Road danger, city by city
People killed or seriously injured on city roads, mapped area by area and, where traffic is counted, road by road, and measured against the people there and the traffic. More cities are added as their open data allows.
- Year
- 2026
- My role
- Solo project, extending my London road safety rebuild to other cities and to individual roads
- Tools
- Python, pandas, GeoPandas and Leaflet
- Data
- DfT STATS19 collision and casualty records, DfT road traffic counts, OS Open Roads, 2011 Census via Nomis, Toronto Police Service KSI and traffic collision open data and City of Toronto neighbourhoods and 2021 neighbourhood profiles
The question
How dangerous are a city’s roads once you allow for how many people are there and how far they travel? My London road safety project answered that for London’s wards. This project asks the same question of other cities, with the same measures so they can be compared, and of individual roads, each measured against its own traffic.
What I did
- Built one pipeline that turns each city’s open collision data into the same set of files: casualties by small area, year, road user and severity, with the population and, where it exists, the traffic in each area. London’s figures come out identical to my first London version.
- Rated London’s main roads one stretch at a time. The Department for Transport counts traffic at points on every motorway and A road. I rebuilt the stretch of road around each count point from Ordnance Survey’s open road map, matched each collision on a numbered main road to the nearest stretch of that road, and divided by the traffic on it: 1,950 stretches covering 1,766 km, with 96.5% of the collisions on those roads matched to a stretch.
- Adjusted for chance. One or two serious collisions on a quiet road can produce a huge rate, so each stretch’s rate is pulled toward London’s average in proportion to how little traffic it carries, a method called empirical Bayes. The map recalculates this for whatever years you pick.
- Mapped Toronto by neighbourhood from Toronto Police’s records of everyone killed or seriously injured from 2006 to 2023, with residents and walking commuters from the 2021 Census.
- Added Birmingham, Bristol, Cardiff, Edinburgh, Glasgow, Leeds, Liverpool, Manchester, Newcastle upon Tyne and Sheffield from the same national data, each mapped by small area (2011 wards, or the Scottish Government’s intermediate zones in Scotland) and, on its counted motorways and A roads, road by road.
- Added 72 more cities in 12 other countries, each from its own police, city or national open data: Argentina (Buenos Aires), Australia (Brisbane, Melbourne and Sydney), Brazil (São Paulo), Canada (Montréal and Ottawa), Colombia (Bogotá), France (Lyon, Marseille and Paris), Germany (Augsburg, Berlin, Bielefeld, Bochum, Bonn, Cologne, Dortmund, Dresden, Duisburg, Düsseldorf, Essen, Frankfurt, Gelsenkirchen, Hamburg, Hamm, Herne, Ingolstadt, Krefeld, Leipzig, Leverkusen, Munich, Mönchengladbach, Mülheim an der Ruhr, Münster, Nuremberg, Oberhausen, Solingen, Wuppertal and Würzburg), Hong Kong (Hong Kong), Japan (Chiba, Fukuoka, Hiroshima, Kawasaki, Kitakyushu, Kobe, Kumamoto, Kyoto, Nagoya, Niigata, Okayama, Osaka, Sagamihara, Saitama, Sakai, Sapporo, Sendai, Shizuoka, Tokyo and Yokohama), New Zealand (Auckland, Christchurch and Wellington), Spain (Barcelona and Madrid) and United States (Austin, Chicago, New York City, San Francisco, Seattle and Washington, DC). Each is mapped by the smallest areas its population figures cover, except São Paulo, which has headline figures only.
Explore the map
Highest in this view
| Area | Rate a year | People a year |
|---|
What I found so far
Per kilometre driven, London’s most dangerous main roads are central. From 2022 to 2025, London’s counted motorways and A roads saw 13.4 people killed or seriously injured per 100 million vehicle-km. The highest rates, even after the adjustment for chance, are on short central stretches with little motor traffic and many people on foot, led by the A40 in Westminster at about 163 per 100 million vehicle-km.
