US housing starts · United States · 12 months ahead

2027-07: the model assigns an 80 percent probability to 1,073 to 1,348 thousands of units, seasonally adjusted annual rate

Point estimate 1,239 thousands of units, seasonally adjusted annual rate from Last value, against 1,239 thousands of units, seasonally adjusted annual rate observed in 2026-07. No challenger beat the last-value baseline for this horizon out of sample, so the baseline is the champion.

ModeledMetric page

Distribution over the history

Fan chart

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.

Quantiles: p5 1,066 · p10 1,073 · p25 1,147 · p50 1,222 · p75 1,272 · p90 1,348 · p95 1,399
Source: empirical quantiles of 36 in-window 12-step residuals, in log space; no normality assumed

The record

Immutable fields

Key: us-housing-starts:USA:12
Id: d48837b4-6b4e-4441-af29-5f931ed00980
Created: 2026-09-07 22:16 UTC
Data cutoff: 2026-09-07 22:16 UTC
Origin: 2026-07 = 1,239 (obs 169883)
Target period: 2027-07 (365 days)
Point: 1,239 thousands of units, seasonally adjusted annual rate
Interval: 1,073 to 1,348 at 80 percent
Model: last_value models_v1.targets_v1
Baseline: last_value
Snapshot: a88771fb19300fc6f9d5cd622c6262bbb0bd90a612ab6f44ab5faed1d5bf9f69
Data mode: ingested_at
Status: active

Drivers

Statistical association, not causal

Last value uses only the target's own history; there are no driver contributions.

Outcome

Not yet realised

The observation for 2027-07 has not been ingested. The weekly forecast.evaluate job attaches it when it arrives; the record above does not change.

Scenario assumptions

What the forecast takes for granted

No structural break in us-housing-starts between the data cutoff and the target period.
The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
The most recent value is the best available guess: recent changes carry no persistent information.

Falsifiers

What would show this forecast wrong

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.

Competing models

Scorecard for this target and horizon

Model	Status	Cutoffs	MAE	RMSE	Direction	80% coverage	Skill vs last value
Last value models_v1.targets_v1	champion	36	93.69	118.3	0%	94%	0.0%
Drift models_v1.targets_v1	challenger	36	93.67	118.2	58%	92%	0.0%
Holt (exponential smoothing with additive trend) models_v1.targets_v1	challenger	36	99.93	122.6	56%	92%	-6.7%
Linear trend models_v1.targets_v1	challenger	36	154.4	193.7	61%	72%	-64.8%
Ridge regression on lags and drivers models_v1.targets_v1	challenger	36	91.6	113.9	58%	92%	2.2%
Seasonal naive models_v1.targets_v1	challenger	36	93.69	118.3	0%	94%	0.0%

Model card

Last value

Purpose. Point and 80 percent interval forecasts of us-housing-starts (USA) 1, 3, 6, 12 months ahead. Owner: Forecast Lab (HV 3.0).

Method

The forecast for every horizon is the last observed value. This is the baseline every other family must beat out of sample. The series is modelled in natural logarithms and transformed back, so intervals are asymmetric in level terms.

Uncertainty

Intervals are empirical: the model is refitted at earlier origins inside the training window, the realised errors at the same horizon are collected, and the interval is the 10th to 90th percentile of those errors around the point forecast. No normality is assumed.

Training and evaluation

Training window: 811 periods, 1959-01-01 to 2026-07-01. Evaluation: rolling-origin backtest over 36 cutoffs using pseudo real-time data (the earliest ingested value per period; no vintage archive exists for this series, so revisions between the cutoff and today are not reproduced).

Known limitations

Structural breaks after the cutoff are not modelled.
Driver contributions are statistical associations, not causal effects.
Promotion to champion requires beating the last-value baseline on out-of-sample MAE by at least 5 percent over the minimum number of cutoffs; otherwise the baseline stays champion.
Retraining: every scheduled forecast.run refits on the data then available; backtests re-run from /admin/forecasts.

Version chain

Supersedes

This is the first record for its key.