HUMANITYVECTOR

USA · snapshot 2026-09-07 22:35 UTC · data to 2026-07-01

US housing construction

Is US home building recovering, and what is holding it back?

US housing starts was 1,239 thousands of units, seasonally adjusted annual rate in 2026-07-01, down 13.5 percent over one year, a move at the 57th percentile of its one-year changes.

DerivedModeledPrimary metric page

Section 1 of 9

What changed?

Observed and derived series only. Changes are latest minus the value one, five and ten years earlier.

AI narrative Modeled

US housing starts stand at 1239 thousands of units seasonally adjusted annual rate in 2026-07-01 [obs:169883], down 193 thousand units or 13.5 percent from 1432 one year earlier at the 57th percentile of one-year moves [obs:169871]. Building permits are 1433 thousands of units, up 2.4 percent year over year [obs:164614][obs:164602]. Residential construction spending is 871205 millions of dollars, down 7.3 percent [obs:182544][obs:182532].

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

Primary series

US housing starts

Observed

Privately owned housing units started, annual rate.

1,239 thousands of units, seasonally adjusted annual rate

2026-07-01

60 points, 2021-08-01 to 2026-07-01, 1,182 to 1,807 thousands of units, seasonally adjusted annual rate1,1821,8072021-08-012026-07-01
SpanFromChangePercentile
1 period1,415 (2026-06-01)-176 (-12.4%)89th
1 year1,432 (2025-07-01)-193 (-13.5%)57th of 798
5 years1,600 (2021-07-01)-361 (-22.6%)56th of 750
10 years1,239 (2016-07-01)0 (0.0%)0th of 690
United States · 811 observations since 1959-01-01 · obs 169883
Supporting series

US building permits

Observed

Permits lead starts by one to three months.

1,433 thousands of units, seasonally adjusted annual rate

2026-07-01

60 points, 2021-08-01 to 2026-07-01, 1,347 to 1,923 thousands of units, seasonally adjusted annual rate1,3471,9232021-08-012026-07-01
SpanFromChangePercentile
1 period1,374 (2026-06-01)59 (+4.3%)63rd
1 year1,400 (2025-07-01)33 (+2.4%)12th of 786
5 years1,666 (2021-07-01)-233 (-14.0%)29th of 738
10 years1,198 (2016-07-01)235 (+19.6%)19th of 678
United States · 799 observations since 1960-01-01 · obs 164614
Observed

Residential construction spending.

871,205 millions of dollars, seasonally adjusted annual rate

2026-07-01

60 points, 2021-08-01 to 2026-07-01, 821,402 to 986,455 millions of dollars, seasonally adjusted annual rate821,402986,4552021-08-012026-07-01
SpanFromChangePercentile
1 period882,821 (2026-06-01)-11,616 (-1.3%)77th
1 year939,959 (2025-07-01)-68,754 (-7.3%)62nd of 282
5 years816,883 (2021-07-01)54,322 (+6.6%)6th of 234
10 years483,205 (2016-07-01)388,000 (+80.3%)59th of 174
United States · 295 observations since 2002-01-01 · obs 182544

Section 2 of 9

How unusual is it?

The primary series against its own stored history (features_v1 rarity and surprise).

AI narrative Modeled

The latest value is at the 26th percentile of the series history [obs:169229][obs:169676]. The one-year change ranks at the 89th percentile of past changes with robust z of -1.7854 and one period since a larger move.

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

Derived
Latest value percentile
26th of 810 values
Latest change percentile
89th of 809 changes
Robust z of latest change
-1.79
Largest move
1 period since a larger move
Record level
no
Previous comparable move
2026-06-01: 233 thousands of units, seasonally adjusted annual rate (obs 169882, 169881)
Prior maximum
2,494 (1972-01-01, obs 169229)
Prior minimum
478 (2009-04-01, obs 169676)
Derived from the stored series; percentiles are rank percentiles with ties split.