Toronto’s roads are deadlier per resident than London’s. From 2021 to 2025, Toronto averaged 49.2 road deaths a year, or 1.76 per 100,000 residents. London averaged 97.2, or 1.10 per 100,000, over the same years, so Toronto’s rate is about 60% higher. Each city is the area its own government publishes: Toronto is the City of Toronto, with 2,794,356 residents in the 2021 Census (Statistics Canada), and London is Greater London, the 32 boroughs and the City of London, with 8,799,800 in the 2021 Census (ONS). Both counts include deaths within 30 days of a collision.
British cities can be compared on serious injuries too, because they all use the same national records. From 2021 to 2025, the number of people killed or seriously injured per 100,000 residents a year ranged from 15.9 in Cardiff to 59.7 in Leeds. Each British city here is its council’s area, except London, which is all of Greater London; the table below gives every boundary and population source.
Pedestrians are about half of road deaths in London and Toronto: 53% in Toronto from 2019 to 2023, and 50% in London from 2021 to 2025.
How the cities compare
What each city counts
| City | Boundary, as the city publishes it | Population and source | Deaths counted within | Years mapped |
|---|---|---|---|---|
| Austin | The City of Austin, within its city limits | 961,855: U.S. Census Bureau, 2020 Census (P.L. 94-171), Austin city, Texas | 30 days | 2010 to 2025 |
| São Paulo | The municipality of São Paulo (IBGE code 3550308) | 11,451,999: IBGE, 2022 Census (table 4714), municipality of São Paulo | 30 days | 2015 to 2025 |
| Bogotá | The Capital District of Bogotá, mapped by its 112 UPZs (urban planning zones) and 20 localidades | 8,034,649: DANE and the Secretaría Distrital de Planeación, population projections by localidad (2018 Census basis) | 30 days | 2010 to 2025 |
| Washington, DC | The District of Columbia, which is the whole city of Washington | 689,545: U.S. Census Bureau, 2020 Census (P.L. 94-171), Washington city, District of Columbia | Not stated | 2010 to 2025 |
| Chicago | The City of Chicago, within its city limits | 2,746,388: U.S. Census Bureau, 2020 Census (P.L. 94-171), City of Chicago | 30 days | 2018 to 2025 |
| Seattle | The City of Seattle, within its city limits | 737,015: U.S. Census Bureau, 2020 Census (P.L. 94-171), Seattle city, Washington | 30 days | 2016 to 2025 |
| San Francisco | The City and County of San Francisco | 873,965: U.S. Census Bureau, 2020 Census (P.L. 94-171), San Francisco city, California | 30 days | 2010 to 2025 |
| Buenos Aires | The Autonomous City of Buenos Aires (CABA), mapped by its 15 comunas | 3,121,707: INDEC, 2022 Census, resident population by comuna (as published by IECBA) | 30 days | 2019 to 2025 |
| New York City | New York City: the five boroughs, within the city limits | 8,804,190: U.S. Census Bureau, 2020 Census (P.L. 94-171), New York city, New York | Not stated | 2013 to 2025 |
| Marseille | The commune of Marseille, made up of its 16 arrondissements | 886,040: INSEE, latest legal population of the commune, via geo.api.gouv.fr | 30 days | 2019 to 2024 |
| Hamm | The City of Hamm, mapped by its 7 Stadtbezirke | 179,108: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (178,788 residents in all) | 30 days | 2019 to 2025 |
| Auckland | The area governed by Auckland Council (territorial authority Auckland) | 1,656,417: Stats NZ, 2023 Census usually resident population, summed over the 627 SA2s in Auckland | 30 days | 2010 to 2025 |
| Ottawa | The City of Ottawa, including its rural wards | 1,017,449: Statistics Canada, 2021 Census, City of Ottawa census subdivision | Not stated | 2013 to 2020 |
| Christchurch | The area governed by Christchurch City Council (territorial authority Christchurch City) | 391,359: Stats NZ, 2023 Census usually resident population, summed over the 176 SA2s in Christchurch City | 30 days | 2010 to 2025 |
| Melbourne | The City of Melbourne: the central council area (the CBD and nearby suburbs), not Greater Melbourne | 177,402: City of Melbourne, population by small area for 2023 (from its 2020 to 2040 forecasts) | 30 days | 2012 to 2025 |
| Okayama | The City of Okayama, mapped by its 4 wards | 707,387: The City of Okayama, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Shizuoka | The City of Shizuoka, mapped by its 3 wards | 658,781: The City of Shizuoka, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Mönchengladbach | The City of Mönchengladbach, mapped by its 44 Stadtteile and 4 Stadtbezirke | 267,176: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (263,914 residents in all) | 30 days | 2019 to 2025 |
