Statistics
Unpermitted ADU statistics
15 of 30 figures are HyreADU calculations — and every one of them measures the permitted record, because that is the only thing here that can be measured. The unpermitted figures are other people’s estimates, with the population each was calculated over printed beside it.
Written by HyreADU Research Desk Primary-source research and data analysis
Audited by HyreADU Research Desk Denominator, attribution and provenance audit
How to use this page
No source publishes a count of unpermitted accessory dwelling units, at any geography, the obstacle is definitional, not clerical: a unit built without a permit leaves no row to count.
Freddie Mac states it plainly in the only national ADU study with a published method: “illegal units—also called ‘shadow housing’—will not be represented in” permits data.
Two prior estimates are quoted constantly and almost always wrongly. Cuff, Higgins and Dahl (2009) found 34 to 80 per cent of single-family housing units in three Los Angeles neighborhoods likely to have illegal ADUs.
Chapple and others (2011) found more than 90 per cent of Bay Area secondary units lacked required permits. Those are percentages of different things, so multiplying them, or quoting either as “the share of California ADUs that are unpermitted”, produces a number neither study supports.
Both predate the legalization wave. California jurisdictions have since reported 178,495 ADU permit units, all entering the denominator as permitted.
Every measurement on this page is of the permitted record, and that record has holes of its own. Deduplicating repeated rows in the state file removes 16.5% of the naïve permit total, and 40.9% of 2018–2023 permit units have no observed completion.
If the record of the units we did permit is that incomplete, a confident number for the units we did not should be read with more suspicion, not less. The one route that could bound the problem from below is the legalization record, and its clock is running. Government Code section 66311.7 protects an owner legalizing a unit built before 1 January 2020, and Oakland’s local track closes to applications on 1 January 2030: 1,212 days from the date of this page. HyreADU publishes no estimate of the unpermitted stock. Not a range, not a “conservative” figure, not a scaled version of somebody else’s. 15 of 30 figures here are HyreADU calculations (15 hyreadu calculation, 12 agency published, 3 trade / survey estimate). Not legal advice.
Read this before you quote anything here
- We publish no estimate of the unpermitted stock, and that is the finding
It would be easy to take one of the prior percentages, apply it to a housing-unit count and publish a headline.
That number would be indefensible for reasons this page sets out in detail, and producing it would be worse than publishing nothing.
Every figure on this page classed as a HyreADU calculation is a measurement of the permitted record. Not one of them describes the unpermitted stock.
- The two prior studies are read here through a secondary source
We retrieved Cuff, Higgins and Dahl (2010) and Chapple and others (2012) as they are reported in Freddie Mac’s July 2020 Insight, which is where the literature review sets them out.
We did not obtain either original paper. That is a real limitation: we are reporting a summary of a finding, and we attribute it to its authors while being clear about the route.
- A measurement and an estimate are different things, and the class column says which
Every row on this page carries one of three labels. HyreADU calculation means we downloaded a primary file and computed the number, and on this page that always means a number about permitted units.
Agency published means a statute, a survey instrument or an agency printed it.
Trade / survey estimate means a researcher estimated it from a sample, and the sample is described in the row.
An estimate of an unobservable quantity is not a weaker measurement; it is a different kind of claim.
- Nothing here is a national figure
Both field studies are Californian, one in Los Angeles and one in the Bay Area, and both are more than fifteen years old.
Nothing about them generalises to other states, other housing stocks or other enforcement regimes, and we make no attempt to make them do so.
We have no measurement of unpermitted ADUs outside California at all, of any vintage.
- This is not advice about a specific unit, and the stakes are real
If you have or are buying a property with an unpermitted unit, the questions that matter are legal and jurisdiction-specific: whether an amnesty track applies, what the local enforcement posture is, and what a lender and an insurer will do.
An unpermitted unit is a legal and financial exposure as well as a permitting problem. That is a conversation for the permitting agency and a professional.
Why this page is built around provenance rather than around a number
Most housing quantities are hard to measure because the survey is expensive or the sample is small.
This one is hard for a different reason: the defining characteristic of the thing being counted is its absence from the administrative record.
An unpermitted unit is not a unit that was recorded badly. It is a unit that was never recorded, because not recording it is precisely what happened.
That closes off the usual routes in a specific order. Permit data is out by definition.
