Research study
Why the national permit data cannot see ADUs
Not “has not counted them yet.” The Census Building Permits Survey is structurally incapable of producing an ADU number, in two separate ways, and this page shows you the instrument.
Written by HyreADU Research Desk Primary-source research and data analysis
Audited by HyreADU Research Desk Survey-methodology and retrieval-date audit
The finding
There is no national ADU permit count because the survey that would produce one does not collect the category. The Census Bureau’s Building Permits Survey classifies residential construction into four structure types, quoted in full below, and accessory is not among them.
No item on the survey form asks whether a unit is accessory to another.
The signal is lost twice, in opposite directions. A garage, basement or other internal conversion is an alteration, captured under Item 434 (“additions and alterations to housekeeping residential buildings”), which carries no housing-unit count; that unit never enters the national units series at all.
A detached new-build ADU does enter the units series, but as a one-unit building, in the same cell as a four-bedroom spec house. So the two errors do not cancel. They point opposite ways and neither is measured. Conversions are a pure undercount.
Detached new-builds are an unattributed overcount inside a category that is used as a proxy for single-family housing production.
HyreADU calculation: in California, where an independent state ADU permit series does exist, HCD-reported single-family-detached plus ADU permits were 9,145 units above the federal one-unit series for California in 2025 and 925 units below it in 2018 — a swing of roughly ten thousand units in a series where ADU permits rose from 11,978 to 31,460 over the same period.
The federal one-unit line barely moves across those eight years. A national ADU figure would require a change to the instrument, not a cleverer analysis of the output: an accessory flag on the permit record itself, plus a unit count on alteration permits that create a dwelling.
This page uses federal and California state data only. HyreADU’s company and permit store is a California store and is used nowhere in this analysis.
Read this first
- No national ADU number is produced here, and that is deliberate
The point of the study is that the instrument cannot yield one.
If you came here for a figure to cite, the companion page — how many ADUs are there in America — sets out the closest defensible estimate that exists and its error bars. It is not derived from permit data, because it cannot be.
- The California comparison is a demonstration, not a validation
Setting HCD’s Annual Progress Report against the Building Permits Survey compares two instruments with different universes, different reporting duties, different timing rules and different definitions of a permit event.
Neither is a gold standard for the other. What the comparison shows is the size of the divergence and its direction; it does not show that one series is right.
- We did not obtain microdata, and we did not survey permit offices
Everything here rests on the published BPS methodology, the published item definitions, the published classification FAQ, the published state history file, and California’s published APR extract.
We did not run a records request against Census, and we did not ask the 19,900 permit-issuing places how they code accessory units in practice. Practice almost certainly varies, and that variation is itself unmeasured.
- One state is not the nation
California is the only state in this analysis with an independent ADU permit series, which is exactly why it is the only state we could run the cross-check in.
That is a limitation of the evidence and not a finding about anywhere else. Nothing on this page describes ADU production in a state we did not measure.
The question this page actually answers
Every few months a national ADU statistic goes into circulation. It is usually round, usually undated, and usually traceable — if it is traceable at all — to a chain of citations that ends at somebody restating somebody else.
The natural instinct of anyone who has worked with housing data is to go and check the permit records, because permits are the one place American housing production is systematically counted.
That instinct fails here, and it fails for a specific and checkable reason. The federal permit series was designed in an era when the interesting distinction was how many units were in a building.
It was not designed to record whether one of those units was subordinate to another on the same lot.
When accessory units became a policy category, the survey did not acquire a category for them, and nothing in the published series can be reprocessed to recover one.
HyreADU analysis: this matters more than a missing row usually would, because the Building Permits Survey is not just one dataset among many.
It is the input to a great deal of downstream analysis — housing starts, state and metro production comparisons, model inputs at lenders and agencies.
When a fast-growing form of housing is invisible to it in one direction and unlabelled in the other, the error propagates quietly into everything built on top.
What the Building Permits Survey is, in its own words
Four definitions, each quoted from the Census Bureau’s published methodology or item definitions, retrieved 2026-09-05. Read them together and the mechanism becomes obvious.
- The universe
- Approximately 19,900 permit-issuing places, as of January 2023. The Bureau adds that “[f]or the nation as a whole, less than 1 percent of all privately owned housing units are constructed in areas not requiring building permits. However, this proportion varies greatly from state to state and among metropolitan areas.” The coverage problem is therefore small nationally and potentially large locally.
