The scoreboard
Every figure, and where it came from.
The state assigns every California city a housing target, broken into income categories the state itself defines. Cities report their own progress to the state each year. This page is those reports, added up, with the arithmetic shown.
The score the state never publishes.
Every city and county reports each new home to the state — its income level, and whether it’s for rent or for sale. The state publishes the income totals. It records the tenure of every project — and has never published the rollup. So we did: of the 506,469 homes completed in California from 2021 to 2025, roughly 1,300 were affordable homes a family could own — deed-restricted, very-low or low income, for sale.
Fewer than 1 in 380.
And that is not an artifact of where we drew the income line: count every deed-restricted home for sale at any income level the state tracks — very-low, low, or moderate — and it is still fewer than 1 in 225.
The arithmetic: 1,306 owner-tenure, deed-restricted, very-low- or low-income completions (accessory dwelling units excluded) of 506,469 total — 1 in 388, published as “fewer than 1 in 380” because a project-level audit found the numerator inflated by rental projects misreported as ownership. Owner-tenure deed-restricted moderate-income completions were 922; with the 1,306 at very-low and low income that totals 2,228, which is 1 in 227 — published rounded as “fewer than 1 in 225.”
What these numbers measure
Building permits issued, as a share of the cycle allocation. HCD's own benchmark for RHNA progress. A permit is not a finished home — it is the point at which a city has approved everything it must approve before construction can start. We use it because it is the benchmark the state uses.
6th cycle RHNA. Windows vary by region — most run 2021–2029 or 2023–2031. Because those windows differ, there is no single statewide “share of the cycle elapsed,” and we draw no on-pace line. For what it is worth as an approximation, weighting each jurisdiction by its allocation puts the cycle about 54% elapsed as of July 2026 — which means neither category below is on pace. We do not claim otherwise anywhere.
Entitled → permitted → built
California, statewide — the five most recent years, 2021–2025, with each unit dated by its own event: the approval date for entitlements, the permit issue date for permits, the certificate-of-occupancy (or equivalent) date for completions. Records are deduplicated before summing, because jurisdictions often re-report the same project in successive annual filings until it is finished. This is where the homepage ratio comes from, and it is a different measurement from the percentages above: these are unit counts over a stated calendar window, with no allocation denominator. Because nothing here is divided by an allocation, the staggered-cycle problem below does not touch it.
| Stage | Very low + low | Moderate | Above moderate | Moderate and above, per 1 lower |
|---|---|---|---|---|
| Entitled approved by a city | 142,334 95.2% rentals | 48,824 16,049 deed-restricted | 411,786 | 3.24 : 1 lower is 23.6% of the stage |
| Permitted building permit issued | 132,220 93.4% rentals | 55,485 8,160 deed-restricted | 457,637 | 3.88 : 1 lower is 20.5% of the stage |
| Built certificate of occupancy or equivalent issued | 84,382 93.7% rentals | 40,327 4,692 deed-restricted | 381,760 | 5.00 : 1 lower is 16.7% of the stage |
The homepage carries the built row and nothing else: 422,087 homes completed at moderate income and above, against 84,382 affordable to very-low- and low-income households — a ratio of 5.00 to 1, displayed there as 5. Without deduplication the ratio is 5.2 to 1 — it does not change the picture, and where a ratio does not round cleanly we round against ourselves. The pipeline is least lopsided at approval and most lopsided at completion: 3.24 at entitlement, 3.88 at permit, 5.00 at completion.
Where we drew the line, and what happens if you draw it elsewhere
Moderate income (81–120% AMI) is counted on the market side, because moderate-income units are overwhelmingly not deed-restricted — 35,635 of the 40,327 completed in this window, 88%. Their initial price fits a moderate band; the data reports no restriction keeping it there. But nothing turns on where that line is drawn, so here is the full range at the built stage. Our reading, with moderate on the market side: 5.0 to 1. A stricter reading, moving the 4,692 deed-restricted moderate units to the affordable side: 4.69 to 1. The maximally generous reading, counting all moderate units as affordable — including the 88% with no deed restriction requiring them to stay affordable — 3.1 to 1.