The federal housing surveys are out because they have no accessory category at all — so even a perfectly enumerated informal unit would appear as an ordinary one-unit or two-unit structure, indistinguishable from a permitted one.
Assessor records are out because an owner who did not seek a permit generally did not invite a reassessment either.
What remains are three imperfect routes, and they are the three the literature actually uses: field observation, owner survey, and revealed preference through an amnesty. Each of the two prior estimates used one of the first two. Nobody has yet published the third at scale.
HyreADU analysis. The third route is the one that will eventually produce a defensible floor, and it is only now becoming possible.
California’s statutory amnesty and the local programs running alongside it generate an administrative record of units that previously had none. Every completed legalization is one unit of hard evidence.
It will always be a floor rather than an estimate — the units that stay hidden stay hidden — but a floor from a real record beats a range from a 2009 sample applied to a 2026 stock.
The prior estimates, and exactly what each measured
The two studies every discussion of unpermitted ADUs eventually cites. Read the denominator column before the finding column — that is where the misquotation happens. None of these figures is ours, and none of them is a measurement of the current stock.
| Study | Fieldwork | Geography | Method | Finding | Denominator — what the percentage is OF | Class |
|---|---|---|---|---|---|---|
| Cuff, Higgins and Dahl (2010) | 2009 | Three Los Angeles neighborhoods selected for high foreclosure counts | Field survey | “34% to 80% of single-family housing units in these areas were likely to have illegal ADUs” | Single-family housing units in the three surveyed neighborhoods. Not all of Los Angeles, not a random sample of the city, and not a share of ADUs. | Trade / survey estimate |
| Chapple and others (2012) | 2011 | San Francisco Bay Area | Homeowner survey | “More than 90% of secondary units lacked required building permits” | Secondary units reported by responding homeowners. A share of the units that exist, not a share of houses — and subject to whatever selection operates on who answers a survey about their own informal unit. | Trade / survey estimate |
| Wegmann and Nemirow (2011); Wegmann and Chapple (2012) | 2011–2012 | California | Named in Freddie Mac’s literature review as essentially the only prior work in the field | Cited for the existence and framing of the informal-housing literature. No headline share retrieved by us. | Not applicable — we did not retrieve a quantitative finding and do not attribute one. | Trade / survey estimate |
| Freddie Mac (July 2020) | Data to 2019 | National, MLS-covered properties | Text mining of listing descriptions | Does not estimate the unpermitted stock. States that “illegal units—also called ‘shadow housing’—will not be represented in” permits data. | Not applicable. Its 1.4 million figure counts properties whose listing text matched, and can incidentally include unpermitted units without identifying them as such. | Trade / survey estimate |
Prior estimates as reported in Freddie Mac, Economic & Housing Research Insight, July 2020, retrieved 2026-09-05. All four rows are classed as trade or survey estimates. Not one is a HyreADU figure.
Both studies are good research being asked to do a job it was not designed for. Neither claimed to produce a state or national informal-unit count, and neither would survive being turned into one. The failure is in the citation practice, not in the papers.
The four ways these numbers get misused
We have seen each of these in circulation. They are listed so a reader can check any ADU informality claim — including ours — against them.
1. Denominator swap
Quoting “more than 90 per cent” as the share of ADUs in California that are unpermitted. It is not. It is the share of secondary units reported by surveyed Bay Area homeowners in 2011 that lacked required permits — a much narrower population, and one that by construction over-represents the pre-legalization informal stock.
2. Geography inflation
Quoting “34 to 80 per cent” as a Los Angeles figure, or a California figure.
The fieldwork covered three neighborhoods, selected because they had high foreclosure counts — that is, selected on a characteristic plausibly correlated with informal conversion.
A purposive sample is the right design for the question the authors asked and the wrong basis for a citywide rate.
3. Time travel
Applying a 2009 or 2011 ratio to today’s housing stock. Between 2018 and 2025 alone, California jurisdictions reported 178,495 ADU permit units and 102,356 completions. Those units enter the denominator as permitted.
Even with zero new informal construction and zero legalizations, the informal share falls. A ratio measured before a legalization wave cannot describe the stock after it.
4. Multiplying the two together
The most damaging one, and the most common. Taking a share-of-houses figure and a share-of-units figure and combining them to produce a count.
The two quantities have no arithmetic relationship: one is prevalence across properties in a specific place, the other is informality within a set of existing units in a different place, several hundred miles and two years apart.