- The structure categories
- Verbatim: “(1) single family houses (attached and detached combined), (2) two-unit buildings, (3) three- and four-unit buildings, and (4) residential buildings with five or more units.” Items 101 to 105 classify buildings by number of units. There is no accessory item and no accessory flag anywhere on the form.
- Item 434 — additions and alterations
- Verbatim: “Includes all permits issued for additions and alterations to housekeeping residential buildings, and conversions of nonresidential and nonhousekeeping buildings to residential buildings. Does not include special ‘installation’ permits issued to cover electrical, plumbing, heating, air-conditioning, or similar mechanical work.” Item 434 records a permit count and a valuation. It does not record housing units.
- Imputation
- Where two of the three items — units, buildings, valuation — are present, the third is imputed from annually updated ratios. Total non-response is imputed at division or state level. Imputation reconstructs the missing item within the existing categories; it cannot create a category the form never collected.
HyreADU note: the classification FAQ is worth reading for what it excludes as well as what it includes.
It directs respondents to exclude “other structures on residential property, such as sheds, garages, pools, etc., when permitted separately.” It offers no guidance at all on accessory dwellings — which is consistent with there being no place to put them.
The two places the signal is lost
The two losses are not two descriptions of the same problem. They are two different failures, affecting two different kinds of ADU, in two opposite directions.
Loss one — conversions vanish entirely
A garage conversion, a basement unit, an internal ADU carved out of an existing house: each of these is an alteration to an existing housekeeping residential building. It is captured by Item 434.
Item 434 collects a permit count and a construction valuation. It has no units field. So a project that creates a new legally occupiable dwelling contributes a permit and a dollar figure to the national record and zero units to the national units series.
This is a pure undercount and it falls hardest on exactly the ADU type that is cheapest and most common in dense, older housing stock. In our own Los Angeles extract, garage and internal conversions are a large share of ADU activity — see the garage conversion study.
Loss two — detached ADUs are counted, but not labeled
A new detached backyard ADU is a new one-unit building. It enters the BPS units series correctly as one unit.
But category (1) merges attached and detached single-family houses into a single cell. A 500 sq ft backyard cottage and a 4,000 sq ft house on a new subdivision lot are the same row in the same column.
This is not an undercount of units. It is an unattributed count: the one-unit series silently becomes a mixture of two very different products, and any analysis that treats “1-unit permits” as a proxy for conventional single-family homebuilding inherits the contamination without being told.
HyreADU analysis: the fact that the two errors point in opposite directions is why no correction factor works. If both were undercounts you could bound the number.
Here, a jurisdiction that builds mostly conversions and a jurisdiction that builds mostly detached units produce national records that are wrong in different ways, and the published series gives you no way to tell them apart.
The California cross-check, 2018–2025
California is the only state with an independent, jurisdiction-reported ADU permit series — the Annual Progress Report each city and county files with HCD.
Setting it against the federal one-unit series for the same state and year shows how large the unlabelled component has become.
The final two columns are HyreADU calculations; the first four are as published.
| Year | HCD ADU permits | HCD single-family detached | Census BPS, CA 1-unit | HCD SFD + ADU (HYRE calc) | Difference vs BPS (HYRE calc) |
|---|---|---|---|---|---|
| 2018 | 11,978 | 45,928 | 58,831 | 57,906 | -925 |
| 2019 | 12,865 | 43,452 | 58,575 | 56,317 | -2,258 |
| 2020 | 12,947 | 40,862 | 59,043 | 53,809 | -5,234 |
| 2021 | 20,843 | 50,983 | 65,890 | 71,826 | +5,936 |
| 2022 | 25,826 | 39,914 | 63,628 | 65,740 | +2,112 |
| 2023 | 28,906 | 39,480 | 58,534 | 68,386 | +9,852 |
| 2024 | 31,006 | 34,979 | 61,143 | 65,985 | +4,842 |
| 2025 | 31,460 | 35,424 | 57,739 | 66,884 | +9,145 |
HCD permit events are deduplicated by tracking identifier, so a unit permitted once and reported in two annual returns is counted once. BPS figures are units, not buildings, for California one-unit structures.
HyreADU note on what this does and does not show. It does not show that the BPS undercounts California by the difference.
The two series have different universes and different timing conventions, and part of the divergence is definitional rather than substantive.