However you draw the moderate line, the picture holds: the ratio never drops below 3 to 1. There is no way to count moderate-income housing that makes the score okay.
The window matters as much as the grouping. We use five years, 2021–2025, because one year swings on a handful of large projects and reporting gaps, and because reaching further back describes a housing market that no longer exists. It is a calendar window, not a cycle — ABAG’s 6th cycle did not begin until 2023, so this is never labelled “the 6th cycle.”
Why we publish the entitled row too
The entitled row is where the pipeline looks least lopsided, and it is published here for that reason. The pipeline is least lopsided at approval and most lopsided at completion — which is also the honest form of the “completions lag permits” objection: homes finished this year were permitted years ago, so the built row is a reading of an older pipeline, not of today’s. We lead with it anyway, because a home someone can live in is the only outcome that has ever housed anyone.
Rent, or own
The state tracks tenure on every project it reports — owner or renter — and never rolls it up. At every stage of the pipeline above, roughly 19 of every 20 affordable homes are rentals: 95.2% at entitlement, 93.4% at permit, 93.7% at completion. The pipeline doesn’t just under-produce affordable homes — the ones a family could buy are a rounding error at every stage: owner-tenure, deed-restricted affordable homes are 1 in 195 of entitlements (3,089 units), 1 in 406 of permits (1,588), and 1 in 388 of completions (1,306). Those are stage counts over the same window, not one set of projects followed through time — but there is no stage at which the share is anything but vanishing.
Of the 84,382 affordable homes completed, 2021–2025
Rentals — 79,042 · 93.7%
For sale — 5,340 · 6.3%
It isn’t only that fewer affordable homes get built
The asymmetry runs deeper than the affordable/market split above. Among homes permitted 2018–2025 — a longer window than the completions figures elsewhere on this page, because it is what the underlying audit computed — 20.0% of renter-occupied units carried a deed-restricted affordability requirement. Among owner-occupied units, 0.65% did — roughly 30 times less often. Deed-restricted affordability is a program built almost entirely around renting.
The genuine article
Even 1,306 overstates it. A project-level audit of the state’s own data found that several of the largest “owner + affordable” entries are rental projects local governments misreported as ownership: Arroyo Crossing II (Indio, LIHTC rental, CTCAC CA-21-656), Villa de Vida (Poway, Mercy Housing rental), Larkin Place (Claremont, supportive rental), and Las Haciendas (Temecula, rental). The bias runs one direction: 1,306 is an upper bound on the genuine count — which is why “fewer than 1 in 380” is conservative. Every correction makes affordable ownership rarer still. The full numerator audit shows how each misreported entry moves the total. The genuine article exists: Stone Pine Cove in Half Moon Bay, 47 for-purchase farmworker homes. It is the exception the audit was built to find.
California, statewide
All 539 reporting jurisdictions
Very low + low income
the state's lower-income categories
9%
91,244 permitted of 1,026,264 allocated — 8.89%, shown rounded above.
Above-moderate income
i.e., market-rate
32%
338,230 permitted of 1,049,216 allocated — 32.24%, shown rounded above.
Building permits issued, as a share of the cycle allocation.
The region that doesn’t fit the pattern
Not every region fits this pattern. San Francisco's 6th cycle (2023–2031, not yet halfway elapsed) reverses it: lower-income permitting is running at 8% (2,703 of 32,881 allocated) against 6% for above-moderate (2,204 of 35,471) — slightly ahead, not behind.
How we checked
The arithmetic, and its limits.
Where the data comes from
The HCD RHNA Progress Report, 6th cycle, published on the California Open Data portal. Reporting year 2025 — the most recent annual progress reports, retrieved from HCD on July 24, 2026. It carries one row per jurisdiction: permits issued and cycle allocation, per income category.