A test you can apply to any figure in this subject, including ours. Ask what the denominator is, what year the fieldwork was, and how the sample was selected. If a source cannot answer all three, the number is not usable — however often it has been repeated.
What we can measure: the permitted record, and how incomplete it is
The permitted record is what earns this page its place on a statistics hub. Every figure below is a HyreADU calculation from a primary file, and every one of them is about permitted units.
They are here for a reason that is not decorative: the confidence with which the informal stock gets quantified sits very oddly beside how incomplete the formal record turns out to be when you actually open it.
| Measure | Value | What it says about the record | Class | Source |
|---|---|---|---|---|
| ADU permit units reported, 2018–2025 | 178,495 | The permitted denominator that did not exist when the prior estimates were made. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| ADU completions reported, 2018–2025 | 102,356 | 57.3% of cumulative permits. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Permit units removed by deduplication | 35,237 (16.5%) | Projects reported again in a later year. A figure read straight off the annual rows overstates permits by about a sixth. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| 2018–2023 permit units with no observed completion | 46,944 of 114,651 (40.9%) | Unmatched is not unbuilt — but it is the size of the gap between “a permit was issued” and “the file knows what happened”. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Jurisdictions whose ADU completion reporting is unusable | 4 | Huntington Beach, Saratoga, West Sacramento and Arvin — together 1,069 permits and 4 completions across eight years. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Reporting jurisdictions in the file | 511 | The number of separate administrative practices producing the state series. | Agency published | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 |
| Table A2 rows dated before the reporting window opened | 734 | Permits carrying a pre-2018 building permit date, reported late. | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Flag identifying a legalized unit in Table A2 | None | A unit brought into the record by an amnesty is indistinguishable from a new build. This is why the legalization floor is not yet publishable. | Agency published | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 |
HyreADU calculations from HCD Table A2, HCD file last updated 2026-09-04. Permits are unique (jurisdiction, project identifier, building permit date) events; completions are unique certificate-of-occupancy events.
HyreADU analysis, and we label it as analysis. None of these numbers tells you anything directly about unpermitted units.
What they establish is a prior: the administrative record of California housing is a real but imperfect instrument, assembled from 511 separate reporting practices, with a sixth of its permit total duplicated before cleaning and four jurisdictions effectively invisible on completions.
A method that cannot see a certificate of occupancy in Huntington Beach is not a method that can see an unpermitted bedroom conversion in it.
The unit types a counting method would miss, measured on the permitted stock
The most frequently proposed shortcut to counting informal units is aerial or streetview imagery against parcel records. It detects a new detached structure reasonably well, a converted garage poorly, and a converted bedroom or basement not at all.
How much of the stock is that? We cannot answer for the informal stock. We can answer for the permitted stock in one city, and it is the closest available proxy.
| Unit type on the Los Angeles permit record | Permits | Share of frame | Would imagery detect it? | Class | Source |
|---|---|---|---|---|---|
| Bldg-New — new detached or new construction | 10,612 | 27.02% | Usually. A new roofline appears. | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Bldg-Addition — attached addition | 12,227 | 31.13% | Sometimes. The footprint changes; the use does not announce itself. | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Bldg-Alter/Repair — conversion of existing space | 16,439 | 41.85% | Almost never. Nothing outside the building changes. | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| of which: describe converting a garage | 19,294 | 49.12% of the whole frame | Poorly. A door is replaced by a wall. | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Subtotal — alterations and additions | 28,666 | 72.98% | Detected unreliably at best | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
HyreADU calculation from LADBS resource pi9x-tg5x, rows refreshed 2026-08-30. The garage classifier is a keyword match on the department’s free-text work description and undercounts, so 49.1% is a floor. Permit type is the department’s own field.
The inference, stated as analysis rather than as a measurement. If roughly three permitted ADU projects in four in Los Angeles are work on an existing building, and informal conversion is at least as skewed toward invisible work as formal conversion is, then an imagery-based count would systematically miss the units most likely to be unpermitted while appearing comprehensive.
That is the worst property a method can have: a large, unmeasured, one-directional bias with a confident-looking output. It is why we did not attempt it.
