Census Counts the Permitted Housing Backlog Separately

The Census Bureau's 20 October housing report puts September permits at a 1.400 million annual rate and housing starts at a 1.221 million annual rate. Subtracting them gives 179,000 annualized units. That calculation cannot tell an investor how many permitted homes are waiting for construction to begin. The report measures that stock separately: a seasonally adjusted 279,000 housing units at September's end.

The distinction matters when a housing headline becomes part of an options research premise. A difference between two activity rates cannot establish a builder's order book, delivery schedule or future earnings.

Census and the Department of Housing and Urban Development released these September estimates on Tuesday, 20 October 2026, at 8:30 a.m. Eastern Daylight Time, or 10:30 p.m. that day in Brisbane. September is preliminary; the August comparators used below are revised. The latest completed relevant NYSE core equity session at preparation was 20 October, ending at 4 p.m. EDT, or 6 a.m. on 21 October in Brisbane. No share-price, option-price or market-reaction evidence is used here.

Keep the annual-rate label through the subtraction

In Census's release, Table 1a reports units authorized by permits and Table 3a reports units started. Both express the month's activity at a seasonally adjusted annual rate.

The author calculation is 1,400,000 minus 1,221,000, giving a difference of 179,000 annualized units. Dividing that difference by 12 gives about 14,917 units on a seasonally adjusted monthly-rate basis. Neither result is the number of homes waiting at month-end. The monthly conversion changes the time unit; it supplies no opening stock or record of which permits led to starts.

Census's definitions explain that an annual rate is the seasonally adjusted monthly value multiplied by 12. It describes that month's pace and is neither a forecast nor a projection. It also differs from an unadjusted count of activity during September.

A start occurs when excavation begins for a building's footings or foundation. A project receiving a permit in September need not start in September, while a September start may follow an earlier permit. The two headline rates do not follow one common group of projects from authorization through excavation.

Read the end-of-period series directly

Table 2a has the relevant heading: housing units authorized but not started at the end of the period. Its September total is 279,000 units, seasonally adjusted, compared with a revised 278,000 for August. These are stock levels, without the annual-rate label used for permits and starts.

The table's note says the stock includes authorizations from all months through the reporting date that remain unstarted. It excludes cancelled, abandoned, expired and revoked permits. The Survey of Construction definition also limits this series to areas requiring a building or zoning permit. It is not a count of every possible future home in the United States.

Even the stock's small monthly increase needs its uncertainty label. Census reports a 0.4% rise with a 90% confidence interval of plus or minus 2.2 percentage points. That interval includes zero. The estimates do not establish a statistically significant monthly increase at that confidence level. The difference between the rounded stock cells is 1,000 units; it should not be described as an exact administrative tally of newly delayed homes.

Our earlier housing-starts asterisk explanation covers the separate sampling issue in more detail. Here the first job is choosing the series that answers the backlog question.

A project register needs an opening balance

Consider a fictional local construction register with 40 permitted homes awaiting a start at the beginning of a month. During the month, it adds 20 newly permitted homes, removes 15 that start construction and removes five cancelled permits. It finishes with 40 unstarted homes.

The calculation is 40 plus 20 minus 15 minus five, equalling 40. Permits minus starts alone would give five, despite an unchanged closing stock. This invented example assumes all starts come from the permitted register and ignores seasonal adjustment and revisions. It explains why a stock reconciliation needs an opening balance and other removals; it does not reconstruct Census's September estimates.

Nor does the actual 279,000-unit stock promise that those homes will start, finish or sell within a specified period. A national count of permitted units cannot identify the portion belonging to a particular listed builder or the revenue and costs attached to its projects. A company-specific investment premise still needs that company's disclosures and an explanation of how its business relates to the national series.

Keep the housing premise separate from the contract

An options research note can preserve the September reference month, 20 October release, preliminary status and the precise Table 2 stock definition. It should then identify the underlying security, strike, expiration, exercise and settlement terms, quote time, bid and ask, and available size separately. The options-chain guide and liquidity guide explain those contract records.

FINRA's options guide describes the loss and assignment risks. A buyer can lose the option premium. A seller's losses can exceed the premium received, and an uncovered call can lose without limit. Assignment can require buying or delivering shares, including before expiration for American-style equity options. Liquidity and transaction costs affect an exit; concentration, funding, settlement and tax consequences depend on the position and account. This housing release determines none of those terms.

Options Matrix Pro publishes this article and offers commercial options-analysis software. This is general education, not personal financial, legal or tax advice or a recommendation to trade. Before carrying a housing-backlog claim into a contract comparison, record the actual end-of-period stock and its scope rather than substituting the permit-start rate difference.

Sources and scope

• Census Bureau and HUD, Monthly New Residential Construction, September 2026, CB26-162, released 20 October 2026 at 8:30 a.m. EDT. Tables 1a, 2a and 3a supply the rates, stock, revision labels and stock-change interval. This current-release PDF rotates; verify its release date and reference month before reuse.
• Census Bureau, Survey of Construction definitions, accessed 21 October 2026 Australia/Brisbane. Source for authorization, authorized-not-started coverage, starts and annual-rate definitions.
• Census Bureau, Survey of Construction release schedule, accessed 21 October 2026 Australia/Brisbane. Confirms the September report's 20 October date and schedules the October report for 18 November. The release PDF supplies the explicit EDT time zone.
• Census Bureau, New Residential Construction program scope, accessed 21 October 2026 Australia/Brisbane. Covers new, privately owned housing units and excludes HUD-code manufactured homes.
• NYSE, holidays and trading hours, accessed 21 October 2026 Australia/Brisbane. Source for the core equity-session reference boundary, not a universal options close or evidence of a market response.
• FINRA, Options, accessed 21 October 2026 Australia/Brisbane. General contract, assignment, funding and loss-risk context, not a view on housing or a trade recommendation.