What we did to it
HCD publishes one row per jurisdiction. The statewide totals here are our sums of those rows. We summed the permitted units, summed the allocations, and divided — we did not average the percentages, which would weight a small city the same as Los Angeles. The very-low and low categories are combined into “lower income” exactly as the state defines them (Gov. Code §65584). We do not relabel the state’s categories.
How we verified it
Every row’s own percentage column was recomputed from its unit and allocation columns; none of the 539 rows disagreed. The reporting year was confirmed as 2025 against HCD’s APR Table A2, whose most recent year of record is 2025.
Where the funnel comes from
The entitled / permitted / built table is a separate computation from the percentages above, out of a separate file: the HCD Annual Progress Reports, Table A2, one row per project per reporting year, from a frozen snapshot of the full dataset taken August 1, 2026. We summed Table A2’s own unit columns over every row for the five most recent years, 2021–2025 — the unprefixed columns for entitlements, the BP columns for building permits, the CO columns for certificates of occupancy — dating each record by its own event, not by the year a city filed the report. Table A2’s completion columns record a certificate of occupancy or another reported form of readiness — a final inspection or notice of completion. We use “built” as shorthand for all of them. “Very low + low” sums the acutely-low, extremely-low, very-low and low columns, deed-restricted and not; those are separate reporting lines on the APR form rather than nested subsets, so adding them does not double-count. We checked that rather than assuming it: of the 201 rows reporting a completed extremely-low deed-restricted unit in 2025, 111 report no very-low unit at all, and 35 of the remaining 90 report fewer very-low than extremely-low — which a nested reading could not produce. HCD publishes one row per project. The statewide totals here are our sums of those rows.
Deduplication
Jurisdictions often re-report the same project in successive annual filings until it is completed — a known issue with APR data also documented by the Terner Center and ABAG — so naive sums overcount. Before summing, we deduplicate records by jurisdiction, tracking ID, parcel, event date, and unit type. Because the key includes the event date and unit type, a project that legitimately reappears at a new stage — entitled one year, permitted the next — is kept. Only records reporting the same units at the same stage and event are removed. Deduplication moves the built ratio from 5.2 to 1 to 5.0 to 1; it does not change the picture, and we publish both. All figures refer to the August 1, 2026 data snapshot — the live dataset updates weekly.
Download the frozen archive — the August 1, 2026 snapshot, the analysis scripts, and their verified outputs, including the numerator audit behind the tenure-misreporting findings above. sha256 of the uncompressed snapshot: 07fc0ff79d83f076e45b50957c9d9b97b694d2a2f8e0a822ef2af0c3fa0243ca.
Why we lead with completions
HCD’s official RHNA benchmark is building permits, and the percentages above use it. We lead the homepage with completions anyway, because a permit is not a home and we are arguing about homes. Both stages are published here so the choice is visible rather than quiet — permits tell a less stark version of the same story.
Staggered cycles
Regional cycle windows do not line up. SCAG’s 6th cycle runs 2021–2029; ABAG’s runs 2023–2031; others differ again. Any percent-of-target figure — including every percentage on this page — therefore mixes jurisdictions that are different distances into their own clocks, and there is no single statewide “share elapsed” that would honestly correct for it. The funnel table avoids the problem entirely by counting units over a stated calendar window instead of measuring against a cycle.
What it can’t tell you
The underlying reports are self-reported by cities and counties, and HCD states plainly that it does not independently verify them. Permits are not completions. And completion reporting is the weaker half of the file: certificates of occupancy are reported less consistently across jurisdictions than building permits are, so the built row is the row most likely to be understated. We use it because it measures the thing that matters, not because it is the cleanest column in the release.
If you find an error here, we want to know. Every figure on this page comes from the release linked above, and we will correct and date-stamp anything that turns out to be wrong.