Research · 6 min read
Seattle’s 311 Surge Was a Reporting Story, Not a Crime Story
311 is a reporting process: its volume surged independently of recorded crime, and it differs sharply by precinct.
Seattle reports a lot more disorder than it used to. Encampment reports climbed 60% across the study window and graffiti 57%, while recorded property crime stayed roughly flat. The surge tracks a shift in how people file requests as much as anything on the street — and precincts report very different amounts of disorder per crime recorded. That gap is the real finding about 311.
Our analysis of 925,306 requests across 43 service types, July 1, 2023–June 30, 2026.
Key takeaways
- Encampment reports rose 60% and graffiti 57% across 35 months while recorded property crime stayed roughly flat (about 2,400–4,100 records a month).
- About nine in ten requests now arrive through the city’s app, so the surge is largely a reporting-channel and engagement story.
- Precincts report between 3.70 and 5.37 disorder reports per recorded property crime — a 45% spread that measures reporting, not risk.
- Month-to-month lead-lag tests between disorder and recorded crime are weak and direction-unstable. 311 is a demand signal, not a crime forecast.
What does Seattle’s 311 disorder data actually measure?
Across 925,306 requests spanning 35 months, encampment reports rose 60% and graffiti 57% while recorded property crime stayed roughly flat, and about nine in ten requests now arrive through the city app. Disorder reports per recorded property crime range from 3.70 in the West to 5.37 in the South. 311 measures who reports and how — not what happened.
The numbers behind the answer
Selected measures only. Denominators and interpretation stay attached so the headline cannot stand alone.
925,306
Disorder requests in 35 months
Across 43 service types, from a feed that runs from 2013 to the present.
+60%
Change in encampment reports
First six months versus last six months of the window (3,720 to 5,955 per month).
3.7–5.4
Disorder reports per recorded property crime
West precinct reports 3.70 per recorded record, South 5.37 — a 45% spread across five precincts.
Disorder reports surged while recorded property crime did not
Across 35 months, the four largest disorder categories all rose: encampment reports 60% and graffiti 57%, with abandoned-vehicle reports up 19% and illegal dumping 13%. In raw volume the biggest single category was unauthorized encampment at 182,544 requests, ahead of abandoned vehicles (170,797), graffiti (93,310), and illegal dumping and needles (82,457). Recorded property crime over the same window did not follow — it stayed inside a band of roughly 2,400 to 4,100 records a month with no matching climb. The two lines simply do not move together.
That alone should change how you read a 311 trend. A rising disorder line is evidence of rising reporting — more people using the app, more attention, more service demand. It is not a crime forecast, and treating it as one mistakes a reporting surge for a safety decline.
The contrast is sharpest at the endpoints. In the first six months of the window, encampment reports averaged 3,720 a month; in the last six they averaged 5,955. Graffiti went from 1,888 to 2,967. The full disorder total reached 925,306 requests across 43 service types — a volume of civic attention that recorded crime never mirrored.
Evidence visual
Change in disorder reports, first six months vs last six
Percent change in monthly volume for the four largest disorder categories, July 2023–June 2026.
Nine in ten requests now arrive through the app
A reporting surge needs a reporting mechanism, and Seattle’s is the app. Month by month, the phone and web channels stay small while app submissions carry almost the entire volume. In August 2025, the busiest month in the extract, the app took 31,612 requests against 551 by web and 1,165 by phone — the app alone was about 94% of that month’s intake. Back in July 2023 the split was already lopsided: 23,314 app to 391 web and 1,170 phone.
That concentration matters for interpretation. When one easy channel grows, measured disorder can rise even if street conditions are steady, because the cost of filing a request fell. The later months also show a slow increase in the “other” intake category, a reminder that channels themselves keep changing. Any comparison across time is partly a comparison of how people reported.
Reading rule
Read a rising 311 line as rising reporting, not rising crime.
The lead-lag relationship is weak and direction-unstable
If disorder reliably foreshadowed property crime, the disorder series should lead the crime series at a consistent lag. It does not. Across the four largest categories the month-to-month relationship is weak and flips sign with the lag: illegal dumping sits at −0.66 at a six-month lag, encampment at −0.56, and graffiti at −0.24 raw against +0.45 on the differenced series. A negative value here means more disorder reports line up with fewer recorded offenses — the opposite of a warning signal.
The formal tests agree. Granger-style checks in both directions mostly fail to clear conventional thresholds, and where a p-value does dip below 0.05 the direction does not hold across lags — encampment shows crime leading disorder at one lag and not the next. The honest reading is that 311 and recorded crime are two different processes, and neither cleanly predicts the other at a monthly cadence.
Precincts report very different amounts of disorder per crime
Normalizing by recorded property crime makes the reporting difference visible. Across Seattle’s five precincts, disorder reports per recorded property-offense record range from 3.70 in the West to 5.37 in the South — a 45% spread. Two precincts with similar crime counts can look very different on a 311 map.
The raw counts show how much of that is volume and how much is ratio. North filed 197,691 disorder requests against 40,866 property records; East filed 88,861 against 21,117; West 98,736 against 26,660; Southwest 57,472 against 11,541; and South 86,239 against 16,052. North files more than twice as many requests as East but only about 1.9 times the property records — and South, mid-pack on volume, carries the highest ratio of all.
This is a reporting signal, not a risk ranking. It reflects who uses the app, what gets routed, and how residents and businesses engage — not which area is more or less safe. Read it as a measure of demand and attention, and never as a neighborhood verdict.