What it does show is that a state series which can see ADUs and a federal series which cannot have moved apart by roughly ten thousand units over eight years, in a period when the ADU component tripled. That is the shape of an instrument problem, not noise.
Every candidate national source, and what it actually measures
If you want a national ADU number, these are the instruments you would reach for. Each is listed with what it genuinely measures and the precise reason it does not answer the question.
| Source | What it measures | Why it cannot give a national ADU count | Retrieved |
|---|---|---|---|
| Census Building Permits Survey | Permits issued for new residential construction at about 19,900 permit-issuing places, by four structure categories, plus alterations under Item 434. | No accessory category exists. Conversions carry no unit count; detached ADUs are merged into “single family houses (attached and detached combined).” | 2026-09-05 |
| American Housing Survey | Detailed characteristics of a national sample of housing units, including structure type, tenure and condition. | The strings “accessory”, “accessory dwelling”, “ADU”, “granny” and “secondary unit” do not appear anywhere in the 2023 AHS definitions document. A detached ADU is “a single unit … detached if it has open space on all four sides” — the same as the main house. An internal ADU makes the host a multi-unit structure — the same as a duplex. | 2026-09-05 |
| American Community Survey, table B25024 | Units in structure, for all occupied and vacant housing units, at tract and place level. | B25024 inherits the AHS structure definition exactly. There is no accessory dimension to condition on, at any geography. | 2026-09-05 |
| Freddie Mac ADU text-mining study (July 2020) | Distinct single-family properties whose MLS listing description text matched ADU key-phrases, across a licensed CoreLogic dataset of about 600 million MLS transactions since the late 1990s. | It is a count of properties that were listed and whose listing text mentioned an accessory unit — not a stock count. Freddie Mac itself names both an understatement bias and an overstatement bias. Discussed in full on our national count page. | 2026-09-05 |
| State APR-style reporting | Permits and completions by unit category, self-reported by each local jurisdiction to a state agency. | It works — California’s ADU series is exactly this. But it exists in one state. There is no federal requirement that any other state collect it, and we located no comparable statewide ADU permit series anywhere else. | 2026-09-05 |
| City-level ADU reporting | Local ADU permit counts published by individual jurisdictions, sometimes under a local reporting ordinance. | Genuinely useful and genuinely local. Seattle’s annual ADU report, mandated by Ordinance 125854, is the strongest example we found — see our growth outside California study. It aggregates to nothing national because coverage is idiosyncratic. | 2026-09-05 |
The candidate instruments for a national ADU count, with the specific failure mode for each. Every source in this table was retrieved on 2026-09-05 and is listed in full at the foot of this page.
What a real national ADU number would require
What follows is a HyreADU recommendation, clearly labeled as one. We are a research desk, not a statistical agency, and none of what follows is a description of anything Census has proposed.
- 1 An accessory flag on the permit record itself
The cheapest possible change and the one that does the most work: a yes/no field on the survey form asking whether the permitted unit is accessory to an existing or proposed primary dwelling on the same lot.
Jurisdictions that already have an ADU permit type — which is most large ones — could populate it from their own system. It would immediately split category (1) into the two products that are currently merged there.
- 2 A units field on alteration permits that create a dwelling
Item 434 would need to distinguish an alteration that creates a dwelling unit from one that does not, and count the units created. Without this, every conversion ADU stays invisible no matter how good the accessory flag is, because conversions never reach the units series in the first place.
- 3 A structure question in the AHS and the ACS that separates accessory from primary
Permits measure flow. A stock estimate needs the housing surveys to be able to identify an accessory unit at all, and at present they cannot: the concept is absent from the 2023 AHS definitions document entirely. Until a structure or relationship question exists, the stock is not measurable from the surveys, only inferred around them.
- 4 Published guidance for permit offices, because coding practice varies
The classification FAQ currently says nothing about accessory dwellings. A flag with no instructions produces a flag that different offices apply differently. Any serious attempt at this would need the guidance published alongside the field, and the variation measured rather than assumed away.
HyreADU note: we are not asserting that these changes are practical, cheap or a priority for the Bureau. Survey changes are expensive and contested, and there are good reasons agencies resist adding fields.
What we are asserting is narrower and, we think, unarguable: without at least the first two, no processing of the published series can produce a national ADU count.