Section 3 of 9

Why might it be happening?

Documented mechanisms, statistical model drivers, hypotheses and unknowns are kept apart.

AI narrative Modeled

Documented mechanism: a higher policy rate raises borrowing cost which reduces construction financing and residential building [mechanism:7006539d-a8a2-4038-86d0-112809badec4]. Construction labour availability is not stored as a series. Local permitting and zoning constraints do not appear in national totals.

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

Reviewed

Documented mechanisms

Statistical

Model-supported drivers

Modeled

No active forecast for this topic records driver contributions (the champion may be a baseline with no drivers).

Candidate or hypothesized relationships

Hypotheses

No candidate or hypothesized relationship touches this topic's entities.

Not settled by stored data

Unknowns

  • Construction labour availability is not stored as a series.
  • Local permitting and zoning constraints do not appear in national totals.

Section 4 of 9

What has not changed?

Fundamentals judged against the 10th to 90th percentile band of their own year-over-year changes (or of their levels when the history is short).

AI narrative Modeled

US unemployment rate is 4.1 percent with year-over-year change inside the 10th to 90th percentile band of past changes [obs:213832]. Average hourly earnings are 37.75 US dollars per hour inside the band [obs:393725]. Nonfarm payroll employment is 159075 thousands of persons inside the band [obs:178884].

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

SeriesLatestBasisNormal band (p10 to p90)Historical rangeReading
US unemployment rate Observed
Unemployment rate; household income supports demand.
4.1 percent
2026-08-01, obs 213832
year-over-year change -0.2 (-4.7%)
930 past changes
-1.1 to 1.72.5 (1953-05-01) to 14.8 (2020-04-01)
since 1948-01-01; latest at the 19th percentile of levels
within its normal band
US average hourly earnings, all private employees Observed
Average hourly earnings.
37.75 US dollars per hour
2026-08-01, obs 393725
year-over-year change 1.13 (+3.1%)
233 past changes
0.44 to 1.4720.04 (2006-03-01) to 37.75 (2026-08-01)
since 2006-03-01; latest at the 100th percentile of levels
within its normal band
US nonfarm payroll employment Observed
Nonfarm payrolls.
159,075 thousands of persons
2026-08-01, obs 178884
year-over-year change 603 (+0.4%)
1039 past changes
-1,114.6 to 3,451.829,923 (1939-01-01) to 159,075 (2026-08-01)
since 1939-01-01; latest at the 100th percentile of levels
within its normal band

Section 5 of 9

What absorbs the shock?

Each buffer names the series or graph entity that evidences it, or says it is not measured.

AI narrative Modeled

Inventory of homes for sale is not measured. Substitution options for softwood lumber list 0 substitutes. Lumber and fixture imports are not ingested. Rate sensitivity of buyers is not measured here. The federal funds rate is 3.63 percent [obs:169072] and the 10-year yield is 4.77 percent [obs:200327].

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

Observed

3.63 percent · 2026-08-01 · yoy -16.2% · obs 169072

The policy rate; mortgage rates follow with a lag.

Observed

4.77 percent · 2026-09-03 · yoy +13.0% · obs 200327

The 10-year yield anchors mortgage pricing.

Inventory

Not measured

No stored series evidences this buffer.

Homes for sale and months of supply are not ingested.

Substitution

Softwood lumber

graph

The graph records no substitute for this entity.

Substitutes recorded on the graph for framing materials.

Imports

Not measured

No stored series evidences this buffer.

Lumber and fixture imports are not ingested.

Demand elasticity

Not measured

No stored series evidences this buffer.

Rate sensitivity of buyers is not measured here.

Section 6 of 9

What response is underway?

Capacity pipeline, corporate pulses, capital spending and production series, with their year-over-year change.

AI narrative Modeled

Residential construction spending shows year-over-year change of -7.3 percent [obs:182544]. Cement production is down 1.2 percent year over year [obs:263553]. Nonfarm payroll employment rose 0.4 percent [obs:178884].