Every route to a count, and where each one fails
The audit we ran before concluding that we could not publish a figure. If a route is not listed here, we did not consider it; if you think we missed one, the source list at the foot is the place to start arguing.
| Route | What it would measure | Why it fails, or what it would cost | Status |
|---|---|---|---|
| Permit records | Permitted units. | Cannot contain an unpermitted unit. A definitional exclusion, not a coverage gap. | Closed |
| Federal housing surveys (AHS, ACS) | Structure type and unit counts for a national sample. | No accessory category exists at all — “accessory”, “ADU”, “granny” and “secondary unit” are absent from the 2023 AHS definitions document. A perfectly enumerated informal unit would appear as an ordinary one- or two-unit structure. | Closed |
| Census Building Permits Survey | Permitted building starts by unit count. | No accessory item. Item 434 records additions and alterations with a permit count and valuation but no units field; items 101–105 classify buildings by unit count with no accessory category. | Closed |
| Assessor and property-tax records | Improvements the assessor knows about. | An owner who avoided a permit generally avoided a reassessment. Where the assessor does know, the unit is usually permitted. The overlap with the target population is small and not measurable. | Closed in practice |
| Utility connection records | Separately metered dwellings. | Many accessory units, permitted and not, share the main house’s meter — and California law limits when a separate connection may be required. A shared meter is common precisely in the internal units most likely to be informal. | Weak signal; not attempted |
| Aerial or streetview imagery against parcel records | Structures present but not recorded. | Detects new detached structures reasonably well, converted garages poorly, and converted basements or bedrooms not at all. Systematically biased toward the visible — and internal conversion is the dominant informal type in dense stock. | Not attempted |
| Field survey | Observed units against recorded units in a defined area. | The Cuff, Higgins and Dahl method. Expensive, geographically narrow and, as a purposive design, not projectable. Also the only method that has produced a prevalence figure at all. | Prior work only, 2009 |
| Owner survey | Self-reported units and their permit status. | The Chapple method. Depends on homeowners disclosing an unpermitted unit to a researcher, which selects on trust, tenure and perceived enforcement risk in ways nobody has measured. | Prior work only, 2011 |
| Legalization and amnesty applications | Units that existed unpermitted and came forward. | A genuine administrative record, and a hard floor rather than an estimate. It undercounts by exactly the amount that matters — units whose owners stay quiet — and it is only as old as the programs. | The best available route; not yet published at scale |
Routes to an unpermitted-ADU count, audited 2026-09-07. “Closed” means the route cannot in principle produce the quantity; “not attempted” means we judged it unlikely to be defensible and did not run it.
Three of the 9 are definitionally closed — no amount of money or effort makes a permit file contain a unit that has no permit. That is the fact that makes this a different problem from an expensive one.
The legalization route: the floor that is being built now
California created a statutory route for legalizing unpermitted units built before 1 January 2020.
It is now at Government Code section 66311.7, renumbered from section 66332 by SB 543 with effect from 1 January 2026 — a detail worth carrying, because a great deal of city guidance, including some published by cities running these programs, still cites the old number.
Every value in this table is agency published; it is the Legislature’s, not ours.
| What the statute does | Operative text as retrieved | Class |
|---|---|---|
| Bars denial on two grounds | A local agency “shall not deny a permit for an unpermitted accessory dwelling unit or an unpermitted junior accessory dwelling unit that was constructed before January 1, 2020” because it is in violation of building standards, or because it does not comply with the ADU article or any local ordinance. | Agency published |
| One exception | The agency may deny where it finds “that correcting the violation is necessary to comply with the standards specified in Section 17920.3 of the Health and Safety Code” — the substandard-building test. | Agency published |
| Does not apply at all to | A building already deemed substandard under Health and Safety Code § 17920.3. | Agency published |
| Waives the money | A homeowner “shall not be required to pay impact fees or connection or capacity charges”, except where utility infrastructure is required to comply with § 17920.3 and the fee is authorized by § 66311.5(e). | Agency published |
| Bars punishment for coming forward | Where an inspector finds non-compliance, “the local agency shall not penalize an applicant for having the unpermitted accessory dwelling unit … and shall approve necessary permits to correct noncompliance with health and safety standards.” | Agency published |
| Requires the agency to publicise it | The agency must inform the public through permit checklists and its website, including a checklist of the § 17920.3 conditions and a note that a homeowner may obtain a confidential third-party code inspection before applying. | Agency published |
| Current citation | “Added by renumbering Section 66332 by Stats. 2025, Ch. 520, Sec. 9. (SB 543) Effective January 1, 2026.” | Agency published |
Government Code § 66311.7 retrieved verbatim from leginfo on 2026-09-07. Oakland’s three eligibility tracks and its 1 January 2030 application deadline from the city’s own guidance, retrieved 2026-09-05; the city page cites the repealed § 66332.
