California ADU statistics — jurisdiction by jurisdiction

Where the permits actually come from

The twenty largest reporting jurisdictions, by cumulative ADU permit units 2018–2025. Read the share column. Los Angeles alone is 27.76% of the state; the five largest are 39.89%; the twenty here are 53.69% of everything California reported. The remaining 46.3% is spread across the other 491 reporting jurisdictions.

HyreADU calculation from HCD Table A2, HCD file last updated 2026-09-04. Permits and completions are deduplicated unique events. Jurisdictions excluded as unusable on completion reporting are not in this table and are named in the reporting-quality section below.

HyreADU analysis. A concentration this high is the single most important thing to know before quoting any California ADU average.

A “statewide average permit time”, a “statewide average unit size” or a “statewide completion rate” is dominated by Los Angeles, whose zoning, housing stock, department practice and construction market are not those of the other 510 reporting jurisdictions.

Where this desk publishes a statewide figure, it is because the underlying series is genuinely statewide — not because we averaged jurisdictions.

Ten jurisdictions, eight years each

The same file cut by jurisdiction and year. Each panel is scaled to its own maximum, so the shapes are comparable and the heights are not. The point of the small multiples is that the trajectories are not the same shape: Los Angeles fell for three years and then recovered, San Diego grew sharply from a small base, and San Francisco and Oakland have been broadly flat on very small numbers.

HyreADU calculation from HCD Table A2. Permit units assigned to the calendar year of the building permit date, deduplicated by jurisdiction and project identifier. 2025 data were due on 1 April 2026 and this extract was taken on 2026-09-05, after HCD’s stated 30 June completeness date.

Los Angeles is the trajectory worth staring at. 7,039 permit units in 2018, falling to 3,514 in 2020, then recovering to 7,439 in 2025.

A statewide series that rose smoothly across the same window was therefore rising despite its largest contributor falling for two years.

That is exactly the kind of composition effect a statewide chart hides and a jurisdiction cut exposes.

Completions per permit unit, and why the spread is not a league table

Cumulative completions divided by cumulative permit units, 2018–2025, for the ten focus jurisdictions against the statewide 57.3%. This is not a measure of how many permitted ADUs got built, and reading it that way is the most common error made with this file.

HyreADU calculation from HCD Table A2. Excluding the four unusable reporters moves the statewide ratio by 0.34 percentage points — small at state level, and decisive for the four jurisdictions themselves.

Three reasons a jurisdiction’s ratio can be low, and this file separates none of them. The permits may be recent and the projects still under construction — a jurisdiction growing fast will always look worse on a cumulative ratio.

The completions may have happened and not been reported. Or the projects may genuinely be stalling.

San Diego reported 0 ADU completions in 2018, 2019 and 2020 combined against 1,318 permits, which looks much more like a reporting practice than like a construction outcome.

The permits against completions study follows permit cohorts forward, which is the right way to ask this question.

How good is California’s own ADU record?

Record quality is the section most ADU statistics pages do not have, and it is the one that decides whether the rest of them mean anything. Every figure here is a HyreADU calculation about the file itself rather than about housing.

HyreADU calculations from HCD Table A2 except the final row, which is a property of the published schema.

Exclusion rules: U1 — 40 or more ADU permits across three or more years with zero ADU completions; U2 — 100 or more permits across five or more years, completions in at most one year, cumulative ratio under 5 per cent.

Why we publish this at all. A statistics page that reports only the outputs of a file, and never its condition, is asking the reader to trust an instrument they cannot see.

The condition here is: usable, genuinely valuable, and imperfect in ways that are measurable and worth measuring. The annual report study is where the exclusion rules were set and argued.

Which vintage you are reading

California’s ADU numbers change after the fact. Jurisdictions file annually, late filings backfill earlier years, and a permit issued in 2019 can appear in the file for the first time in 2023.

A figure from Table A2 without a retrieval date is not a figure, and these are the specific facts that make that true.

HyreADU calculations from HCD Table A2 except the first two rows. This is the reason Terner’s contemporaneous 2018 figure and ours differ: later filings backfill earlier permit dates, so an extract taken today reports more 2018 permits than an extract taken in 2020 could.

The practical instruction. If you are comparing a California ADU figure across two sources, check the extraction date before you check anything else. Two honest analysts reading the same table a year apart will get different numbers for 2018, and neither is wrong.

The approval clock, where a jurisdiction publishes enough to measure it

The state file records the date a permit was issued. It does not record the date the application was filed, so the approval clock cannot be measured from it at all.

Three California jurisdictions publish both dates on a complete permit extract, and those three are the whole of what can be measured. Three is not a state.

HyreADU calculation from each jurisdiction’s own open permit portal, retrieved 2026-09-05. Los Angeles: earliest non-supplemental permit per address. San Francisco: unique permit numbers on the department’s own ADU flag.

Marin: unincorporated county only, matched on a work-description regular expression. The three definitions are not identical and the medians should be read as three separate measurements rather than as a ranked list.

The spread is the finding, not the median. City of Los Angeles at 169 days against City and County of San Francisco at 445 — and in City and County of San Francisco, 62.2% of ADU permits took more than a year from filing to issue.

The statute allows 60 days from a completed application, and the gap between those two numbers is mostly the completeness question rather than the decision question.

The permit times study is where that distinction is worked through, and where the jurisdictions we could not measure are logged.

Every figure on this page, with its class

The full ledger, so the 50-of-59 split can be checked rather than taken on trust. Nothing appears anywhere on this page that is not in this table.

50 HyreADU calculation · 9 Agency published · 0 Trade / survey estimate. The hub gate requires at least one third to be ours; this page is at 85 per cent.