last year, about 497 a day. Cranes in a start-rate headline are a sample. Permits and finished homes are the tape.

Related