Driving Mastery Research · 2025
Where and when young drivers crash
It isn't busy junctions in town. Across 15,065 reported collisions involving a car driver aged 17 to 24 in 2025, 11.2% happened in darkness, on a rural road, with a speed limit of 60mph or more — against 4.5% of collisions involving drivers 25 and over.
- Dark, rural and 60mph+
- 2.5×
- Involved losing control
- 20.0%
- People killed
- 269
- Casualties who were not the driver
- 34.3%
Analysis by Niall Cullen, Approved Driving Instructor
Analysis by Driving Mastery using Department for Transport STATS19 road casualty statistics, published under the Open Government Licence v3.0. DfT does not endorse this analysis.
Executive summary
Key findings
-
1
The dangerous combination is darkness, a rural road and a 60 limit.
11.2% of young-driver collisions had all three, against 4.5% for drivers 25 and over.
-
2
One in five involved the car skidding or overturning.
20.0% against 8.8% — the single widest behavioural gap in the data.
-
3
Most were simply driving along, not performing a manoeuvre.
59.3% were going ahead, and 47.3% were nowhere near a junction.
-
4
A third of the people hurt were not the driver.
Of 21,995 casualties, 7,546 were passengers or people outside the car. 269 people died.
Every comparison on this page is a share of each group's own collisions. This data counts collisions, not miles driven, so none of it says how likely anyone is to crash — see what this cannot tell you.
Where and when
Dark roads, fast roads, country roads
Each condition on its own is more common in young-driver collisions. Together they compound: the three at once account for 11.2% of young-driver collisions and 4.5% of the comparison group.
Coloured bar: drivers 17–24. Thin grey bar beneath: drivers 25 and over, on the same scale.
Speed limit of the road
Share of each group's collisions, by hour
Solid line: drivers 17–24. Dashed: drivers 25 and over. Both are shares of their own group's collisions, so the lines are comparable in shape rather than in volume.
Losing control
One in five involved a skid or a roll
DfT records whether a vehicle skidded, overturned, or both. It is the widest behavioural gap in the dataset: 20.0% of cars driven by 17 to 24 year olds in a reported collision had skidded or overturned, against 8.8% for drivers 25 and over.
Drivers 17–24
20.0%
of 15,691 cars with a known outcome
Drivers 25 and over
8.8%
of 77,472 cars with a known outcome
Skidding is not itself a cause — it is what happens when speed, surface and steering stop agreeing. It sits alongside DfT's own finding that travelling too fast for the conditions contributes to 18% of these collisions.
The moment itself
Mostly, just driving along
The manoeuvres a driving test spends most of its time on — junctions, turning, reversing — are not where these collisions concentrate. 59.3% were going ahead, more than the comparison group, and the largest single group was nowhere near a junction at all.
What the car was doing
Position relative to a junction
Who is hurt
A third of the people hurt were not driving
21,995 people were injured in these collisions in 2025, and 269 died. 7,546 of the injured — 34.3% — were passengers or people outside the car.
Fatalities are stated as recorded. They are the one severity that needs no statistical adjustment, and the only figures on this page that are directly comparable between places and years.
Cause
What DfT records as the reason
Police record contributory factors at the scene. That field is withheld from the open dataset, so these are DfT's own published figures, not our analysis — taken from their younger-driver factsheet for 2024–2025.
Shaded rows are the ones that are taught and practised rather than simply warned about — observation, reading conditions, and judging a junction. A collision can have more than one factor, so these do not sum to 100%.
Where
Every one of them, on a map
All 15,065 collisions, each at the location the Department for Transport publishes for it. Zoom in to your own area; click a point for what DfT recorded. The 233 in red are the fatal ones — they stay visible at every zoom.
- A group of collisions — the number is how many. Click to zoom in and separate them.
- One collision. Click for what was recorded.
- One fatal collision. Never grouped, visible at every zoom.
Map © OpenStreetMap contributors, tiles by OpenFreeMap. Collision locations and details: Department for Transport, Open Government Licence v3.0. Nothing about any person is recorded in this data.
Zoomed out, the points follow the road network and the cities, because that is where the driving is. Tick Fatal and they don't: the red scatters into the gaps between towns, which is the same finding as the rest of this report, arrived at from a different direction.
233 fatal collisions killed 269 people: some killed more than one.
