STATS19 DATA LAB · GREAT BRITAIN
Which road risks persist?
Compare injury-collision outcomes, test geographic persistence and explore the conditions behind the totals.
Comparable road outcomes
The panel contains only authorities with valid, unflagged mileage and outcomes in every released year. Great Britain geography; eligible authority totals are not whole-country totals.
Great Britain · fixed panel
Vertical marks show conditional 95% Poisson intervals. They exclude traffic-estimation, reporting and spatial uncertainty.
| Year | Authorities | Excluded | Injury collisions | Adjusted KSI | Motor miles | Selected outcome |
|---|---|---|---|---|---|---|
| 2016 | 190 | 17 | 124,267 | 28,316.8 | 289,941,000,000 | 42.86 |
| 2017 | 190 | 17 | 119,059 | 27,395.3 | 293,991,100,000 | 40.5 |
| 2018 | 190 | 17 | 112,309 | 27,879.3 | 295,449,700,000 | 38.01 |
| 2019 | 190 | 17 | 107,763 | 26,992.6 | 299,246,700,000 | 36.01 |
| 2020 | 190 | 17 | 83,751 | 21,101.3 | 235,369,100,000 | 35.58 |
| 2021 | 190 | 17 | 93,105 | 24,043.6 | 263,254,200,000 | 35.37 |
| 2022 | 190 | 17 | 97,758 | 26,291.2 | 286,397,700,000 | 34.13 |
| 2023 | 190 | 17 | 95,724 | 26,295 | 292,640,700,000 | 32.71 |
| 2024 | 190 | 17 | 91,727 | 26,156.2 | 297,202,300,000 | 30.86 |
| 2025 | 190 | 17 | 92,007 | 26,873.8 | 302,823,900,000 | 30.38 |
Method & limits
STATS19 records police-reported personal-injury collisions on public roads in Great Britain. Damage-only incidents, insurance claims costs and fault are not measured here.
The annual geographic view defaults to the same eligible authorities in every year. Rates divide pooled outcomes by matched mileage. Adjusted KSI means killed or seriously injured casualties; adjusted FSC means fatal or serious collisions. These estimates account for changes in severity recording.
Persistence uses frozen discovery groups, a separate pandemic stress period and later-year comparisons. Common reporting differences and road mix can persist too. Descriptive persistence does not establish a causal effect or insurer pricing lift.
Detailed collision and vehicle records are joined using collision identifiers, retaining unknown categories. STATS19 specification changes around 2024 affect some classifications. Source vintages and code mappings are archived with the dataset.
Every view, export and Actuary Lens interpretation reads the same immutable release. A missing denominator produces an unavailable rate, never zero.
Record-level validation
513,801 collision records · 937,265 vehicle records · 652,821 casualty records.
- Unique collision, vehicle and casualty keys; no orphan records.
- Every collision’s vehicle and casualty record counts reconcile to its reported totals.
- Every dimension, including unknown categories, reconciles to annual source record counts.
- DFT adjusted serious-collision weights are within [0,1]; fatal collisions receive weight 1.
- Vehicle-mile joins use all 11 distinct GB regions and compatible vehicle classes.
- Licensed stock uses Great Britain, Cars, Licensed and four quarter-end observations per year.
- All raw inputs and the source codebook are retained with SHA-256 checksums.
Archived inputs and exact source vintages
These files are the frozen inputs to this release. Each download is gzip compressed; source hashes identify the uncompressed bytes.
- STATS19 collision records, latest five final years · 2021–2025 · retrieved 2026-09-25
Official source file · Archived input06a03c0e5504175e23ed98e3e0b4767ac61d396c073405a1bddab9339c18f948Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- STATS19 vehicle records, latest five final years · 2021–2025 · retrieved 2026-09-25
Official source file · Archived input05453dca2651ad8a27b4ac0e6a0450a852c506e4d495111ffca398f3d311e26aOriginal bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- STATS19 casualty records, latest five final years · 2021–2025 · retrieved 2026-09-25
Official source file · Archived input870be5a934ae10baf5c52163d59eccba0ab2f40327ffa6b2bd354d2e7ee6b108Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- STATS19 codebook and historical mappings · 2021–2025 · retrieved 2026-09-25
Official source file · Archived input51b08606ef3dad1ae5d4c8726241b022fc8e4bbdc910ef0e5cc2fb92b054fc12Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- Regional vehicle mileage · 2021–2025 · retrieved 2026-09-25
Official source file · Archived input2db1cc2ac20f6a0cce917207b9f8c8d310119393b67e7bdc94dcb1adfa19d480Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- Regional road-category mileage · 2021–2025 · retrieved 2026-09-25
Official source file · Archived inputb07cbe882042158d2bc2650bb71c865187cb4694f3069597f6d867d7ef636441Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
- VEH0120 GB licensed vehicle stock · 2021–2025 · retrieved 2026-09-25
Official source file · Archived inputafdc4feb04a9577c84080926e06769fe60fc8d0a445df1032a95ec2bb7766a92Original bytes archived losslessly as gzip; sha256 applies to decompressed bytes. Initial reproducible extraction, no human source review asserted.
Sources & release
14 sources need attention. The source table is open below; each row explains whether a check failed or new evidence awaits review.
Inspect 14 relevant sources
| Source | Recorded source period | Checked in release | Validated in release | Current source check |
|---|---|---|---|---|
| RAS0403: collisions and casualties by local authority | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Local authority motor vehicle mileage | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Rainfall — England | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Rainfall — Wales | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Rainfall — Scotland | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Air frost — England | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Air frost — Wales | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Air frost — Scotland | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Wet days ≥1 mm — England | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Wet days ≥1 mm — Wales | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| Wet days ≥1 mm — Scotland | 2016–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| STATS19 collision records, latest five final years | 2021–2025 | 21/09/2026 | Not recorded | New evidence awaiting review — not yet in findings |
| Regional vehicle mileage | 2021–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
| VEH0120 GB licensed vehicle stock | 2021–2025 | Not recorded | Not recorded | New evidence awaiting review — not yet in findings |
Technical release ID
lens-2026-09-25-a1fc94c9307e