| Leeds | The area governed by Leeds City Council (local authority E08000035) | 811,956: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Manchester | The area governed by Manchester City Council (local authority E08000003) | 551,938: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Birmingham | The area governed by Birmingham City Council (local authority E08000025) | 1,144,919: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Sheffield | The area governed by Sheffield City Council (local authority E08000019) | 556,521: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Sydney | The City of Sydney, the area governed by the City of Sydney council (the ABS's local government area 'Sydney') | 248,706: Australian Bureau of Statistics, estimated resident population at 30 June 2025, summed over the 16 SA2s in the City of Sydney | 30 days | 2018 to 2024 |
| Bielefeld | The City of Bielefeld, mapped by its 10 Stadtbezirke | 331,419: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (329,764 residents in all) | 30 days | 2019 to 2025 |
| Würzburg | The City of Würzburg, mapped by its 13 Stadtteile | 133,258: Destatis, Gemeindeverzeichnis: official population at 31 December 2025 (Census 2022 basis); the map's areas use the 2022 Census's 100 m grid (131,308 residents in all) | 30 days | 2018 to 2025 |
| Brisbane | The area governed by Brisbane City Council (the ABS's local government area 'Brisbane') | 1,375,046: Australian Bureau of Statistics, estimated resident population at 30 June 2025, summed over the 135 SA2s in Brisbane City | 30 days | 2010 to 2024 |
| Wellington | The area governed by Wellington City Council (territorial authority Wellington City) | 202,683: Stats NZ, 2023 Census usually resident population, summed over the 86 SA2s in Wellington City | 30 days | 2010 to 2025 |
| Chiba | The City of Chiba, mapped by its 6 wards | 994,846: The City of Chiba, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Lyon | The commune of Lyon, made up of its 9 arrondissements | 519,127: INSEE, latest legal population of the commune, via geo.api.gouv.fr | 30 days | 2019 to 2024 |
| Düsseldorf | The City of Düsseldorf, mapped by its 50 Stadtteile and 10 Stadtbezirke | 619,444: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (611,152 residents in all) | 30 days | 2019 to 2025 |
| Paris | The commune of Paris, made up of its 20 arrondissements | 2,103,778: INSEE, latest legal population of the commune, via geo.api.gouv.fr | 30 days | 2019 to 2024 |
| Sakai | The City of Sakai, mapped by its 7 wards | 802,648: The City of Sakai, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Leverkusen | The City of Leverkusen, mapped by its 13 Stadtteile and 3 Stadtbezirke | 168,299: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (166,341 residents in all) | 30 days | 2019 to 2025 |
| Münster | The City of Münster, mapped by its 22 Stadtteile and 6 Stadtbezirke | 306,368: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (303,595 residents in all) | 30 days | 2019 to 2025 |
| Toronto | The City of Toronto, the amalgamated city that includes the former East York, Etobicoke, North York, Scarborough and York | 2,794,356: Statistics Canada, 2021 Census, City of Toronto census subdivision | 30 days | 2006 to 2023 |
| Kitakyushu | The City of Kitakyushu, mapped by its 7 wards | 903,592: The City of Kitakyushu, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Kobe | The City of Kobe, mapped by its 9 wards | 1,494,997: The City of Kobe, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Mülheim an der Ruhr | The City of Mülheim an der Ruhr, mapped by its 9 Stadtteile and 3 Stadtbezirke | 171,674: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (170,976 residents in all) | 30 days | 2019 to 2025 |
| Krefeld | The City of Krefeld, mapped by its 11 Stadtteile and 9 Stadtbezirke | 230,738: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (230,532 residents in all) | 30 days | 2019 to 2025 |
| Ingolstadt | The City of Ingolstadt, mapped by its 12 Stadtbezirke | 141,185: Destatis, Gemeindeverzeichnis: official population at 31 December 2025 (Census 2022 basis); the map's areas use the 2022 Census's 100 m grid (136,407 residents in all) | 30 days | 2018 to 2025 |