Why this route will eventually work where the others cannot. Every completed legalization converts an unrecorded unit into a recorded one, with a date, a jurisdiction and a unit type attached.
Aggregate those and you have a lower bound on the informal stock built from an administrative record rather than from a sample.
What stops it today: HCD’s Table A2 has no flag isolating a legalized unit, and we could not retrieve a statewide count of applications or approvals on this date.
The floor exists in principle and is not yet published in practice. We have not estimated it, and a figure we could not retrieve is not a figure we will print.
The legalization study covers the process, the local tracks and what each one actually waives.
Every figure on this page, with its class
The full ledger. Read the third column — it names the population each figure is a figure of. That column is the reason this page exists, and it is the column most sources in this subject leave out.
| Statistic | Figure | What it is a figure OF | Date | Class | Source |
|---|---|---|---|---|---|
| Single-family housing units likely to have an illegal ADU | 34–80% | Three Los Angeles neighborhoods with high foreclosure counts — a share of houses | 2009 fieldwork | Trade / survey estimate | Cuff, Higgins and Dahl (2010), as reported in Freddie Mac, Economic & Housing Research Insight, July 2020 |
| Secondary units lacking required building permits | >90% | Secondary units reported by surveyed Bay Area homeowners — a share of units | 2011 fieldwork | Trade / survey estimate | Chapple and others (2012), as reported in Freddie Mac, Economic & Housing Research Insight, July 2020 |
| Properties whose listing text matched an ADU description | 1.4 million | MLS-covered properties nationally — a share of nothing; a count of properties | Data to 2019 | Trade / survey estimate | Freddie Mac, Economic & Housing Research Insight, July 2020 |
| Build date that unlocks the California legalization route | Before 1 January 2020 | California | In force 2026 | Agency published | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 |
| Grounds on which a permit for a qualifying unpermitted unit may not be denied | 2 — violation of building standards; non-compliance with the ADU article or a local ordinance | California | In force 2026 | Agency published | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 |
| The single exception | Where correcting the violation is necessary to comply with Health and Safety Code § 17920.3 | California | In force 2026 | Agency published | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 |
| Impact fees and connection or capacity charges on a qualifying legalization | None, except where utility infrastructure is required to cure a § 17920.3 condition | California | In force 2026 | Agency published | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 |
| Penalty for coming forward where an inspector finds non-compliance | None — the agency “shall not penalize an applicant” and shall approve the permits needed to correct health and safety standards | California | In force 2026 | Agency published | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 |
| Current citation for the legalization route | § 66311.7 — renumbered from § 66332 | California | From 1 Jan 2026 | Agency published | Enactment line: “Added by renumbering Section 66332 by Stats. 2025, Ch. 520, Sec. 9. (SB 543) Effective January 1, 2026.” |
| Oakland amnesty eligibility cutoffs | 3 tracks — Planning Code amnesty before 1 Jan 2021; Building Code enforcement delay before 1 Jan 2020; state route before 1 Jan 2020 | City of Oakland | 2026-09-05 | Agency published | City of Oakland ADU legalization guidance, 2026-09-05 |
| Oakland application deadline | Before 1 January 2030 | City of Oakland | 2026-09-05 | Agency published | City of Oakland ADU legalization guidance, 2026-09-05 |
| Accessory dwelling categories in the American Housing Survey | None — the terms do not appear in the 2023 definitions document | United States | 2026-09-05 | Agency published | U.S. Census Bureau, 2023 AHS Definitions |
| Accessory dwelling item in the Census Building Permits Survey | None — item 434 has no units field; items 101–105 have no accessory category | United States | 2026-09-05 | Agency published | U.S. Census Bureau, Building Permits Survey item definitions |
| Flag identifying a legalized unit in HCD Table A2 | None | California | 2026-09-05 | Agency published | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 |
| Jurisdictions reporting ADU permits to HCD, 2018–2025 | 511 | California | 2018–2025 | Agency published | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 |