A deliberate omission
Why there is no league table
Showing where collisions happened and ranking places against each other are different claims, and only the first one is supported here.
Police forces record injury severity under two different systems and moved over at different times. We measured it across the 43 forces with enough serious injuries to compare: applying DfT's adjustment moves one force's count by 38.0% and another's by -4.5% — a spread of 42.5 percentage points. A table ordering areas by serious injuries would substantially be ordering paperwork, so there isn't one. We do not name the forces either, for the same reason: the size of the distortion is the finding, not which force happened to carry it.
The numbers on the map's clusters are not a ranking either. They count every reported collision in a group, whatever its severity, and a place with more driving has more of them: a large number over a city says how busy its roads are, not how dangerous.
Fatalities are the exception. They need no adjustment, so they are directly comparable between places — which is why the map keeps them visible at every zoom rather than folding them into the clusters.
Methodology
How these figures were produced
-
1
A young driver is 17 to 24, by exact age.
DfT's own definition. The dataset also offers age bands, but they run 16–20 and 21–25 — using them returns 18% more records and disagrees with every figure DfT publishes.
-
2
The comparison group excludes mixed collisions.
A collision involving both a young and an older driver counts for the young group only. Counting it in both would compare each group partly against itself.
-
3
Missing values are excluded, never filled in.
16.7% of car records have no driver age recorded. They are dropped and counted, not assumed.
-
4
We reproduce DfT's own published total to within half a per cent.
Their factsheet gives 4,735 killed or seriously injured in these collisions; the same rebuild here gives 4,710. Using unadjusted severity instead gives 4,604 — measurably wrong, and the reason the adjustment matters.
Limitations
What this data cannot tell you
- Nothing about how likely anyone is to crash. STATS19 counts collisions, not miles driven. Without a mileage denominator, every figure here is a share of each group's own collisions — not a rate, and not a risk.
- Nothing about experience. There is no record anywhere in this data of when a driver passed their test. 17 to 24 includes people who passed last week and people who passed seven years ago. These are young drivers, not new ones.
- Nothing about collisions the police never saw. Only reported personal-injury collisions appear. Damage-only incidents and unreported injuries are absent and unmeasured.
- Serious-injury counts are not comparable between areas without the statistical adjustment described above. Fatalities are.
- Correlation is not cause. That young-driver collisions cluster on dark rural roads does not establish why. DfT's contributory factors are the closest thing to a recorded reason, and they are the attending officer's judgement.
Data sources
Every file this was built from
Published by the Department for Transport under the Open Government Licence v3.0. The checksum is of the exact file we read, so anyone can download it and confirm they have the same data.
collision — Road safety data: collisions
annual · DfT last updated 30 July 2026 · calendar year 2025 · 19,349,153 bytes
sha256 633249567e95479a1c30b3f10b0a6271ced11684fb338d0b8b4054ca94b80aa6
vehicle — Road safety data: vehicles
annual · DfT last updated 30 July 2026 · calendar year 2025 · 20,688,988 bytes
sha256 78b23539d0e7a1ccae9906c769b577c32f8b80ecc08d002e57abc6516b0431eb
casualty — Road safety data: casualties
annual · DfT last updated 30 July 2026 · calendar year 2025 · 10,062,743 bytes
sha256 cae4912c6cc29efcd7a20b6cbc7132e310478700dd06d80f46d2d92c86687c76
guide — Road safety open dataset data guide
annual · DfT last updated 30 July 2026 · code lists for all three tables · 81,378 bytes
sha256 5ee0d473e212ed50b2723ae52feeda86487861ed6a5a627b900f7b93c9bbf6b6
Open data: https://www.gov.uk/government/statistical-data-sets/road-safety-open-data. Contributory factors are quoted from DfT's younger driver factsheet. Licence: Open Government Licence v3.0. Built 2026-09-05T06:16:10Z.
About
About Driving Mastery
We published this because the pattern in it — dark, fast, rural, losing control — is the opposite of what most people picture when they imagine a young driver crashing, and because the skills it points at are teachable.
Using these figures? They are free to quote with credit to Driving Mastery and to the Department for Transport. If a number here looks wrong we would rather hear about it — the checksums above identify the exact files, and the method is described in full.
Analysis by Driving Mastery using Department for Transport STATS19 road casualty statistics, published under the Open Government Licence v3.0. DfT does not endorse this analysis.