| Kumamoto | The City of Kumamoto, mapped by its 5 wards | 730,882: The City of Kumamoto, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Newcastle upon Tyne | The area governed by Newcastle City Council (local authority E08000021) | 300,125: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Leipzig | The City of Leipzig, mapped by its 63 Ortsteile and 10 Stadtbezirke | 633,592: Stadt Leipzig, Amt für Statistik und Wahlen: residents with their main home in Leipzig, 31 December 2025 (the city's register) | 30 days | 2016 to 2025 |
| Cologne | The City of Cologne (Köln), mapped by its 86 Stadtteile and 9 Stadtbezirke | 1,025,523: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the city's count of people with a main or second home there (1,100,076 at 31 December 2025) | 30 days | 2019 to 2025 |
| Hiroshima | The City of Hiroshima, mapped by its 8 wards | 1,171,915: The City of Hiroshima, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Niigata | The City of Niigata, mapped by its 8 wards | 759,139: The City of Niigata, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Osaka | The City of Osaka, mapped by its 24 wards | 2,817,490: The City of Osaka, population estimate, 1 September 2026 | 24 hours | 2019 to 2025 |
| Liverpool | The area governed by Liverpool City Council (local authority E08000012) | 486,088: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Kyoto | The City of Kyoto, mapped by its 11 wards | 1,430,062: The City of Kyoto, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Oberhausen | The City of Oberhausen, mapped by its 3 Stadtbezirke | 213,178: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (211,550 residents in all) | 30 days | 2019 to 2025 |
| Sagamihara | The City of Sagamihara, mapped by its 3 wards | 718,908: The City of Sagamihara, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Glasgow | The area governed by Glasgow City Council (council area S12000049) | 654,330: National Records of Scotland, mid-2025 population estimate, via Nomis | 30 days | 2010 to 2025 |
| Augsburg | The City of Augsburg, mapped by its 17 Stadtteile | 301,105: Destatis, Gemeindeverzeichnis: official population at 31 December 2025 (Census 2022 basis); the map's areas use the 2022 Census's 100 m grid (294,632 residents in all) | 30 days | 2018 to 2025 |
| Hamburg | The city-state of Hamburg, mapped by its 99 Stadtteile | 1,862,565: Statistikamt Nord, official population at 31 December 2024 (Census 2022 basis); the map's Stadtteile use registered residents at 31 December 2024 (1,968,531 in all) | 30 days | 2018 to 2025 |
| Herne | The City of Herne, mapped by its 13 Stadtteile and 4 Stadtbezirke | 156,266: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (155,344 residents in all) | 30 days | 2019 to 2025 |
| Frankfurt | The City of Frankfurt am Main, mapped by its 16 Ortsbezirke | 756,021: Destatis, Gemeindeverzeichnis: official population at 31 December 2025 (Census 2022 basis); the map's areas use the 2022 Census's 100 m grid (743,448 residents in all) | 30 days | 2021 to 2025 |
| Nuremberg | The City of Nürnberg, mapped by its 10 Stadtteile | 529,508: Destatis, Gemeindeverzeichnis: official population at 31 December 2025 (Census 2022 basis); the map's areas use the 2022 Census's 100 m grid (522,429 residents in all) | 30 days | 2018 to 2025 |
| Bochum | The City of Bochum, mapped by its 28 Stadtteile and 6 Stadtbezirke | 358,880: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (354,103 residents in all) | 30 days | 2019 to 2025 |
| Montréal | The City of Montréal and its 19 boroughs (not the rest of the island's municipalities) | 1,762,949: Statistics Canada, 2021 Census, City of Montréal census subdivision | Not stated | 2012 to 2021 |
| Bristol | The area governed by Bristol City Council (local authority E06000023) | 472,465: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Gelsenkirchen | The City of Gelsenkirchen, mapped by its 18 Stadtteile and 5 Stadtbezirke | 266,199: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (264,123 residents in all) | 30 days | 2019 to 2025 |
| Solingen | The City of Solingen, mapped by its 5 Stadtbezirke | 164,621: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (164,527 residents in all) | 30 days | 2019 to 2025 |
| Nagoya | The City of Nagoya, mapped by its 16 wards | 2,354,205: The City of Nagoya, population estimate, 1 October 2026 | 24 hours | 2019 to 2025 |
| Bonn | The City of Bonn, mapped by its 51 Ortsteile and 4 Stadtbezirke | 323,587: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (321,351 residents in all) | 30 days | 2019 to 2025 |
