Research · 2 min read
Does Construction Really Attract Theft? Seattle Data Complicates the Story
Permit phase alone is an unstable proxy for construction-site theft exposure; site conditions and project moments deserve the attention.
Everyone knows construction sites attract theft. We matched six years of Seattle larceny records to 3,997 permitted projects to check — and the answer refused to sit still. Depending on how you draw the circle and weight the projects, active construction looks slightly worse, slightly better, or exactly the same. That wobble is the story.
Our analysis of 3,997 permit projects; 172,545 accepted larceny-theft records in the matching pool, permits completed july 1, 2020–june 30, 2025; offense records through june 30, 2026.
Is permit-active status a reliable signal of increased nearby larceny?
Across 3,997 eligible Seattle permit projects completed from July 2020 through June 2025 and nearby larceny records through June 2026, no stable direction appears: the 250-meter exposure-weighted active rate is 3.1% below the pre-period rate, the 100-meter ratio is effectively flat, and the unweighted project mean is slightly higher during active permits.
The numbers behind the answer
Selected measures only. Denominators and interpretation stay attached so the headline cannot stand alone.
−3.1%
Active vs pre rate at 250 m
The exposure-weighted active-period rate was 259.1 versus 267.4 matches per 1,000 project-days before issuance.
0.965–1.006
Active/pre ratio by radius
The direction ranges from effectively flat at 100 m to modestly lower at 250 m and 500 m.
3,997
Eligible permit projects
Projects had usable issue and completion dates and contributed exposure time to all three periods.
A permit-active interval is not the same as a theft at a construction site
First, an honesty check about what's being measured. A police record near a project is not a theft from that site. The data can't say whether the event happened on the project, involved materials, or had anything to do with construction at all. It's a larceny record inside a circle drawn around a permit address.
And the permit itself is a proxy. Work can start weeks after issuance, pause, surge, or wind down before completion. So the real question isn't "do permits raise theft?" It's whether the pattern survives every defensible way of asking. It doesn't.
Answer
It is not stable enough to support a simple ‘construction increases nearby theft’ rule.
Reasonable summaries point in different directions
Draw a 250-meter circle and weight projects by exposure time: the active-period rate is 259.1 matches per 1,000 project-days against 267.4 before issuance — 3.1% lower while active, and 4.6% lower than the post period. Shrink the circle to 100 meters and the active/pre ratio flips to 1.006, a hair higher. Widen to 500 meters: 0.965, lower again.
Or skip exposure-weighting and average each project's own rate equally. Now the active period looks slightly higher: 276.0 versus 267.4 before. Same city, same records, same projects — three reasonable methods, three directions. That disagreement is the finding.
Evidence visual
Nearby record matches per 1,000 project-days at 250 meters
Exposure-weighted aggregate rates. These are project-relative record matches, not unique thefts from construction sites.
One police record can be matched to many nearby projects
One more number keeps everyone honest. The analysis produced 1,025,178 project-offense matches from 108,337 unique offense records. Each record was matched, on average, to nearly ten projects — buffers overlap, and comparison periods repeat.
That's valid for measuring project exposure. But if you skimmed the match count as "incidents near construction," you'd overstate the problem almost tenfold. Both numbers have to travel together.
Inspect project moments and site exposure instead of presuming risk
So what do you do with a non-answer? Stop asking the permit to do a job it can't. What shapes theft exposure at a project is physical: when materials arrive and where they're stored, which sections of the perimeter are open this week, who's on site after hours, how the layout changed since last month.
The aggregate neither convicts construction nor acquits it. It shows that "there's a permit, therefore there's risk" is not a defensible shortcut — and neither is "the data says it's fine." The defensible version is a walk through the actual site at the actual phase. The checklist is the starting set.
- Confirm when physical activity actually begins and ends.
- Review high-value delivery and storage periods.
- Track perimeter and access changes by phase.
- Inspect adjacent parking, alleys, and public routes.
- Refresh coverage when the site layout changes.
Questions property teams ask
Are these thefts from construction sites?
Not necessarily. They are accepted larceny-theft records within the selected radius of a permitted project; the source does not identify them as construction-site losses.
Why are there more matches than unique offense records?
Project buffers overlap. One offense record can be near several projects and can therefore appear in several project-relative comparisons.
Does the lower exposure-weighted active rate prove construction prevents theft?
No. The unweighted project summary points slightly higher during active periods, and neither comparison controls for factors needed to establish cause.
Why use 250 meters?
It is a practical nearby-area specification, not a natural boundary. The article also shows 100 m and 500 m results because the choice can change the direction.
Should a construction site reduce security because of this result?
No. The result says permit phase is an unstable proxy. Decisions should be based on the project's actual materials, access, layout, schedule, and surroundings.
Our methods, limits, and sources
How we calculated this
This is original 911 Sentinel research — we gathered the records, ran every calculation below, and published the aggregate dataset.
We matched accepted Seattle larceny-theft records to 3,997 permit projects, normalized matches by project-days, and compared pre, permit-active, and post periods.
- We required usable permit issue and completion dates.
- We created equal-length pre and post comparison periods around each active interval.
- We counted project-relative record matches within 100 m, 250 m, and 500 m.
- We compared exposure-weighted rates with equally weighted project distributions and quantified repeated matching.
What this analysis cannot establish
- Permit dates proxy project activity and do not establish the timing or intensity of work.
- Nearby police records are not identified as thefts from construction sites.
- SPD coordinates are approximate and overlapping project buffers repeat records.
- The comparisons are descriptive and unadjusted for neighborhood, season, project type, or exposure.
- Association cannot establish an increase or decrease caused by construction.
Sources
The raw records come from the sources below; the study design, analysis, charts, and conclusions are our own.
Related questions and practical guides
New coverage — intake open
Review the project as it changes
A construction assessment can map deliveries, material storage, perimeter openings, access routes, lighting, and coverage changes to the actual build schedule.