Same-day permits didn't disappear. The record of them did.
What happens when you try to measure permit approval times across 130 US and Canadian jurisdictions — and why the answer is about software migrations rather than speed.
Every figure here was measured on and is not updated. This report is a claim about that date.
The short version
We tried to measure how permit approval times have changed over the last decade. We hold permit data for 143 jurisdictions; 130 of them publish an issue date on at least some records, and that is the population this report measures. Of those 130, 22 publish enough to attempt the question, and 6 survive the checks needed to believe the answer.
In all six, same-day permit issuance has largely disappeared. Chicago went from 64% same-day in 2015 to 26% in 2025. Denver from 46% to 5%. Calgary from 40% to 1%.
That looks like cities getting slower. In at least one case it demonstrably isn’t — and in the other five we can locate the year it changed, and in three of them name a plausible cause without being able to date it.
The finding is not about speed. It’s that permit velocity cannot be read off this data without knowing when each city changed its permitting software — and nothing in the data records that.
Why only 6 of 130
The obvious way to measure permit turnaround is to subtract the application date from the issue date. Most cities publish both columns. Most of those columns can’t support the subtraction.
Thirty jurisdictions publish an application date on under 1% of their records, and thirty-five fall below any usable threshold. Others publish two date columns that hold the same date on every row: Reading, Pennsylvania publishes 111,174 filings of which 108,776 carry both dates, and the gap between them is zero on 100.00% of them. Houston does the same across 30,516 records. These score highest on any completeness measure — the field is present, populated, and empty of information.
Of the 22 that clear the bar and have enough measurable rows in every year from 2015 to 2025, six survive two further checks.
Corpus growth. If we ingested more of a city over time, a trend in our data might be a trend in our coverage. We check total permit volume per year per city and discard anything where the corpus moved enough to explain the effect.
Composition. Harder and more important: if the share of rows carrying a measurable timeline changes across the window, then early and late years are different subpopulations even when total volume is flat. New York City looked like it had a rising same-day rate. Its three data sources use different date vocabularies and their mix shifts across the decade — the rise is composition, not behaviour. Fort Worth showed a rise on a corpus that grew 8.3×, and its upstream feed turns out to carry no issued-date field at all.
What the six show
| City | Same-day 2015 | Same-day 2025 | Median lag 2015 | Median lag 2025 |
|---|---|---|---|---|
| Calgary, AB | 39.5% | 0.9% | 6 days | 15 days |
| Denver, CO | 46.1% | 5.4% | 2 days | 40 days |
| Winnipeg, MB | 26.2% | 0.6% | 15 days | 21 days |
| Austin, TX | 34.5% | 4.3% | 21 days | 28 days |
| Chicago, IL | 64.2% | 26.3% | 0 days | 5 days |
| Mesa, AZ | 14.9% | 2.8% | 15 days | 21 days |
The first explanation to rule out is that the mix of permit types changed — that cities are issuing proportionally fewer of the simple permits that used to go same-day. It isn’t that. Decomposing within permit type, 86% to 97% of the change happens inside individual types. In two of the six the type mix actually moved toward faster categories, offsetting part of the decline; in the other four it contributed, but never more than 6.2 points against overall moves of 12 to 41. Chicago’s plumbing permits went from 91.1% same-day to 14.2% — inside one permit class, holding everything else constant.
| City | Overall | Within | Between | Within share |
|---|---|---|---|---|
| Calgary | −38.6% | −38.5% | −6.2% | 86% |
| Denver | −40.7% | −41.6% | +3.1% | 93% |
| Winnipeg | −25.6% | −25.3% | −3.7% | 87% |
| Austin | −30.2% | −29.7% | +0.8% | 97% |
| Chicago | −37.9% | −36.5% | −2.3% | 94% |
| Mesa | −12.2% | −12.5% | −1.3% | 90% |
So something real changed. The question is what.
Chicago, where we can name the date
Every one of the six has a locatable inflection year: a point where a flat or slowly-drifting series breaks. Chicago’s is 2024, and it is unusually sharp — the series sits between 57% and 67% every year from 2013 to 2022, then breaks. A step model fits at R² 0.89 against 0.44 for a linear trend, with 71% of the total movement occurring in a single year.