| Hong Kong | The Hong Kong Special Administrative Region, mapped by its 18 District Council districts | 7,497,800: Census and Statistics Department, table 110-02001: land-based mid-year population, 2025 | 30 days | 2018 to 2025 |
| Kawasaki | The City of Kawasaki, mapped by its 7 wards | 1,559,571: The City of Kawasaki, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Wuppertal | The City of Wuppertal, mapped by its 10 Gemarkungen | 357,243: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (356,419 residents in all) | 30 days | 2019 to 2025 |
| Duisburg | The City of Duisburg, mapped by its 7 Stadtbezirke | 500,810: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (501,657 residents in all) | 30 days | 2019 to 2025 |
| Dortmund | The City of Dortmund, mapped by its 12 Stadtbezirke | 600,880: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (598,307 residents in all) | 30 days | 2019 to 2025 |
| Yokohama | The City of Yokohama, mapped by its 18 wards | 3,753,462: The City of Yokohama, population estimate, 1 October 2026 | 24 hours | 2019 to 2025 |
| Sapporo | The City of Sapporo, mapped by its 10 wards | 1,961,996: The City of Sapporo, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| London | Greater London: the 32 London boroughs and the City of London | 8,799,800: ONS, Census 2021: usual residents of Greater London | 30 days | 2010 to 2025 |
| Berlin | The city-state of Berlin, mapped by its 542 LOR planning areas (Planungsräume, 2021) | 3,685,265: Amt für Statistik Berlin-Brandenburg, official population at 31 December 2024 (Census 2022 basis); the map's areas use registered residents at 31 December 2025 (3,913,644 in all), the only count by planning area | 30 days | 2018 to 2025 |
| Barcelona | The municipality of Barcelona: its 10 districts and 73 neighbourhoods | 1,729,963: Ajuntament de Barcelona, municipal register (Padró), 1 January 2026 | 24 hours | 2016 to 2025 |
| Fukuoka | The City of Fukuoka, mapped by its 7 wards | 1,662,793: The City of Fukuoka, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Edinburgh | The area governed by the City of Edinburgh Council (council area S12000036) | 531,370: National Records of Scotland, mid-2025 population estimate, via Nomis | 30 days | 2010 to 2025 |
| Sendai | The City of Sendai, mapped by its 5 wards | 1,096,146: The City of Sendai, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Dresden | The City of Dresden, mapped by its 10 Stadtbezirke and 9 Ortschaften | 562,764: Statistisches Landesamt des Freistaates Sachsen, official population at 31 December 2025, as published on the city's open data portal; the map's areas use the city's register (571,510 with their main home in Dresden) | 30 days | 2016 to 2025 |
| Essen | The City of Essen, mapped by its 50 Stadtteile and 9 Stadtbezirke | 573,618: IT.NRW, official population at 31 December 2025 (Census 2022 basis, table 12411-01i); the map's areas use the 2022 Census's 100 m grid (570,607 residents in all) | 30 days | 2019 to 2025 |
| Munich | The City of Munich (Landeshauptstadt München), mapped by its 25 Stadtbezirke | 1,505,005: Bayerisches Landesamt für Statistik, official population at 31 December 2024 (Census 2022 basis); the map's Stadtbezirke use the city's register of residents with their main home there (1,603,776 in all) | 30 days | 2018 to 2025 |
| Tokyo | Tokyo's 23 special wards (the ku-bu, the former City of Tokyo), part of the Tokyo Metropolis | 9,991,053: Tokyo Metropolitan Government, population estimate for the 23 wards, 1 September 2026 | 24 hours | 2019 to 2025 |
| Saitama | The City of Saitama, mapped by its 10 wards | 1,344,865: The City of Saitama, population estimate, 1 October 2025 (Census) | 24 hours | 2019 to 2025 |
| Cardiff | The area governed by Cardiff Council (local authority W06000015) | 362,308: ONS, Census 2021 (table TS001), usual residents, via Nomis | 30 days | 2010 to 2025 |
| Madrid | The municipality of Madrid (Ayuntamiento de Madrid), mapped by its 131 barrios | 3,497,277: Ayuntamiento de Madrid, municipal register (Padrón municipal), 1 de enero de 2026, summed over the 131 barrios | 24 hours | 2019 to 2025 |
Killed or seriously injured, per 100,000 residents a year, for cities that define serious injuries the same way (STATS19): Leeds 59.7, Manchester 42.3, Birmingham 42.1, Sheffield 51.8, Newcastle upon Tyne 34.8, Liverpool 55.3, Glasgow 36.7, Bristol 47.4, London 43.2, Edinburgh 29.9, Cardiff 15.9.
