Permits, starts, completions, spending: what each measures

Question

What can US Census Bureau new residential construction and construction spending data establish for a planner, and what can they not establish?

Who can use this

• Public-sector and regional planners reading monthly housing construction releases alongside local permit data.
• Analysts and researchers choosing which series matches a question about authorization, activity, output or dollars.
• Architecture and construction professionals locating official sources, including where adopted building codes are tracked.

Chronology

1. April 2001 Census implemented the current methodology for compiling starts, under construction and completions estimates, with revisions back to January 1999.
2. January 2005 The monthly Survey of Construction sample design in current use took effect, with about 900 permit offices.
3. January 2023 The Building Permits Survey moved to an annually updated universe, at approximately 19,900 permit-issuing places as of that month.
4. January 2025 Permit imputation moved from region-level factors to a mix of division and large individual state factors.
5. 2026-10-05 FEMA's BCAT page shows this last-updated date, with statistics based on FY2026 Q4 data.
6. 2026-10-20 Census and HUD released September 2026 new residential construction estimates.
7. 2026-11-02 Census released September 2026 construction spending estimates.
8. 2026-11-18 Next new residential construction release, as scheduled by the source.
9. 2026-12-01 Next construction spending release, as scheduled by the source.

Claims

Observed

C-01 In September 2026 privately owned units authorized by permits ran at a seasonally adjusted annual rate of 1,400,000; starts were 1,221,000 and completions 1,192,000. Census Bureau and HUD
Source text: “rate of 1,400,000. This is 0.2 percent below the revised August rate of 1,403,000 and is 3.0 percent below the”

C-02 Census defines a start as the point at which excavation begins for footings or the foundation, not the date a permit is issued. Census Bureau
Source text: “of construction occurs when excavation begins for the footings or foundation of a building.”

C-03 A house counts as completed when finished flooring is installed, or at occupancy if occupied earlier; this is a statistical definition. Census Bureau
Source text: “A house is defined as completed when all finished flooring has been installed”

C-04 The authorized but not started series is an end-of-period stock, 279,000 units seasonally adjusted at September 2026, and it excludes cancelled, abandoned, expired and revoked permits. Census Bureau and HUD
Source text: “Cancelled, abandoned, expired, and revoked permits are excluded.”

C-05 Census reports that permit-issuing places contain more than 99 percent of privately owned residential buildings constructed. Census Bureau
Source text: “Over 99 percent of all privately-owned residential buildings constructed are in permit-issuing places.”

C-06 Starts and completions from the Survey of Construction are tabulated only for the United States and four Census Regions. Census Bureau
Source text: “The SOC does not have a large enough sample size to make state or local area estimates.”

C-07 Permits are the only new residential construction series available below the regional level. Census Bureau
Source text: “The only series on new residential construction that is available at a smaller geographic area is the housing units authorized by building permits.”

C-08 Permit, start and completion statistics cover privately owned buildings; public housing units are outside them. Census Bureau
Source text: “These statistics only include privately-owned buildings. Publicly-owned housing units are excluded from the statistics.”

C-09 The statistics also exclude HUD-code manufactured homes, moved buildings and units created inside existing structures, so conversions do not appear in the counts. Census Bureau, HUD, Census Bureau and HUD
Source text: “manufactured (mobile) homes, moved or relocated buildings, and housing units created in an existing residential or nonresidential structure.”

C-10 Census states that a seasonally adjusted annual rate describes the month for which it is calculated and is not a prediction. Census Bureau
Source text: “The seasonally adjusted annual rate is neither a forecast nor a projection”

C-11 Census states that establishing an underlying trend may take three months for permits and six months each for total starts and total completions. Census Bureau and HUD
Source text: “It may take three months to establish an underlying trend”

C-12 A monthly change is statistically significant only when its 90 percent confidence range excludes zero; the September 2026 starts change of 5.2 percent (±11.8 percent) includes zero. Census Bureau and HUD
Source text: “If a range does not contain zero, the change is statistically significant.”

C-13 Permit estimates in Table 1 are not sampling estimates, but they carry nonsampling error and include imputed activity, with a total quantity response rate of 76.3 percent. Census Bureau and HUD, Census Bureau
Source text: “are based on a non-probability sample and not subject to sampling error”

C-14 Permit valuation is the estimate on the permit document, or an estimate from the permit official when none is listed. Census Bureau
Source text: “The valuation is the estimated value of the residential structure as shown on the building permit.”