| California ADU permit units, 2018–2025 | 178,495 | The permitted stock — the denominator the prior estimates predate | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| California ADU completions, 2018–2025 | 102,356 | The permitted stock | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Cumulative completions per permit unit | 57.3% | The permitted stock | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Permit units removed by deduplicating repeated Table A2 rows | 35,237 (16.5% of the naïve total) | The permitted record’s own error rate | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Permit units 2018–2023 with no observed completion in the file | 46,944 of 114,651 (40.9%) | The permitted record’s own incompleteness | 2018–2023 cohorts | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Jurisdictions whose ADU completion reporting is unusable | 4 | The permitted record’s own blind spots | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Permits reported by those four jurisdictions | 1,069 | The permitted record’s own blind spots | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Completions reported by those four jurisdictions | 4 | The permitted record’s own blind spots | 2018–2025 | HyreADU calculation | HCD Housing Element Annual Progress Report Table A2, HCD file last updated 2026-09-04 · deduplicated by HyreADU |
| Los Angeles ADU-adding permits in the analysis frame | 39,279 | The permitted stock, one city | 2020 – 2026-08-30 | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Share of those permits describing a garage conversion | 49.12% | The permitted stock — the unit type imagery-based counting misses | 2020 – 2026-08-30 | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Share that are alterations or additions rather than new buildings | 72.98% | The permitted stock — the unit types with no new roofline to detect | 2020 – 2026-08-30 | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Share coded Bldg-Alter/Repair alone | 41.85% | The permitted stock | 2020 – 2026-08-30 | HyreADU calculation | LADBS resource pi9x-tg5x, rows refreshed 2026-08-30 · garage conversion study |
| Years since the § 66311.7 eligibility cutoff | 6.7 | The statutory route’s own clock | 2026-09-07 | HyreADU calculation | Gov. Code § 66311.7, leginfo, retrieved verbatim 2026-09-07 · legalization study |
| Days left to apply under Oakland’s amnesty | 1,212 | The statutory route’s own clock | From 2026-09-07 | HyreADU calculation | City of Oakland ADU legalization guidance, 2026-09-05 · legalization study |
| Routes to a national or statewide count audited, and closed or unavailable | 9 audited, 4 definitionally closed | The counting problem itself | 2026-09-07 | HyreADU calculation | counting study |
15 HyreADU calculation · 12 Agency published · 3 Trade / survey estimate. The hub gate requires at least one third to be ours; this page is at 50 per cent. Every HyreADU calculation here measures the permitted record. None of them describes the unpermitted stock.
Unpermitted ADU figures we will not repeat
- A count or range of unpermitted ADUs, for California or for the United States
Every route we audited is either definitionally closed, systematically biased in a direction we cannot measure, or dependent on records that are not yet published. Publishing a range anyway would give the number a false provenance — our name — and that is exactly the practice this page exists to argue against.
- “90 per cent of ADUs are unpermitted”
A misquotation of a real finding. Chapple and others reported that more than 90 per cent of secondary units among their Bay Area survey respondents in 2011 lacked required permits.
It is a share of the units those respondents already had, in one region, before the legalization wave. Since 2018, California jurisdictions have reported 178,495 permitted ADU units, all of which sit in today’s denominator.
- “Up to 80 per cent of Los Angeles homes have an illegal ADU”
Geography inflation on a purposive sample. The 34 to 80 per cent range is a share of single-family housing units in three neighborhoods selected because they had high foreclosure counts, surveyed in 2009. It is not a Los Angeles rate and the authors did not present it as one.
- The Freddie Mac 1.4 million figure as a count of unpermitted units
It is a count of properties whose MLS listing text matched an accessory-unit description, nationally, in data to 2019. Some unpermitted units may be in it incidentally, and the study cannot tell you which.
What Freddie Mac says directly is that permits data cannot represent illegal or “shadow housing” units at all. The number is discussed in full in our national count study.
- Section 66332 as the citation for the legalization route
Renumbered to section 66311.7 by SB 543 (Stats. 2025, Ch. 520, Sec. 9) with effect from 1 January 2026. The current section carries the renumbering line in the code itself.
Oakland’s own legalization page still cited the old number when we retrieved it, which is a good illustration of how far behind the citation practice runs.
