Housing Starts

Housing Starts is the starts leg of New Residential Construction, the monthly Census Bureau survey program run with HUD that tracks privately owned homes. A start means excavation began for footings or foundations, not permit approval, and public housing is excluded. Fieldwork comes from the Survey of Construction: 900 permit places plus 80 non-permit areas, covering the nation and 4 regions only. It does not measure permits, existing-home sales, or prices, so each series answers a different question.

Collection runs monthly: sampled permits are followed until the start month, while non-permit areas are canvassed at least once every 3 months. Release falls on the 12th workday after the month at 8:30 a.m. Eastern with permits, completions, units under construction, and authorized-but-not-started units. The prior 2 months are revised monthly, seasonal factors for 63 months with April data, and February data finalizes the prior year. A move counts only if its 90% confidence interval excludes zero, and trends need 3 months for permits and 6 each for starts and completions.

Starts split by building size and by place.

• Single-family homes: detached and qualifying attached houses with separate systems, the largest share of starts.
• 2-to-4-unit buildings: small multifamily structures reported as their own column.
• 5-or-more-unit buildings: apartment or condo-type structures where every large permit is sampled.
• 4 Census regions: the nation plus Northeast, Midwest, South, and West, the only published geography.

January 2024 data, released February 16, 2024, put starts at a 1,331,000 SAAR, the seasonally adjusted monthly value × 12. Down 14.8% (±10.2%) from 1,562,000 in December, a real drop because the interval excluded zero, while permits stood at 1,470,000 and completions at 1,416,000. The year-on-year move was 0.7% below January 2023's 1,340,000 (±11.7%), not significant. Single-family starts at 1,004,000 fell 4.7% (±11.6%), not significant, with 314,000 in 5-or-more-unit buildings.

1. Permit control total: start from the monthly Building Permits Survey permits issued, the universe total for permit areas.
2. Start ratios: match sampled permits to their start months and build 14 ratios from 5 years of data. Cover the issuance month, each of the prior 11 months, months 13 to 18 pooled, and months 19 to 60 pooled.
3. Apply ratios: multiply sampled starts for each month group by its ratio, separately in 8 sets: single-family and 2-or-more-unit structures across 4 regions. For example, 90,000 sampled starts × a 1.20 ratio = 108,000 ratio-adjusted starts.
4. Timing adjustment: add about 4% for units started before permits and late reports, using history; no adjustment covers units built with no permit. For example, 108,000 × 1.04 = 112,320 permit-area starts.
5. Non-permit add-on: weight up starts found by driving non-permit roads at least once every 3 months and add them to permit-area starts. For example, add 2,000 weighted non-permit starts for 114,320 actual starts.
6. Seasonal adjustment: remove seasonal, trading-day, and holiday effects with X-13ARIMA-SEATS software. Sum 4 regional single-family series plus the US 2-or-more-unit total, then multiply by 12: 114,320 / 0.95 × 12 ≈ 1,444,000.