SAFEROUTE

Crime-data transparency · 29 cities · 4 countries · 2026-10-01

Which cities let you see their crime?

This is not a ranking of how safe places are. It measures something different and rarely looked at: how well each police force actually publishes its crime data — how current it is, how much detail each record carries, how much of the criminal code is visible, and how far back you can look.

Two British police forces publish essentially nothing

Both appear in the national open-data service. Both have force pages, neighbourhood listings and a published API. Neither supplies meaningful street-level crime, in their own city centres or anywhere else.

  • Gloucestershire Constabulary6 crimes in the busiest neighbourhood sampled
  • Greater Manchester Police2 crimes in the busiest neighbourhood sampled

Measured across 44 UK forces. For comparison, a point a few miles outside Greater Manchester — in Lancashire — returns 260 crimes for the same month, and Bristol returns 1,691.

Why this matters more than it sounds

Anyone building a "safest cities" ranking from open data — and plenty of people do — is quietly ranking publishing practice alongside crime. A city that releases nothing looks flawless. Greater Manchester and Gloucestershire would top a British safety league table built this way, purely because there is nothing to count.

The same problem runs through the detail. Washington D.C. publishes only Part I offences, so no simple assault, drugs, vandalism or public-order incidents appear at all. Toronto publishes five categories. New Orleans publishes 911 dispatches rather than confirmed reports. London's feed covers a rolling month; Dallas's covers a year. Put those side by side on one scale and the narrower feed always wins.

That is why SafeRoute's own safety index is calibrated within each city and never across them — and why the honest thing to publish globally is this: not who is safest, but who lets you check.

The index

Scored out of 100 across four equally weighted measures. Every bar below is the score taken apart — a number you cannot decompose is exactly what this page argues against.

FreshnessDetailBreadthHistory Not measurable — excluded

Measured on all four

23 cities whose feeds could be tested on every measure. Dallas leads on 99; Toronto is lowest of this group on 53.

CityScoreHow it breaks down
1 🇺🇸 DallasDallas Open Data 99
2 🇺🇸 Los AngelesLA City Open Data 98
3 🇺🇸 San FranciscoDataSF 93
4 🇺🇸 SeattleSeattle Open Data 93
5 🇺🇸 NashvilleNashville Open Data 93
6 🇺🇸 ClevelandCity of Cleveland Open Data 93
7 🇺🇸 Kansas CityKansas City Open Data 93
8 🇺🇸 ChicagoChicago Data Portal 92
9 🇺🇸 DetroitCity of Detroit Open Data 91
10 🇺🇸 New OrleansCity of New Orleans / OPCD 91 1000m probe box: 2000m exceeds its row cap
11 🇺🇸 MemphisCity of Memphis Open Data 90
12 🇺🇸 BaltimoreOpen Baltimore 89
13 🇺🇸 DenverOpen Data Denver 87
14 🇺🇸 Las VegasLVMPD Open Data 84 no incident time
15 🇺🇸 Washington DCOpen Data DC 80 only 6 offence categories
16 🇺🇸 CharlotteCity of Charlotte Open Data 78 no incident time
17 🇺🇸 MinneapolisCity of Minneapolis Open Data 76 no incident time
18 🇺🇸 HartfordCity of Hartford 76 no incident time
19 🇨🇦 VancouverVancouver Police Department 74 only 7 offence categories · no incident time
20 🇺🇸 Long BeachCity of Long Beach 68 no incident time
21 🇺🇸 TucsonCity of Tucson 63 42-day window · no incident time
22 🇺🇸 New YorkNYC Open Data 62
23 🇨🇦 TorontoToronto Police Service 53 only 5 offence categories

Measured on three of four

These 6 feeds report only the month of their newest record, not the date, so freshness could not be tested. They are scored on the other three and shown separately — ranking them against the group above would compare different measurements, which is the error this page exists to point out.

CityScoreHow it breaks down
1 🇺🇸 Fort WorthCity of Fort Worth Open Data 91 freshness not measurable
2 🇺🇸 BostonAnalyze Boston 89 freshness not measurable
3 🇺🇸 PhiladelphiaOpenDataPhilly 88 freshness not measurable
4 🇲🇽 Mexico CityHoyo de Crimen (api.hoyodecrimen.com) · CDMX FGJ 85 freshness not measurable · only 7 offence categories
5 🇺🇸 San DiegoCity of San Diego 68 freshness not measurable · no incident time
6 🇬🇧 Londondata.police.uk 49 freshness not measurable · 31-day window · no incident time

Not measured this time: Houston — its feed could not be reached when this snapshot was taken. No score is shown rather than a wrong one.

How it is measured

Each city's live feed is queried directly — a 2000m box at each city's central coordinate, halved (not below 500m) where a provider refuses the pull at its row cap. Nothing here is taken on trust or from documentation; every value comes from asking the feed what it actually returns today.

Publishing at all is a gate, not a scored category. A force that releases nothing is not "40% of a good feed" — it is the headline above, and it is excluded from the table rather than given a low score.

Where something could not be measured, it is excluded and the score normalised over the rest, marked with a hatched segment. 6 of the 29 feeds report only the month of their newest record rather than the date, so freshness is not scored for those — scoring it would rank cities on our own summarising rather than on their data.

Weights: freshness 25 · detail 25 · breadth 25 · history 25. Disagree with them? The underlying components are all shown, so any row can be recomputed by hand.

Walking somewhere tonight?

SafeRoute scores every walking route against this same live crime data, and shows how much of each route runs on lit streets. Free, no account.

Get SafeRoute on the App Store