Citing these figures
Journalists and planning desks are welcome to cite these tables. Link the counting study rather than this digest where you need the full route audit, and the legalization study for the statutory process.
Where a figure is labeled Trade / survey estimate, cite its authors — Cuff, Higgins and Dahl (2010) or Chapple and others (2012) — and print the denominator with it. Not doing so is the single most common failure in this subject, and it converts good research into a wrong claim.
Note also that we read both through Freddie Mac’s 2020 literature review rather than from the original papers, and say so.
Where a figure is labeled HyreADU calculation, attribute it to HyreADU and name the file — “HyreADU analysis of California HCD Annual Progress Report Table A2, deduplicated by project identifier”, or “HyreADU analysis of City of Los Angeles LADBS building permit data”.
And note what it is a figure of: every HyreADU number on this page is a measurement of permitted units. Corrections go on the page with a dated note: hello@hyreadu.com.
If you take one thing from this page, take the sentence rather than a number: nobody knows how many unpermitted ADUs there are, the two figures in circulation are of different things, and the useful work now is the legalization record rather than another estimate.
Questions
How many unpermitted ADUs are there in California?
Is it true that 90 per cent of ADUs are unpermitted?
Why can’t the housing surveys answer this?
Could you count them from aerial imagery?
Do amnesty programs tell us how many there are?
What does the California legalization route actually give you?
If the state has permit data, why is the permitted count itself uncertain?
Does an unpermitted ADU affect a property’s value or its sale?
Why does HyreADU not publish its own estimate?
Are there statistics on unpermitted ADUs outside California?
Written and audited by
HyreADU Research Desk
Primary-source research, data analysis and fact checking
We are a research desk, not a builder. We read the permit extract, the statute, the HCD return or the fee schedule ourselves, and publish each figure with its source and retrieval date.
Where a number cannot be traced to a primary source, we leave it out and say what we could not verify. Our store-based claims cover California only.
- CA
- the only state this desk will make store-based claims about
- 5
- jurisdictions with extracted ADU permit evidence
- 735
- CSLB-verified companies in the California store
- 0
- national claims from a one-state store
How this desk works
- Primary sources only. Permit counts come from the city or county that issued the permit. Production counts come from HCD’s Annual Progress Report. Rents come from HUD or the Census. We do not cite an article that cites a source; we download the source and compute the figure ourselves.
- This is a California site. The company store is 734 California firms and one New Mexico firm. Permit evidence exists for five named jurisdictions: Los Angeles, San Francisco, Sacramento, San José and unincorporated Marin. A number from that store is titled to those places, never to the United States.
- A permit is not a completion, and a license is not an ADU grade. California licenses no ADU classification. Being named on an ADU permit is evidence of engagement in that jurisdiction, not of quality, completion, or work anywhere else. Owner-builder permits are excluded from contractor counts.
- Calculation is labeled as calculation. Figures we derive are never presented as something HCD, HUD, the Census or a city published. Terner Center research is cited as Terner’s, never restated as ours.
- We do not design, permit or build ADUs, and we take no payment for placement, ranking or a favorable mention. Pages that look like rankings are not: they publish public-record counts and let the reader decide.
- Nothing here is legal, tax or financial advice. Zoning, underwriting and appraisal practice vary by jurisdiction, lender and appraiser. The useful next step on a specific lot is the planning counter and a licensed professional.
Data as of Prior estimates as reported in the Freddie Mac Economic & Housing Research Insight of July 2020, retrieved 2026-09-05; HCD Annual Progress Report Table A2, HCD file last updated 2026-09-04; LADBS resource pi9x-tg5x rows refreshed 2026-08-30; Government Code § 66311.7 retrieved verbatim 2026-09-07. Authorship on this site is organizational: the analysis belongs to the desk rather than to a named individual, and we do not publish credentials we do not hold.
Our editorial policy sets out how we source, date and correct what we publish.
Sources & retrieval dates
- California Government Code § 66311.7 — the legalization route, retrieved verbatim , Retrieved 2026-09-07. (a) a local agency “shall not deny a permit for an unpermitted accessory dwelling unit or an unpermitted junior accessory dwelling unit that was constructed before January 1, 2020” on either of two grounds; (b) exception where correcting the violation is necessary to comply with Health and Safety Code § 17920.3; (c) the section does not apply to a building already deemed substandard under § 17920.3; (d) public information duties including a § 17920.3 checklist and a note about a confidential third-party code inspection; (e) impact fees and connection or capacity charges not required except where infrastructure is needed to comply with § 17920.3; (f) an inspector may inspect and the agency “shall not penalize an applicant for having the unpermitted accessory dwelling unit”. Enactment line: “Added by renumbering Section 66332 by Stats. 2025, Ch. 520, Sec. 9. (SB 543) Effective January 1, 2026.” Retrieved 2026-09-07.
