Most US cities publish permit timeline data. Almost none publish it on most of their permits.
What 143 US jurisdictions actually publish about how long a building permit takes — and why a number computed across them is measured on a subset nobody has described.
Every figure here was measured on and is not updated. This report is a claim about that date. For current numbers, see the live status map.
Correction, 26 August 2026: our corpus count was understated. One jurisdiction publishes a filing identifier that is present on every record and identical across all of them, so our de-duplication collapsed its 674,787 records into a single filing. Corrected, the corpus is 23,398,188 rather than 22,718,868 — and the two shares below move against us: the share publishing both dates falls from 76.5% to 74.3%, and the share supporting a measured timeline from 14.8% to 14.4%. No finding in this report changes. Every jurisdiction count is unaffected, and the corrected figures make the gap slightly wider, not narrower.
Correction, 26 August 2026: this report said Norfolk publishes an application date and no issue date at all. That was wrong, and the error was ours, not the city’s. Norfolk publishes an issue date on 106,553 of its 125,845 records — 84.7% — and our pipeline did not ingest it. The same check found a second: Howard County publishes a file date on 100% of its records and we did not ingest that either. Both are verified against the cities’ own endpoints. Norfolk has been removed from the excluded list below, and the partition is re-derived: 100 jurisdictions publish both dates rather than 98, 17,558,046 filings and 75.0% rather than 17,381,346 and 74.3%, and 43 rather than 45 publish no usable pair — the issue-date-only group falls to 5,622,345 filings from 5,682,188, and the application-date-only group to 224,100 from 334,640. The count that can be measured is unchanged — we hold both dates for 98 as of this correction. See the coverage note below — that has since changed. We checked all six excluded cities against their live endpoints again; the other five reproduce exactly.
Coverage note, 26 August 2026: we now hold both dates for 100 jurisdictions rather than 98, and 94 of those 100 publish informative lag data rather than 92 of 98. Nothing about the cities changed. We began ingesting an issue date that Norfolk publishes on 86.5% of its records, and a file date that Howard County publishes on 100% of its records — fields both cities were serving and we were not reading. The gap was ours, we found it while checking this report, and this line exists so the improvement is not silent.
The short version
If you subtract a permit’s application date from its issue date across American cities, you will get a number for nearly all of them. That number is real. It is also computed on whichever permits happened to carry both dates — and in most jurisdictions that is a minority of the corpus, with no published account of how that minority differs from the rest.
- 94 of the 100 we hold both dates for publish informative lag data.
- Only 14 publish it on a majority of their records.
- 2 publish two date columns holding the same date on every record.
This is not a story about cities publishing nothing. The data is better than it is usually assumed to be. It is simply not usable in the way it is usually used.
Two tests, two answers
We applied two independent tests to every jurisdiction that publishes both an application date and an issue date.
Does the lag data carry enough distinct values to describe a distribution? 94 of 100 pass. These jurisdictions publish hundreds of distinct application-to-issue intervals — real, varied, informative timeline data.
Is it published on most of the records? 14 of 100 pass, on a threshold of 99% dual-date coverage.
The two answers disagree, and the disagreement is the finding. The information exists almost everywhere. The coverage does not exist almost anywhere.
For the 84 jurisdictions that pass the first test and fail the second, a duration is computable and its denominator is unstated. The permits missing a date are not missing at random — an application still in process is exactly the kind that carries no issue date yet — so the measured subset skews toward permits that completed, and a median computed on it runs fast by an unknown amount.
The scale
We hold 23,398,188 permit filings across 143 jurisdictions.
- 100 publish both an application date and an issue date — 17,558,046 filings, 75.0% of everything we hold.
- 30 publish only an issue date. You can see that a permit was issued; you can never see when it was asked for. 5,622,345 filings.
- 13 publish only an application date — the mirror case. You see the request, never the outcome. 224,100 filings.
43 jurisdictions publish no usable pair of dates at all. Of the 100 that publish both, only 14 publish them on essentially every record: 3,363,641 filings, 14.4% of the corpus.
The extreme case: two columns, one date
Two jurisdictions publish an application date and an issue date that are the same date on 100.00% of records, with 0 distinct non-zero intervals between them.
Houston — 30,516 filings, every one same-day. The source publishes a single permit date, written into both fields.
Reading, Pennsylvania — 111,174 filings, 108,776 dual-dated, every one same-day. The issue date is set equal to the application date whenever the record’s status implies the permit issued.
Neither is a fast permitting system. Both are one column subtracted from itself. These are the jurisdictions that score highest on a naive completeness check — both fields present, both fully populated, and empty of information.