Did street changes coincide with fewer casualties?
From 2010 to 2025, London’s streets changed: the share of its roads mapped at 20 mph rose from 0.8% to 52.7%, and mapped cycle tracks grew from 338 km to 1,414 km. I measured both for every ward and year from OpenStreetMap’s full edit history (through the ohsome API) and set them against the people killed or seriously injured there.
Wards that went 20 mph didn’t see bigger falls. I compared 2010 to 2014 with 2021 to 2025, grouping wards by whether at least 50 percentage points more of their roads were mapped at 20 mph, and whether they were in the top third of wards for new cycle track (at least 0.58 km per 10 km of road). The number killed or seriously injured fell 10.6% in wards that went 20 mph without new cycle track, and fell 16.5% in wards with neither change, against 14.0% fewer across London.
| What changed | Wards | KSI a year, 2010 to 2014 | KSI a year, 2021 to 2025 | Change |
|---|---|---|---|---|
| Neither | 180 | 954 | 797 | −16.5% |
| 20 mph only | 236 | 1,614 | 1,443 | −10.6% |
| Cycle tracks only | 86 | 587 | 499 | −15.0% |
| Both | 123 | 1,260 | 1,059 | −15.9% |
| All of London | 625 | 4,414 | 3,797 | −14.0% |
Comparing each ward with itself finds no clear link either. A regression of each ward’s yearly count on its street measures, allowing for every ward’s own level and for London-wide changes each year, puts each measure’s association within a wide range around zero. None is clearly different from no change.
| Measure | Associated change | 95% interval |
|---|---|---|
| All of a ward’s roads at 20 mph, against none | +3.6% | −4.6% to +12.6% |
| One more km of cycle track per 10 km of road | +0.2% | −3.6% to +4.2% |
| One more km of painted cycle lane per 10 km of road | −3.1% | −10.1% to +4.5% |
Outside London, 20 mph limits do coincide with bigger falls. I ran the same comparison for the 10 other British cities on this page, pooled: 500 wards and intermediate zones. The number killed or seriously injured fell 26.2% where areas went 20 mph without new cycle track and fell 36.3% where they gained both, against 9.9% fewer where neither changed. Comparing each area with itself, and letting each city have its own year-to-year changes, moving all of an area’s roads to 20 mph was associated with −13.1% (95% interval −21.9% to −3.2%). Cycle tracks again show no clear link (−1.9%, −6.8% to +3.2%).
| City | 20 mph, 2010 | 20 mph, 2025 | KSI a year, 2010 to 2014 | KSI a year, 2021 to 2025 | Change |
|---|---|---|---|---|---|
| Edinburgh | 1.3% | 72.3% | 319 | 159 | −50.2% |
| Bristol | 0.1% | 58.7% | 266 | 224 | −15.8% |
| Manchester | 0.0% | 45.1% | 284 | 233 | −17.9% |
| Cardiff | 0.0% | 42.2% | 103 | 58 | −44.3% |
| Liverpool | 0.0% | 31.6% | 292 | 269 | −7.8% |
| Sheffield | 0.1% | 29.8% | 336 | 288 | −14.2% |
| Newcastle upon Tyne | 0.0% | 23.3% | 137 | 104 | −23.8% |
| Leeds | 0.0% | 20.2% | 425 | 485 | +14.0% |
| Birmingham | 0.0% | 16.5% | 452 | 482 | +6.7% |
| Glasgow | 0.0% | 11.8% | 339 | 240 | −29.2% |
Edinburgh, with the most roads mapped at 20 mph (72.3%), also saw the biggest fall (−50.2%). Giving each city its own year-to-year changes shrinks the association from −28.1% to −13.1%, but it doesn’t remove it. London’s result may differ because many of its central wards already had low speeds and heavy walking and cycling before their limits changed.