Evidence visual
Disorder reports per recorded property crime, by precinct
Aggregate 311 disorder reports divided by recorded property-offense records, July 2023–June 2026.
Treat 311 as a demand signal, and normalize before comparing
None of this makes 311 useless. It makes it a report on reporting: which services people ask for, where engagement is high, and how the mix is changing. As a measure of demand, attention, and service load, it is genuinely informative.
Before comparing places or periods, normalize by population and channel, and resist the pull toward causation. This study normalizes at the precinct level; tract-level population normalization is a separate, geocoded analysis. The checklist below is the standard this release applies.
What it cannot do is stand in for incidence. A block can generate few requests because it is quiet or because its occupants report little; a busy corridor can generate many for either reason. Fix the numerator, the denominator, and the channel before you compare — and keep the causal claim off the table.
- Read 311 as reporting behavior, not incidence.
- Compare the disorder trend to recorded crime, not instead of it.
- Normalize volume by population and reporting channel.
- Use precinct ratios to spot reporting differences, not risk.
- Never use 311 to forecast a specific property.
The volume is concentrated, and the lead-lag flips with the method
Six service types carry most of the 925,306 requests. Unauthorized encampment leads at 182,544, ahead of abandoned vehicles (170,797), graffiti (93,310), illegal dumping and needles (82,457), parking (74,115), and general inquiry (56,258) — a disorder “trend” is largely a story about the biggest two categories.
The lead-lag tests are direction-unstable in a way that rules out a clean warning signal. Illegal dumping correlates with property crime at −0.659 raw but −0.457 once the series are differenced at a six-month lag; encampment runs −0.564 raw at the same lag; abandoned vehicles flip from +0.081 raw to +0.503 differenced. When differencing moves the sign, the shared volume trend is doing the work, not a real coupling.
That is the established reading of service-request data. Sampson and Raudenbush found disorder and crime share underlying causes and that perceived disorder is shaped by neighborhood context, and a review of 311 data as a social indicator warns it reflects participation and routing as much as conditions. Treat the surge as a demand signal: pair it with recorded crime and keep the property-level view separate.
Frame
311 is a demand and engagement signal; the disorder literature warns against reading it as incidence.
Questions property teams ask
Can 311 reports predict property crime?
Not reliably. Encampment and graffiti reports rose sharply while recorded property crime stayed roughly flat, and the month-to-month lead-lag is weak and direction-unstable.
Does a 60% rise in encampment reports mean encampments rose 60%?
No. It means reports rose 60%. Volume reflects both underlying activity and how many people report, so it cannot be read as an incidence change on its own.
Why does the reporting channel matter?
About nine in ten requests now arrive through the city app. When one low-effort channel grows, measured disorder can rise even if street conditions are steady, because the cost of filing a request fell.
Do disorder reports lead property crime?
In this data, no clear lead appears. Correlations are weak and change sign by lag, and Granger-style tests are mostly insignificant and direction-unstable across categories.
Does a high precinct ratio mean that precinct is worse?
No. It means more disorder is reported there relative to recorded property crime. That is a reporting signal, not a safety ranking.
At what geography is this analysis normalized?
Disorder reports are normalized against recorded property crime at the precinct level. Tract-level population normalization is a separate, geocoded analysis and is outside this study’s scope.
Is 311 a reasonable measure of neighborhood disorder?
Only with care. Research on service-request data shows volume reflects reporting behavior, access, and administrative routing as much as conditions on the street, so it is a demand signal rather than a disorder census.
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 aggregated Seattle 311 service requests and SPD property-offense records to monthly and precinct-level counts (July 2023–June 2026), compared category trends, decomposed intake by channel, computed disorder reports per recorded property crime by precinct, and examined lead-lag on differenced series.
- We pulled monthly 311 counts by service type, by precinct, and by intake method for 35 months and 43 service types.
- We pulled monthly property-offense counts by precinct and by month.
- We computed per-category first-six vs last-six-month change, channel shares, and precinct disorder-to-crime ratios.
- We compared raw and differenced lead-lag correlations and ran Granger tests on differenced series in both directions.
- We dropped the trailing current month as incomplete.
What this analysis cannot establish
- 311 is a reporting process, not a disorder census; reporting propensity varies by neighborhood.
- Precinct comparisons are aggregate and ecological; they describe reporting behavior, not risk.
- Normalization is at the precinct level; tract-level population normalization is outside this study’s scope.
- Intake-channel shares describe how requests were filed, not who filed them or why.
- The trailing current month is dropped as incomplete.
- Monthly resolution can hide shorter dynamics.
Sources
The raw records come from the sources below; the study design, analysis, charts, and conclusions are our own.
- City of Seattle — Customer Service Requests (311)
- Seattle Police Department — SPD Crime Data: 2008-Present
- U.S. Census Bureau — American Community Survey (ACS) 5-Year Estimates
- Published aggregate result (JSON)
- Source manifest (JSON)
- Robert J. Sampson & Stephen W. Raudenbush (American Journal of Sociology) — Systematic Social Observation of Public Spaces: A New Look at Disorder in Urban Neighborhoods
- Ariel White & Kristine-Sarah Trump (Urban Affairs Review) — The Promises and Pitfalls of 311 Data
Related questions and practical guides
No obligation · free property walk
Separate reporting behavior from site conditions
A site assessment looks at the property itself — access, lighting, vacancy, and camera views — rather than inferring risk from citywide reporting patterns.