Four more jurisdictions publish intervals that vary but take too few distinct values to describe a distribution: 99,074 filings between them.
The 14 where the question is answerable
| Jurisdiction | Filings | Dual-date coverage | Median days | 75th pct | Same-day share |
|---|---|---|---|---|---|
| Austin, TX | 1,625,074 | 99.96% | 0 | 9 | 58.66% |
| Chicago, IL | 867,084 | 99.95% | 0 | 11 | 55.22% |
| College Station, TX | 180,158 | 99.99% | 0 | 1 | 71.28% |
| Winnipeg, MB | 161,670 | 100.00% | 18 | 44 | 15.81% |
| Baton Rouge, LA | 142,587 | 99.99% | 1 | 14 | 47.21% |
| San Antonio, TX | 121,624 | 100.00% | 0 | 2 | 67.93% |
| Denver, CO | 116,717 | 99.88% | 15 | 69 | 29.04% |
| Gainesville, FL | 96,691 | 99.96% | 0 | 6 | 56.09% |
| Milwaukee, WI | 16,385 | 99.83% | 25 | 65 | 21.71% |
| Cambridge, MA | 14,594 | 100.00% | 20 | 43 | 2.84% |
| Aurora, CO | 11,273 | 100.00% | 1 | 9 | 42.73% |
| Allentown, PA | 5,538 | 100.00% | 7 | 21 | 17.66% |
| Santa Monica, CA | 2,810 | 99.47% | 43 | 206 | 15.46% |
| Alpharetta, GA | 1,321 | 100.00% | 36 | 61 | 0.08% |
The spread is the point. Median 0 days in five of them; 43 in Santa Monica. Same-day share runs from 0.08% to 71.28%. This is not a ranking and it should not be read as one — see the limits below.
What we cannot tell you
We do not know what these cities mean by “applied.” We verified that each writes two distinct source fields into two distinct columns. We did not decode whether a given city’s application date marks submission, acceptance after intake review, or formal filing — that is a per-municipality policy question and no permit dataset documents it. Across our adapters, many different source fields feed each of the two columns, and they are not synonyms: one city’s application date is the moment a file was opened, another’s is the moment intake review accepted it. Any comparison here is a comparison of published fields, not of equivalent process milestones.
There is no national figure in this report and there cannot be. 14 jurisdictions, self-selected by which happen to publish completely, is a sample of American permit publishing — not of American permitting.
We are not naming most excluded cities. Naming one requires verifying against its own live endpoint that the gap is the city’s and not ours. We ran that check on the five largest and it held for all five, but it is one verification per city and we have not done the rest.
| Jurisdiction | Filings | What the source serves |
|---|---|---|
| Philadelphia | 927,838 | Carto permits table, 45 fields, five date fields — none an application date |
| Boston | 661,135 | the city's catalogue reports 660,840 rows in its only permits dataset, 25 fields, no application date |
| Las Vegas | 436,465 | 38 fields, one date field |
| Baltimore | 289,884 | 18 fields, two date fields, neither an application date |
| Dallas | 126,840 | 11 fields, one date field |
In every one of the five, the absence is the source’s, not ours. We ingest 925,473 of Philadelphia’s 927,838 issue dates — 99.75% — and all 25 of Boston’s fields. We checked, because the alternative was telling a city’s own reporters that it publishes nothing when the gap was on our side.
Why this matters
Permit timelines are a live question in American housing policy. How long approval takes, whether it is getting worse, which cities are outliers — people are actively trying to answer these with published data.
The answer for most of the country is not that the data is missing. It is that the data is partial in a way that is invisible at the point of use. A researcher pulling both date columns gets a well-formed number from nearly every city. For 86 of 100, that number describes an unstated fraction of that city’s permits, selected by a mechanism nobody has published.
We think that is worth knowing before the number gets used.
Method
143 jurisdictions, 23,398,188 permit filings, deduplicated to filing grain. All-time window. Records are excluded from any duration when the issue date is flagged as a placeholder, future-dated, a parse artefact, or a repeated sentinel value.
A jurisdiction publishes informative lag data if its non-zero intervals take at least 100 distinct values and its same-day share is below 80%. It publishes on most records if at least 99% of its filings carry both dates with the issue date on or after the application date.
Both conditions are required for the 14, and neither alone is sufficient: a jurisdiction can show many distinct intervals on a small fraction of its records, and one can show near-complete coverage of two columns that hold the same value.
Classification precision, per-jurisdiction coverage and the measurement date for every figure are published at permitcore.io/methodology/certification. No jurisdiction in this report is certified across all cohort families, and the coverage figures above describe date completeness, not classification accuracy — they are different measures and this report does not blend them.