C-15 Starts Rates tables relate units started in a month to the units authorized in a given authorization month, at region and division level. Census Bureau
Source text: “Each estimate in these tables is the ratio of the number of housing units authorized by building permits in the Authorization Month”

C-16 Starts estimates are not adjusted for units built in permit areas without a permit. Census Bureau
Source text: “No adjustment is made for units in permit areas built without a permit.”

C-17 Construction spending measures work put in place over a period, a dollar flow rather than a unit count. Census Bureau
Source text: “is a measure of the value of construction installed or erected at the site during a given period”

C-18 The period total sums work on all projects underway, independent of project start dates and contractor payment dates. Census Bureau
Source text: “regardless of when work on each individual project was started or when payment was made to the contractors”

C-19 Maintenance and repairs are excluded from construction spending; land acquisition is also excluded. Census Bureau
Source text: “Maintenance and repairs to existing structures or service facilities.”

C-20 Some spending categories rest on payments rather than work done in the period. Census Bureau
Source text: “For some categories, published estimates represent payments made during a period rather than the value of work actually done during that period.”

C-21 New single-family spending is modeled by distributing estimated cost over time rather than reported month by month. Census Bureau
Source text: “by applying fixed patterns of monthly construction progress.”

C-22 Total construction spending in September 2026 was estimated at a seasonally adjusted annual rate of $2,186.1 billion. Census Bureau
Source text: “$2,186.1 billion, 1.9 percent (±0.7 percent) below the revised August estimate of $2,227.4 billion.”

C-23 Census states that establishing an underlying trend may take 2 months for total spending and as long as 8 months for specific categories. Census Bureau
Source text: “It may take 2 months to establish an underlying trend”

C-24 FEMA reports that 34 percent of natural hazard-prone jurisdictions meet its current-code measure as of Q4 2026, based on FY2026 Q4 data. FEMA
Source text: “Of natural hazard-prone jurisdictions have adopted current hazard resistant building codes.”

C-25 FEMA cautions that its adoption percentage does not measure enforcement or the condition of older structures. FEMA
Source text: “Adoption is just the first step in natural hazard resistance.”

C-26 NIST describes building codes as primarily a state and local matter, typically built from model codes that jurisdictions may tighten or loosen. NIST
Source text: “In the U.S., building codes mostly fall under the purview of state and local governments.”

C-27 Manufactured housing has a separate federal standards regime and a separate statistical survey. HUD, Census Bureau and HUD
Source text: “HUD enforces standards both directly and through state agencies”

Derived

C-28 The September 2026 annual rates differ by 179,000 between permits and starts and by 29,000 between starts and completions, but these gaps do not measure projects lost or delayed. Census Bureau and HUD, Census Bureau
Basis: Arithmetic on the rates in C-01. The three rates describe activity in one month on different units, and the definitions in C-02 and C-03 tie each stage to a different event, so the same units are not being compared across stages.

C-29 Published 2025 averages put new one-unit buildings at 1.4 months from authorization to start and 7.4 months from start to completion; buildings with two or more units took 2.4 and 16.5 months. Census Bureau
Basis: Read from the 2025 rows of the two Average Length of Time workbooks linked from SRC-07. These are historical averages for buildings in permit-issuing places and imply that, on average, a month's completions come from buildings started in earlier months.

C-30 Of the September 2026 changes stated in the release, the starts change from August and from a year earlier is not statistically significant, while the 14.0 percent (±11.1 percent) year-over-year decline in completions is. Census Bureau and HUD
Basis: Applies the confidence-range rule in C-12 to the ranges printed in SRC-01: ±11.8 on 5.2 percent and ±11.5 on 7.4 percent include zero, and ±11.1 on 14.0 percent does not.

Inferred

C-31 Dividing residential spending by unit counts does not yield a cost per home. Census Bureau and HUD, Census Bureau, Census Bureau
Basis: Spending measures work in a period on all projects underway (C-17, C-18), and SRC-08 includes improvements to existing residences in residential spending, while unit counts cover new private units at one stage (C-01, C-08, C-09).

C-32 Permit counts alone support local comparison of authorizations, but no reviewed source supports a local measure of starts, completions or construction spending from these monthly series. Census Bureau, Census Bureau, Census Bureau
Basis: Follows from C-06 and C-07 for residential series. SRC-09 and SRC-10 describe national monthly spending estimates; annual geographic spending tables exist but were not assessed here.

C-33 The Census series and FEMA BCAT cannot be combined to establish which code governed any counted unit. Census Bureau, FEMA, NIST
Basis: The Census series report counts without code attributes in the reviewed definitions, and BCAT reports jurisdiction adoption status with the enforcement caveat in C-25 and the local variation in C-26.