Housing Starts Explained: How Permits, Starts, and Completions Signal New Housing Supply

Housing starts measure when construction actually begins on new privately owned housing units. Under U.S. Census Bureau methodology, a unit is counted as started when excavation begins for the footings or foundation of its building. Starts sit in the middle of the construction pipeline. Building permits come earlier, when a local jurisdiction authorizes construction. Completions come later, when units are finished.

The three measures are related, but they describe different stages and are not interchangeable. A permit is an authorization, not construction. A start means work has begun, not that a home exists. A completion is much closer to physical delivery, but it still does not show whether a unit is occupied, available, or affordable. Together, they describe how new housing moves from approval to delivery. None of them, on its own, shows whether a market has enough housing.

This article explains what each measure captures, how to read the latest national figures, what "seasonally adjusted annual rate" means, and why a single month of construction data is rarely enough to establish a trend.

About the data

National figures in this article come from the U.S. Census Bureau and U.S. Department of Housing and Urban Development's New Residential Construction release for August 2026, published September 17, 2026. Headline permits, starts, and completions are reported as seasonally adjusted annual rates. These are annualized rates, not counts of units for the month. The figures are preliminary and subject to revision. Definitions come from Census Survey of Construction and Building Permits Survey documentation.

What housing starts actually measure

The Census Bureau defines the start of construction as the point when excavation begins for the footings or foundation of a building. Since September 1992, starts have also included units in structures being totally rebuilt on an existing foundation.

For multifamily buildings, all housing units in the building are counted as started when excavation for that building begins.

Example: a 200-unit apartment building enters the statistics as 200 starts when excavation begins. This helps explain why multifamily starts can move sharply from month to month.

The statistics cover new privately owned housing units. Coverage excludes:

• publicly owned housing
• HUD-code manufactured or mobile homes
• group quarters such as dormitories
• transient accommodations such as hotels
• units created inside existing buildings

A housing start is not:

• a completed home
• an occupied home
• proof of current move-in availability

A housing start does not mean the unit is complete or ready for occupancy. Units may be marketed, sold, or leased before completion, so the starts measure itself does not establish current move-in availability.

How building permits, starts, and completions fit together

The three measures form a conceptual pipeline, from authorization to delivery.

Building permit

Authorization

↓

Authorized but not started

↓

Housing start

Construction begins

↓

Under construction

↓

Completion

Construction completed

Stages are related, but monthly totals do not represent the same exact cohort of units.

1. Authorized by permit. A local jurisdiction approves construction by issuing a building or zoning permit.
2. Authorized but not started. Permit issued, excavation not begun.
3. Started. Excavation for footings or foundation begins.
4. Under construction. Started but not complete.
5. Completed. Reached the Census definition of completion.

Monthly permits, starts, and completions do not represent the same exact cohort of units. August permits are not necessarily the units that started in August. August starts are not necessarily the units completed in August.

What the latest national data shows

The Census Bureau and HUD reported the following seasonally adjusted annual rates for August 2026.

What seasonally adjusted annual rate means

SAAR stands for seasonally adjusted annual rate. It has two parts.

Seasonally adjusted

Construction has recurring seasonal patterns. Adjustment removes estimated recurring seasonal effects. The Census Bureau notes this does not account for abnormal weather or year-to-year weather differences.

Annual rate

The Census Bureau multiplies the seasonally adjusted monthly value by 12. The August 2026 figure of 1.275 million housing-start SAAR means the seasonally adjusted August pace expressed at an annual rate. It does not mean 1.275 million homes physically began construction during August.

Monthly activity

↓

Seasonal adjustment

↓

Annualized rate

× 12

Example: 1.275M housing starts SAAR — not 1.275M homes physically started during August.

SAAR is neither a forecast nor a projection. It does not imply the annualized pace will actually persist for a year.

Why building permits are an earlier signal

A building or zoning permit is authorization from a local jurisdiction. Authorization comes before physical construction. A few key points follow from that:

• permit does not equal start
• some authorized units remain not started
• permit does not guarantee future construction
• not all areas require building permits
• permit coverage differs from starts and completions coverage

The principal permit statistics are based on a non-probability sample. The Census Bureau states they are not subject to sampling error, but they remain subject to nonsampling error.

In August 2026, building permits were estimated at 1.394 million SAAR, about 2.7% below July's revised rate and about 3.5% above August 2025.

Why housing starts are not completed housing supply

A housing start confirms construction has physically begun. But it is the beginning of construction, not delivery.

• A start is not a completion.
• A start is not occupancy.
• A start is not current move-in availability.
• A start is not affordability.