OpenStreetMap measures mapping as well as building. Its count of London’s traffic-calming features went from 3,592 at the start of 2020 to 44,828 at the start of 2023, a mapping drive rather than a building one, so I left traffic calming out. A road can also be mapped at 20 mph long after its signs changed, which blurs any effect, and wards that changed differ from those that didn’t in ways a regression can’t fully remove, such as how much people walk and cycle there. These are associations, not causes.
Checking OpenStreetMap against TfL’s own survey. TfL’s Cycling Infrastructure Database recorded every cycle lane and track on London’s streets between January 2017 and September 2019. Ward by ward, OpenStreetMap at the start of 2019 ranks London’s wards broadly as the survey does: a rank correlation of 0.70 for cycle tracks and off-road routes, and 0.63 for painted lanes, where 1 would be perfect agreement. Its totals are lower, 1,062 km of cycle track against the survey’s 2,015 km, partly because the survey draws each side of a road separately and includes shared footways, which OpenStreetMap often maps as footpaths. Random errors like these tend to pull an estimate toward no association, so the lack of a clear link for cycle tracks is weaker evidence than it looks.
How the cities are measured
- London uses the Department for Transport’s STATS19 records of collisions reported to the police, which separate fatal, serious and slight injuries and correct serious injuries for changes in how police record them. Areas are 2011 wards and boroughs, measured against the workday population, residents and people walking to work (2011 Census), and against motor, cycle and bus traffic.
- Toronto uses Toronto Police’s records of people killed or seriously injured, which leave out slight injuries, and its separate count of deaths within 30 days for the headline figures. The detailed records count deaths within a year rather than 30 days, but the two counts match from 2014 to 2023. Toronto’s definition of a serious injury differs from Britain’s, so comparisons between the two cities use deaths only.
- Every other city uses its own police or city records. The table under the chart gives each city’s boundary, population source and definitions. Three differences matter most when comparing them:
- Deaths. Most cities count deaths within 30 days of a crash, the international standard, but the 20 cities in Japan, Barcelona and Madrid count only deaths within 24 hours. That leaves out people who die later, so their death rates run low.
- Serious injuries. Almost every country defines a serious injury differently, so the page compares serious injuries only among the British cities. Montréal, Bogotá, the 20 cities in Japan and New York City don’t separate serious from slight injuries at all.
- People or collisions. Augsburg, Berlin, Bielefeld, Bochum, Bonn, Cologne, Dortmund, Dresden, Duisburg, Düsseldorf, Essen, Frankfurt, Gelsenkirchen, Hamburg, Hamm, Herne, Hong Kong, Ingolstadt, Krefeld, Leipzig, Leverkusen, Mönchengladbach, Mülheim an der Ruhr, Munich, Münster, Nuremberg, Oberhausen, Solingen, Wuppertal and Würzburg come from Germany’s accident atlas, which records collisions rather than casualties, so their maps count collisions by their worst injury.
Caveats
- Only collisions reported to the police are included, and many slight injuries are never reported.
- Per-resident rates use one population figure for every year: each city’s latest census or official estimate.
- Road-by-road rates cover only London’s motorways and A roads, the roads with traffic counts, and the stretches are rebuilt around the count points, so their ends are approximate.
- Small areas and short stretches rest on small numbers, even after the adjustment for chance.
What’s next
- More cities where open data allow, especially in parts of the world the page doesn’t cover yet: Africa, South and Southeast Asia, and more of Latin America.
- Traffic counts for more cities, so their main roads can be measured against their traffic as Britain’s are.