Chicago’s Express Permit Program launched on 6 November 2023, replacing paper “easy permit” and “short form” processes. It expanded on 16 September 2024, replacing the city’s twenty-year-old electrical-only permit process.
Electrical permits were Chicago’s largest permit class at the start of our window — 36.5% of measurable rows in 2015, 29.3% by 2025. Their same-day rate went from 69.3% to 38.7%.
The permit class the city moved online is the class whose same-day rate collapsed, in the year it moved online.
This is not Chicago getting slower. Under the old counter process, an application and an issuance were one transaction, recorded once — so the application date and the issue date were the same date by construction. Under the new system, a submission gets its own timestamp. The work may take exactly as long as it always did. What changed is that the record now shows the gap.
Nothing in Chicago’s permit data says this happened. The field names are unchanged. The values are all valid dates. If you subtract one column from the other across 2023 and 2024, you get a number, and the number is wrong about the world.
The other five
We searched for a comparable event at each inflection. What we found, stated at the confidence we have:
- Winnipeg — Confirmed.
- Permits Online expanded to homeowners in May 2020, and digital development permits took effect 13 October 2020. The inflection is 2020, and the series runs 12.1% → 3.7% → 2.0% across it.
- Calgary — Consistent, undated.
- A Calgary online permit portal was covered as new in February 2019. The inflection is 2018. Close, and not close enough to call.
- Denver — Consistent, undated.
- Denver runs Accela E-permits with electronic plan review that was pushed hard during 2020. The inflection is 2020. Plausible and unproven.
- Mesa — Consistent, undated.
- Mesa issued an Accela procurement in 2014 to replace a four-plan-set paper process. The inflection is 2015.
- Austin — Unconfirmed.
- Austin Build + Connect opened on 30 March 2015 for online payment of invoiced trade permits, and expanded over time. But paying an invoice online cannot create an application date — only application intake can — and 2015 is two years before Austin's 2017 inflection. Austin has a documented digitisation programme and no documented event at its break.
Six cities, six digitisation events at or near the inflection, none refuted. Two dated to the month, three consistent but undated, one where a date exists and doesn’t fit.
We want to be careful about Austin. It is the city this analysis started from and the only one where the search came back without a match. When Austin was the only city checked, “unconfirmed” left room for a genuine slowdown. Now that five of six carry a digitisation event, an unpublished date is more likely evidence about Austin’s website than about Austin’s permitting — but that is an inference, not a measurement, and it constrains what Austin’s label may imply rather than upgrading it. The weakest evidence in an analysis should not become its most favourable interpretation.
What this means if you use permit data
A permit timeline computed across cities is comparing systems, not processes. A city that digitised in 2020 has a different relationship between its two date columns before and after 2020, and nothing in the data marks the boundary.
Any national permit-velocity figure is computed on a base that hasn’t been checked. 22 of 130 jurisdictions can support a trend at all. Six survive the composition check. A number aggregated across “all available cities” is aggregating cities where the measurement means different things.
And the change is invisible in the field itself. No schema version, no note, no flag. The columns keep their names and the values stay valid. The only way to find it is to look for the inflection and then go read the city’s press releases.
Method
143 jurisdictions held; 130 publish an issue date on at least some records, across roughly 27.6 million such rows. A jurisdiction qualifies as measurable if it has a usable application-to-issue lag on at least 50% of those rows and at least 500 measurable rows in every year from 2015 to 2025.
Trend measured as Spearman correlation of annual same-day share against year, alongside median lag. Two controls applied to every candidate: annual corpus size, and the annual share of rows carrying a measurable timeline. A jurisdiction fails if either moves enough to account for the effect. Within-type decomposition run at permit-type grain. Step-versus-trend judged by comparing a best-single-breakpoint step model against ordinary least squares on year, requiring at least two points either side of the break.
Known-bad rows removed. Not: clean rows certified. The issue-date quality flags are populated on roughly 0.3% of the corpus; a null flag means the row was never measured, not that it passed.
Attribution is labelled per city and the labels are not interchangeable: confirmed (a dated event matching the inflection), consistent-undated (an event exists, no date), unconfirmed (searched, nothing matching found).
Every city named was checked against its own upstream source to confirm it publishes two genuinely distinct date fields, because the alternative — a city listed as incomplete when the gap is in our ingestion — is a mistake we have made and corrected before.