Throughout construction, units are not yet complete or ready for occupancy, although some may be marketed, sold, or leased before completion. The starts measure records that construction has begun. It does not show whether a unit has been:

• sold
• pre-leased
• marketed
• completed
• occupied

Why completions are closer to physical housing delivery

A house is considered completed when all finished flooring, or carpeting used in its place, has been installed. If a unit is occupied before construction is otherwise finished, it is classified as completed at occupancy. For privately owned buildings with two or more units, all units are counted as completed when 50% or more are occupied or available for occupancy.

In August 2026, housing completions were estimated at 1.128 million SAAR. Completions are closer to a physical addition to the housing stock than starts or permits are. But completion does not establish:

• occupancy
• listing status
• rent or sale availability
• affordability
• local demand fit
• absorption

Why one month of housing-start data is not a trend

Monthly construction data can move irregularly, so the Census Bureau publishes interpretive guidance for how many months of data it typically takes to see an underlying trend, rather than normal month-to-month movement.

This is guidance, not a forecasting rule or a hard threshold.

Census notes that preliminary seasonally adjusted estimates of total permits, starts, and completions are revised, on average, by 3.8% or less. Because these first estimates are preliminary, later releases can revise earlier values as additional reports are received.

Why statistical significance matters

Housing starts and completions are sample-survey estimates, and the Census Bureau evaluates month-to-month changes using 90% confidence intervals. If the interval includes zero, the change is not statistically significant. If the interval does not include zero, the change is statistically significant.

-14.6%0%+9.4%

Point estimate: -2.6%90% confidence interval: ±12 ppResult: Not statistically significant

The confidence interval spans both negative and positive values, so it includes zero — meaning the survey cannot confirm the change was different from zero.

Point estimates vs. confirmed change

Comparison	Point estimate	90% CI	Significant
Total starts vs. July 2026	-2.6%	±12 pp	No
Total starts vs. August 2025	-1.2%	±10.8 pp	No
Single-family starts vs. July 2026	+7.6%	±14 pp	No
Completions vs. July 2026 (revised)	-11.9%	±9.7 pp	Yes
Completions vs. August 2025	-27.1%	±8.9 pp	Yes

Swipe horizontally to view all columns →

Source: U.S. Census Bureau and HUD, New Residential Construction, August 2026. “No” means the change was not statistically different from zero at the 90% confidence level. It does not mean the value was unchanged.

The August point estimate for total starts was 2.6% below July, but Census reported that the change was not statistically significant.

Completions moved differently. The August completions estimate was down 11.9% from July's revised figure and down 27.1% from August 2025, and the Census Bureau found both changes statistically significant. That contrast is useful methodologically: it shows what a statistically significant change looks like next to one that is not. It does not mean completions have established a new trend on their own — one statistically significant month is still not the same as a trend confirmed across several months.

Single-family and multifamily starts are different signals

All units in a multifamily building are counted as started when excavation begins. Because of that, one large building can add many units to a monthly starts estimate on its own, in a way a single-family start cannot.

Structure type is not the same as tenure. Multifamily construction is not automatically rental housing, and single-family construction is not automatically owner-occupied.

What construction data can—and cannot—tell you about housing supply

Construction data can help describe:

• the construction pipeline
• authorization activity
• starts
• later-stage delivery
• the single-family versus multifamily mix
• broad construction patterns over time

On its own, it cannot independently establish:

• a housing shortage
• excess supply
• affordability
• the direction of rents
• the direction of home prices
• current occupancy
• local demand fit
• future absorption
• future housing demand
• investment attractiveness

Permits

Question

What has been authorized?

Starts

Question

Where has construction begun?

Completions

Question

What has reached the completion stage?

None independently answers: does this market have enough housing?

How to read housing starts alongside other housing indicators

A fuller interpretation can consider construction data alongside:

• rental vacancy
• existing inventory
• household formation
• population growth
• migration
• household income
• affordability
• financing conditions
• employment and broader economic context

These are contextual indicators, not a formula.

The Metro Analytics Insight on what the rental vacancy rate can and cannot tell you about housing supply covers how existing inventory complements construction data. The Insight on what population growth and migration can and cannot tell you about housing demand covers the demand side in more detail.

You can explore Metro Analytics market pages for local context, and the Metro Analytics methodology explains how these measures are defined and sourced.

The takeaway

Building permits, housing starts, and housing completions are different stages of the same housing-production pipeline: authorization, the beginning of construction, and the end of it.

Housing starts are valuable because they show when construction begins. But starts are not completed supply, and one monthly estimate is not an established trend.

Strong interpretation requires:

• understanding which pipeline stage is measured
• recognizing SAAR versus literal monthly counts
• checking statistical significance
• accounting for revisions
• looking across several months
• using broader housing-market context

If you want to look at construction trends alongside other signals for a specific area, you can start with Metro Analytics Markets, or review the Methodology to see how these measures are used.