STATS19 Data Lab

Police-reported injury collisions in Great Britain · geographic history 2016–2025 · record analysis 2021–2025

Methods and limitations

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.

Evidence views, exports and Actuary Lens interpretations read the same immutable release. Research protocols and prepared research panels are versioned separately and identify their evidence 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 input
06a03c0e5504175e23ed98e3e0b4767ac61d396c073405a1bddab9339c18f948
Original 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 input
05453dca2651ad8a27b4ac0e6a0450a852c506e4d495111ffca398f3d311e26a
Original 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 input
870be5a934ae10baf5c52163d59eccba0ab2f40327ffa6b2bd354d2e7ee6b108
Original 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 input
51b08606ef3dad1ae5d4c8726241b022fc8e4bbdc910ef0e5cc2fb92b054fc12
Original 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 ·

Clock changes and road collisions · Protocol, data and description · Phases 1–3 →Parking and collision risk · Adjusted analysis · Phase 4 →Frost days and collision frequency · Persistence and validation · Phase 8 →Social factors · Learned adjacent bands · v3 →National feature studies · Actuarial joint GLM review →

Exports Sources

All investigations

Search investigationsWorkspace

22 investigations · study pages identify their own frozen release.

Frequency per 100,000 milesCompare road, time and vehicle rates with their exposure basis.Descriptive result available · Derived rates describe reported injury outcomes, not insurance claim frequency.
Annual trendsFollow pooled rates for the same eligible authorities.Descriptive result available · A fixed authority panel supports descriptive annual comparisons.
Road, time and vehiclesExplore counts, injury shares and supported mileage rates.Descriptive result available · Some breakdowns have counts only; rates require a compatible denominator.
Road mix and geographic persistenceCompare raw and road-mix-adjusted geography, then later years.Comparison unavailable · This evidence release does not contain the required inputs. Choose another investigation or open its published release.
Map, authority and historyInspect two authorities and a fixed-panel benchmark.Descriptive result available · Collision location and through-area mileage describe places, not residents.
PersistenceFreeze discovery groups and inspect later separation.Exploratory analysis · Retrospective persistence is not evidence of insurer pricing lift.
Residence and collisionInspect recorded home-area matches and distance bands.Descriptive result available · Recorded involvements are the denominator; this is not a resident risk rate.
Vehicle mileage and fuel exposureSeparate licensed stock, observed MOT mileage and usable exposure.Coverage research · Fuel-specific rates depend on compatible mileage; stock is not exposure.
Vehicle cohortsInspect propulsion cohorts and stock-normalised indicators.Descriptive result available · Licensed-stock indicators are not frequency per mile.
Comparable EV / ICECompare conditional injury severity on common support.Exploratory analysis · Unsupported selections remain unavailable; no causal fuel effect is estimated.
ADAS and vehicle generationsAudit equipment mappings and generation coverage.Comparison unavailable · This evidence release does not contain the required inputs. Choose another investigation or open its published release.
Driver age and journey purposeInspect survey mileage and compatible recorded subsets.Comparison unavailable · This evidence release does not contain the required inputs. Choose another investigation or open its published release.
Injury burdenCompare casualty counts and injury severity by recorded group.Descriptive result available · Injury burden is not monetary claim severity.
Local rainfall and measured trafficInspect the M4 winter pilot, traffic coverage and sensitivity.Comparison unavailable · This evidence release does not contain the required inputs. Choose another investigation or open its published release.
Annual country weatherCompare matched country-year observations.Exploratory analysis · Country-year associations do not identify local or causal weather effects.
Clock changes and road collisionsCompare autumn and spring development-year counts, severity and young-driver involvement, with a frozen protocol and auditable hourly panel.Descriptive result available · Descriptive evidence · external covariates pending
London baseline and qualificationsReproduce the frozen London baseline and no-qualifications study with all model inputs, independent calculation checks, fresh bootstrap refits and five mileage scenarios.Exploratory analysis · Reproduced calculations · exploratory predictive evidence
Social factorsFollow the original matched authority study through exploratory models, robustness checks and learned adjacent decile bands. The national joint GLM is a separate study with a different population.Exploratory analysis · Exploratory adjacent-band analysis
Parking and collision riskCompare raw and adjusted rates across 137 English authorities, with rurality, traffic, housing and socioeconomic controls and whole-region prediction checks.Exploratory analysis · Exploratory adjusted analysis
Frost days and collision frequencyInspect frozen 2025 authority predictions, geographic validation and annual tail comparisons alongside the separate conditional LSOA mileage pilot.Exploratory analysis · Exploratory persistence analysis
National LSOA · current joint GLMReview all 40 nationally banded features together and ten separate raw-value extensions. Inspect calibration, conditional gain intervals and mileage sensitivity before recording feature decisions.Exploratory analysis · Exploratory joint GLM review
Social factors · historical panelInspect the frozen preparation panel and research protocol.Coverage research · Historical preparation; use the current social-factor study for modelling.

Comparison unavailable

Can documented equipment be linked to collision risk?

This evidence release does not contain the required inputs. Choose another investigation or open its published release.

Charts displayed; exact values remain available with each chart.

ADAS generation research is unavailable in this release

This release predates the reviewed equipment dictionary and mapping audit.

Current selection

Full data & reproducibility

Geographic inputs & release manifest Detailed STATS19 data & manifest Read the actuarial interpretation →

Sources & release

Loading source monitoring. Selected results are already available; live checks do not change them.

Source	Evidence period	Source published	Checked in release	Latest check	Monitoring state
RAS0403: collisions and casualties by local authority	2016–2025	Date unavailable	23 Sept 2026	21 Oct 2026	No source issue recorded
Local authority motor vehicle mileage	2016–2025	Date unavailable	23 Sept 2026	21 Oct 2026	No source issue recorded
Rainfall — England	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Rainfall — Wales	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Rainfall — Scotland	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Air frost — England	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Air frost — Wales	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Air frost — Scotland	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Wet days ≥1 mm — England	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Wet days ≥1 mm — Wales	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
Wet days ≥1 mm — Scotland	2016–2025	Date unavailable	23 Sept 2026	23 Sept 2026	Source content changed; review pending
STATS19 collision records, latest five final years	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
STATS19 vehicle records, latest five final years	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
STATS19 casualty records, latest five final years	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
STATS19 codebook and historical mappings	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
Regional vehicle mileage	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
Regional road-category mileage	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked
VEH0120 GB licensed vehicle stock	2021–2025	Date unavailable	25 Sept 2026	Date unavailable	No live check linked

Loading approved evidence

ADAS generation research is unavailable in this release

Sources & release

Sources & release