Why this page is the foundation of our own scope rule
HyreADU verifies one state. Our permit evidence covers five California jurisdictions and our company store is 735 CSLB-verified firms, almost all Californian.
It would be easy, and dishonest, to scale those numbers up to a national figure with a population weight and call it research.
This page is why we do not: the national denominator that such an extrapolation would need does not exist, and anyone who produces one has either built it from somewhere they should name, or has not built it at all.
When you read a round national ADU number, the useful question is not whether it is plausible. It is which of the instruments above it could possibly have come from.
Questions
How many ADUs are permitted in the United States each year?
Could you not just subtract to get the ADU number?
Does the American Housing Survey ask about ADUs?
Is California’s number reliable, then?
Why does the federal one-unit line for California barely move while ADU permits triple?
What about housing starts — do ADUs show up there?
Does the 1 per cent of housing built without permits matter here?
Has the Census Bureau said it will add an ADU category?
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 Census Building Permits Survey methodology, item definitions and classification FAQ retrieved 2026-09-05; BPS Annual History by State, annual 2025 release of 14 May 2026; California HCD Annual Progress Report Table A2 extract, HCD file last updated 4 September 2026. 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
- U.S. Census Bureau, Building Permits Survey — Methodology , Source of the four structure categories quoted verbatim on this page, the universe of approximately 19,900 permit-issuing places as of January 2023, the statement that for the nation as a whole less than 1 per cent of privately owned housing units are constructed in areas not requiring building permits, and the imputation procedure for missing items. Retrieved 2026-09-05.
- U.S. Census Bureau, Building Permits Survey — item definitions (items.txt) , Source of the Item 434 definition quoted verbatim, and of items 101 to 105, which classify buildings by number of units. No accessory or secondary-unit item appears anywhere in the file. Retrieved 2026-09-05.
- U.S. Census Bureau, Building Permits Survey — classification FAQ , Contains no accessory-dwelling guidance. Directs respondents to exclude other structures on residential property such as sheds, garages and pools when permitted separately. Retrieved 2026-09-05.
- U.S. Census Bureau, Building Permits Survey — Annual History by State and Territory , Annual 2025 release, 14 May 2026. California column, units rather than buildings, for total privately owned units and one-unit structures. Supplies the “Census BPS, CA 1-unit” and total columns used in the cross-check. Retrieved 2026-09-05.
- U.S. Census Bureau, 2023 American Housing Survey — Definitions , Source of the housing unit and units-in-structure definitions quoted on this page. Searched for “accessory”, “accessory dwelling”, “ADU”, “granny” and “secondary unit”: none appears. The structure definition establishes that a detached ADU is its own single-unit detached structure and an internal ADU makes the host a multi-unit structure. 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 4 September 2026. 922,102 rows, of which 281,321 carry UNIT_CAT = ADU. Permit and completion events deduplicated by tracking identifier: 178,495 ADU permit units and 102,356 completions for 2018–2025 across 511 reporting jurisdictions. Supplies the HCD ADU and single-family-detached columns in the cross-check. Retrieved 2026-09-05.
- 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. Identified 1.4 million distinct single-family properties with accessory dwellings by text-mining a licensed CoreLogic dataset of about 600 million MLS transactions since the late 1990s. Cited here for its explicit statement that permit data cannot represent “illegal units — also called ‘shadow housing’”, and as the only national estimate with a published method. Retrieved 2026-09-05.
- City of Seattle Office of Planning and Community Development, “Accessory Dwelling Units 2024 Annual Report” , October 2025, the fourth annual report, produced under a reporting requirement in Ordinance 125854. Cited here only as the example of a local ADU permit series that works — 913 ADUs permitted in 2024, against 283 in 2019 — and as evidence that the data can be collected where somebody is required to collect it. Retrieved 2026-09-05.
Every figure on this site carries the instrument it came from
Naming the instrument is the whole method. If a number cannot be traced to a document with a retrieval date, we publish the shorter page and say what we could not verify. The research index is where the rest of that work lives.
HyreADU does not design, permit or build accessory dwelling units. This page is informational and is not legal, tax or financial advice.
It describes published federal survey documentation and published California state data as retrieved on 5 September 2026, and it deliberately does not publish a national ADU count.
The four-part change described under “what would fix it” is a HyreADU recommendation and is not a description of anything the Census Bureau has proposed.
The California comparison sets two different instruments beside one another to show the size of a divergence; it does not establish that either series is correct.