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

GroupSeriesLatestYear earlierChange
Capital spendingUS residential construction spending Observed871,205 millions of dollars, seasonally adjusted annual rate (2026-07-01, obs 182544)939,959 (2025-07-01)-68,754 (-7.3%)
ProductionCement production Observed84,000 thousand metric tons (2025, obs 263553)85,000 (2024)-1,000 (-1.2%)
WorkforceUS nonfarm payroll employment Observed159,075 thousands of persons (2026-08-01, obs 178884)158,472 (2025-08-01)603 (+0.4%)

The US pipeline and response measures are on the capacity page; bottleneck assessments are on bottlenecks.

Section 7 of 9

Scenarios

Base is the champion model's point estimate; upside and downside are the bounds of its 80 percent interval. No probability beyond the interval's own is assigned.

AI narrative Modeled

The 12-month base case is 1239 thousands of units with 80 percent interval 1073 to 1348 [forecast:d48837b4-6b4e-4441-af29-5f931ed00980]. The 6-month base case is 1239 with interval 1109 to 1350 [forecast:3ab4a587-ad7a-4d3c-b172-d30a6832c256]. The model assigns a 10 percent probability to values below the lower bound and above the upper bound.

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

12 month horizon · target 2027-07-01

US housing starts

Modeled
60 observed points to 2026-07, then 1 forecast with 80 percent intervals1,0731,4401,8072021-082022-042022-122023-082024-042024-122025-082026-042027-07
Solid: observed (thousands of units, seasonally adjusted annual rate). Dashed and shaded: model point estimates and 80 percent intervals from empirical residual quantiles. Filled dot: the record on this page. Green or red dot: the realised value, inside or outside its interval.

Base case

1,239 thousands of units, seasonally adjusted annual rate

The point estimate: the model's central path.

Upside

1,347.67 thousands of units, seasonally adjusted annual rate

Upper bound of the interval: what would need to go better than the central path.

Downside

1,072.99 thousands of units, seasonally adjusted annual rate

Lower bound of the interval: what would need to go worse than the central path.

The model assigns 80 percent to the interval; the site assigns no other probability.

Falsifiers and assumptions
  • If the first published value of us-housing-starts (USA) for 2027-07 is below 1,073 thousands of units, seasonally adjusted annual rate or above 1,348 thousands of units, seasonally adjusted annual rate, the 80 percent interval is falsified.
  • The model assigns a 10 percent probability to a value below 1,073 thousands of units, seasonally adjusted annual rate and a 10 percent probability to a value above 1,348 thousands of units, seasonally adjusted annual rate; interval misses should occur about one time in five over many forecasts, and a run of misses well above that rate falsifies the calibration.
  • Assumption: No structural break in us-housing-starts between the data cutoff and the target period.
  • Assumption: The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
  • Assumption: The most recent value is the best available guess: recent changes carry no persistent information.
Record d48837b4 · last_value (baseline is champion) · data cutoff 2026-09-07 22:16 UTC · created 2026-09-07 22:16 UTC · USA
6 month horizon · target 2027-01-01

US housing starts

Modeled
60 observed points to 2026-07, then 1 forecast with 80 percent intervals1,1091,4581,8072021-082022-042022-122023-082024-042024-122025-082026-042027-01
Solid: observed (thousands of units, seasonally adjusted annual rate). Dashed and shaded: model point estimates and 80 percent intervals from empirical residual quantiles. Filled dot: the record on this page. Green or red dot: the realised value, inside or outside its interval.

Base case

1,239 thousands of units, seasonally adjusted annual rate

The point estimate: the model's central path.

Upside

1,350.48 thousands of units, seasonally adjusted annual rate

Upper bound of the interval: what would need to go better than the central path.

Downside

1,109.35 thousands of units, seasonally adjusted annual rate

Lower bound of the interval: what would need to go worse than the central path.