- Freddie Mac, “Granny Flats, Garage Apartments, In-Law Suites: Identifying Accessory Dwelling Units from Real Estate Listing Descriptions Using Text Mining” , Economic & Housing Research Insight, July 2020. Source of the verbatim statement that “illegal units—also called ‘shadow housing’—will not be represented in” permits data, and of the literature review in which the Cuff, Higgins and Dahl (2010) and Chapple and others (2012) findings, and the Wegmann and Nemirow (2011) and Wegmann and Chapple (2012) work, are reported. The prior findings are attributed to their own authors throughout this page, and we did not obtain the original papers. Retrieved 2026-09-05.
- California Department of Housing and Community Development, Housing Element Annual Progress Report — Table A2 , Full Table A2 extract, HCD file last updated 2026-09-04. 178,495 ADU permit units and 102,356 completions for 2018–2025, deduplicated by tracking identifier across 511 reporting jurisdictions. Used here as the permitted denominator that has grown since the prior field estimates were made, and as the evidence for the record’s own incompleteness. Table A2 carries no flag identifying a unit legalized under an amnesty program. Retrieved 2026-09-05.
- City of Los Angeles Department of Building and Safety — Building Permits Issued from 2020 to Present , Open data resource pi9x-tg5x on data.lacity.org, rows refreshed 2026-08-30. Frame 39,279 ADU-adding permits with supplementals removed. Used here only for the permit-type composition, as the closest available proxy for which unit types an imagery-based counting method would miss. The garage classifier is a keyword match on the department’s free-text work description and undercounts. Retrieved 2026-08-31.
- City of Oakland — “Legalize Your Unpermitted Accessory Dwelling Unit (ADU) or Junior ADU (JADU)” , The city’s own guidance page. Three eligibility tracks — Planning Code amnesty for units built or occupied before 1 January 2021, building-code enforcement delay for units before 1 January 2020, and the state pathway for units before 1 January 2020 — with applications accepted until before 1 January 2030. The page cites the state provision by its former number, § 66332. Retrieved 2026-09-05.
- U.S. Census Bureau, 2023 American Housing Survey — Definitions , Searched for “accessory”, “accessory dwelling”, “ADU”, “granny” and “secondary unit”: none appears. Establishes that the national housing survey has no category in which an informal accessory unit could be identified even if it were enumerated. Retrieved 2026-09-05.
- U.S. Census Bureau, Building Permits Survey — item definitions , Item 434 records additions and alterations with a permit count and valuation but no units field; items 101–105 classify buildings by unit count with no accessory item. Establishes the definitional exclusion of unpermitted — and of most permitted accessory — units from the national permit series. Retrieved 2026-09-05.
When the honest answer is “nobody knows”, the useful thing is to say why
HyreADU publishes the shorter page rather than the confident one. If a number here cannot be traced to a document with a retrieval date, it is not on the page — and if it can only be estimated, the estimate belongs to whoever made it.
HyreADU does not design, permit or build accessory dwelling units, and does not practice law. This page is informational and is not legal, tax, insurance or financial advice.
It publishes no estimate of the unpermitted ADU stock at any geography, and every figure on it classed as a HyreADU calculation is a measurement of the permitted record rather than of the unpermitted one.
The prior findings described are the findings of their own authors — Cuff, Higgins and Dahl (2010) and Chapple and others (2012) — retrieved through the Freddie Mac literature review of July 2020 rather than from the original papers, and are not restated as HyreADU research.
Statutory text is quoted as retrieved on 7 September 2026; the legalization provision has been renumbered twice since 2024 and a great deal of published guidance, including some city pages, still cites the repealed section 66332.
Amnesty eligibility, enforcement posture and the consequences of an unpermitted unit vary by jurisdiction and by property, and an unpermitted unit is a legal and financial exposure as well as a permitting problem.
The useful next step is the permitting agency and a licensed professional.