Market Analysis • February 18, 2026
Housing Starts Up 6.2%? The Confidence Interval Says “Maybe Not”
On February 18, 2026, Census/HUD released December’s New Residential Construction. The headline cheered a 6.2% m/m rise in housing starts to 1,404,000 SAAR, but the fine print cut the music: the 90% confidence interval is ±10.7%, which includes zero. In other words, the “bounce” may be statistically indistinguishable from no change.
Here’s what the release actually shows:
- Starts +6.2% m/m (1,404,000 SAAR), CI ±10.7%; single-family starts +4.1%, CI ±13.9%—not statistically different from zero.
- Permits +4.3% m/m (1,448,000 SAAR), but single-family permits fell -1.7% to 881,000; December single-family starts (981,000) exceeded permits (881,000)—a backlog drawdown, not fresh demand.
- Multi-family (5+ units) permits +18.1% m/m and +18.7% y/y; total permits still -2.2% y/y, and single-family authorizations down -10.9% y/y.
- Full-year 2025 weakened: permits -3.6% (1,425,200), starts -0.6% (1,358,700), completions -7.9% (1,497,800) vs 2024.
- Backlog mixed by adjustment: SA authorized-but-not-started -1.5% m/m (±4.0 pp); NSA backlog up 249.5k → 254.5k m/m but down 273.6k → 254.5k y/y.
The press release leaned on “above revised November” framing and late-year upticks but downplayed statistical uncertainty, composition shifts toward multi-family, and a weaker 2025 supply picture.
The Headline Bounce That Might Not Exist
The narrative: momentum into year-end. The math: maybe, maybe not.
- December’s starts +6.2% and completions +2.3% both carry 90% CIs that include zero (±10.7% for starts; ±11.8% for completions). Single-family tells the same story: starts +4.1% (±13.9%) and completions -0.1% (±13.8%).
- Headlines declared increases; the statistics whispered “inconclusive.” For professional investors, that’s not pedantry—it’s risk calibration. If the measured change is inside the error band, treat it as trend-agnostic.
The revision angle
The release also acknowledges average preliminary revisions up to 1.9%. Yet the textual emphasis sits squarely on month-over-month changes. Stacking “could be zero” moves on top of revision-prone estimates is not a foundation for durable conclusions.
Backlog, Not Breakout: Permits, Starts, and Composition
December’s split personality is the real story.
- Total permits +4.3% m/m to 1,448,000 SAAR, but single-family permits -1.7% m/m to 881,000. Meanwhile, single-family starts 981,000 outpaced permits 881,000.
- Translation: builders are converting backlog in single-family rather than greenlighting new projects at the same pace. That’s throughput, not ignition.
- The multi-family engine did the heavy lifting. 5+ unit permits jumped +18.1% m/m and +18.7% y/y. The headline rise in total permits is a multi-family story wearing a “housing” label.
December indicators at a glance (SAAR unless noted)
| Metric | Level (Dec) | m/m % change | 90% CI on m/m | y/y % change |
|---|---|---|---|---|
| Total Permits | 1,448,000 | +4.3% | — | -2.2% |
| Single-Family Permits | 881,000 | -1.7% | — | -10.9% |
| Total Starts | 1,404,000 | +6.2% | ±10.7% | — |
| Single-Family Starts | 981,000 | +4.1% | ±13.9% | — |
| Total Completions | 1,525,000 | +2.3% | ±11.8% | — |
| Single-Family Completions | — | -0.1% | ±13.8% | — |
| 5+ Unit Permits (composition) | — | +18.1% | — | +18.7% |
Note: Levels for 5+ unit permits not specified here; percent changes per release.
Seasonal adjustment shapes the y/y story
On a SAAR basis, December permits were -2.2% y/y (1,448,000 vs 1,480,000). On a not seasonally adjusted (NSA) basis, raw December permits were higher y/y (116.9k vs 112.9k). Both can be true; only one made the narrative. Investors should watch the composition and the adjustment filter—both materially alter conclusions.