The model assigns 80 percent to the interval; the site assigns no other probability.

Falsifiers and assumptions
  • If the first published value of us-housing-starts (USA) for 2027-01 is below 1,109 thousands of units, seasonally adjusted annual rate or above 1,350 thousands of units, seasonally adjusted annual rate, the 80 percent interval is falsified.
  • The model assigns a 10 percent probability to a value below 1,109 thousands of units, seasonally adjusted annual rate and a 10 percent probability to a value above 1,350 thousands of units, seasonally adjusted annual rate; interval misses should occur about one time in five over many forecasts, and a run of misses well above that rate falsifies the calibration.
  • Assumption: No structural break in us-housing-starts between the data cutoff and the target period.
  • Assumption: The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
  • Assumption: The most recent value is the best available guess: recent changes carry no persistent information.
Record 3ab4a587 · last_value (baseline is champion) · data cutoff 2026-09-07 22:16 UTC · created 2026-09-07 22:16 UTC · USA

Section 8 of 9

What to watch

Each indicator is a stored series with a numeric threshold; the reading says which side of it the latest value sits on.

AI narrative Modeled

US building permits at 1433 thousands show year-over-year percent 2.3571 against threshold 0, status easing [obs:164614]. US housing starts show -13.4777 percent, status worsening [obs:169883]. Federal funds rate at 3.63 percent below threshold 4, status easing [obs:169072]. 10-year yield at 4.77 percent above threshold 4, status worsening [obs:200327].

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

Observed
Year over year: +2.4% (2026-07-01)Threshold: 0%

On the easing side of the threshold · rising confirms easing

Permits turning up year over year lead starts.

obs 164614, 164602 · United States
Observed
Year over year: -13.5% (2026-07-01)Threshold: 0%

On the worsening side of the threshold · rising confirms easing

Starts rising year over year confirm the recovery in the metric itself.

obs 169883, 169871 · United States
Observed
Level: 3.63 percent (2026-08-01)Threshold: 4 percent

On the easing side of the threshold · falling confirms easing

A policy rate below 4 percent has historically coincided with stronger building.

obs 169072 · United States
Observed
Level: 4.77 percent (2026-09-03)Threshold: 4 percent

On the worsening side of the threshold · falling confirms easing

The 10-year yield below 4 percent brings mortgage rates toward the range that clears more buyers.

obs 200327 · United States

Section 9 of 9

Forecast track record

Backtests are rolling-origin, out of sample; live evaluations compare published forecasts with the realised values.

AI narrative Modeled

For us-housing-starts USA 6 month horizon the champion is last_value with backtest MAE 93.78571428571433 over 42 cutoffs and 80 percent interval coverage 81 percent. For 12 month the MAE is 93.69444444444454 over 36 cutoffs with 94 percent coverage. No live evaluations yet.

Written by grok-4.3 over the numbers in this snapshot only, validated against them (every figure and cited id had to be in the snapshot), run 2b0739ab, prompt 1.0.0. It adds no facts.

ForecastChampionBacktest MAESkill vs last value80% coverageLive evaluations
us-housing-starts:USA:6last_value93.79 over 42 cutoffs0%+81%none yet
us-housing-starts:USA:12last_value93.69 over 36 cutoffs0%+94%none yet

The full scorecard is at /forecasts/scorecard.

Provenance

Snapshot
5 · 2026-09-07 22:35 UTC · navigator_view_v1+navigator_topics_v1
Gaps stated
0
AI narrative
grok-4.3, run 2b0739ab-cf9b-4beb-acb0-1925b748cd2e, prompt 1.0.0, 17 ids cited
Earlier snapshots
none

Every value on this page is a stored observation (obs ids shown), a forecast record, a mechanism or a relationship row, or a percentile computed from those the same way the relevance engine computes its features. Series the config references but the site does not store are listed as gaps under their section, never filled in.