The Year That Shrunk: 2025’s Supply Slippage
Zooming out removes the spin. 2025 delivered fewer permits, starts, and notably fewer completions than 2024. The late-year bump didn’t close the gap.
| Full-Year 2025 vs 2024 | 2025 Level | % change vs 2024 |
|---|---|---|
| Permits | 1,425,200 | -3.6% |
| Starts | 1,358,700 | -0.6% |
| Completions | 1,497,800 | -7.9% |
- Completions are where rubber meets road for actual supply: -7.9% is not a rounding error. Even with completions rising Oct → Nov → Dec (1,430,000 → 1,490,000 → 1,525,000 SAAR), the year’s reality is a downshift.
- Permits traced a softening arc—starting near 1.46 million SAAR in January, bottoming at 1.33 million in August, and ending December at 1.448 million, still below December 2024’s 1.48 million.
Structural tilt toward multi-family
The shift is no longer anecdote. In 2025:
- Single-family permits fell 7.4%.
- 5+ unit permits rose 4.3%.
- December reinforced it: 5+ permits +18.1% m/m, +18.7% y/y, while single-family authorizations fell m/m and -10.9% y/y.
Owner-occupied supply is where affordability relief typically comes from; it’s also where the pipeline looks softest.
Adjustments and Regions: The Hidden Story
Backlog and regional dynamics complicate the near-term supply outlook.
- Backlog (authorized-but-not-started) ended December down -1.5% m/m SA (±4.0 pp)—again, not statistically clean—while NSA backlog rose to 254.5k from 249.5k m/m and fell from 273.6k y/y. The pipeline is thinning year-over-year, but the month-to-month optics swing with the adjustment lens.
- Regionally (annual NSA, 2025 vs 2024), the Midwest held up (total +3.0%, 1-unit +0.1%), while the South (total -5.2%, 1-unit -8.5%) and West (total -1.9%, 1-unit -10.6%) weakened, with the Northeast also soft (total -7.7%, 1-unit -0.8%).
The release supplies the figures but not the narrative that ties them to supply where demand is deepest. And it offers no context on affordability, mortgage rates, or builder sentiment—omissions that matter when single-family authorizations are contracting while multi-family ramps.
What This Means for Markets
- Homebuilders and construction: The single-family pipeline is not accelerating. With single-family permits -1.7% m/m and -10.9% y/y, December’s starts outpacing permits signals backlog conversion. Expect delivery mix to favor projects already in the queue; land-light, spec-averse builders may defend margins, but topline unit growth looks constrained.
- Multi-family and rentals: 5+ permits up 18.1% m/m and 18.7% y/y point to continued multi-family resilience into 2026. This benefits multi-family developers, large GC’s tilted to mid/high-density, and select building products leveraged to concrete/steel over lumber.
- Supply and inflation optics: Completions -7.9% for 2025 tightens the near-term delivery outlook. With owner-occupied supply lagging, the relief valve for shelter inflation resides more in rentals than in single-family turnover—an asymmetry worth noting for macro positioning.
- Rates and MBS: The report doesn’t change the rates narrative by itself—but the composition shift and weaker annual completions argue against a sudden supply-led disinflation impulse in shelter. For MBS investors, watch coupon performance in regions where single-family pipelines have cooled most (South/West).
- Regional exposure: The Midwest shows relative permit stability; the South and West show notable single-family softness. Public builders with heavier Midwest exposure may have cleaner starts-to-permits trajectories in early 2026.
Positioning and what to watch
- Tilt toward operators and suppliers leveraged to multi-family starts; be cautious on pure-play single-family volume stories until permits turn.
- Focus on companies with healthy backlog-to-permit ratios and consistent conversion—December’s single-family start/permit gap won’t persist without new authorizations.
- Track the next two prints for:
The market loves a neat narrative. December’s release offered one—momentum into year-end. The data offered a different story: confidence intervals that erase the bounce, a single-family pipeline that’s thinning, and a multi-family sector doing the heavy lifting. For investors, the edge is in respecting the composition and the error bars, not the headline.