Retrieving and verifying the requested analytical release. Result data loads before optional map boundaries and source checks.
Your filters and release stay in the address. Navigation remains available.
ORIGINAL PHASE 3 · ENGLAND & WALES · 2022–2025
All 35,672 LSOAs, every year, including zero-collision areas. The E8 feature values and national bands remain fixed.
The panel contains 142,688 LSOA-year rows and 386,051 distinct, spatially assigned injury collisions involving a confirmed motor vehicle. 34,215 rows have zero such collisions. A collision is counted once, regardless of its number of vehicles or casualties.
These rows have positive modelled mileage and pass the declared reporting, geography and official-comparison gates. Remaining rows retain their observations and feature bands, with a reason for withholding the rate.
Authority-total reconciliation does not validate the distribution of miles between LSOAs. These are research inputs for baseline modelling and sensitivity tests; insurer-claims performance is untested.
The 2026 road network is retrospective for every study year. The 2025 outcomes have already been examined and are exploratory, not an untouched validation sample. STATS19 covers police-reported injury collisions, not all insurance claims.
| Country | Year | LSOAs | Motor collisions | Allocated miles, bn | Conditional rows |
|---|---|---|---|---|---|
| England | 2022 | 33,755 | 96,249 | 276.346 | 31,192 |
| Wales | 2022 | 1,917 | 3,262 | 18.153 | 1,916 |
| England | 2023 | 33,755 | 94,461 | 282.460 | 31,192 |
| Wales | 2023 | 1,917 | 3,197 | 18.499 | 1,916 |
| England | 2024 | 33,755 | 91,357 | 287.020 | 32,928 |
| Wales | 2024 | 1,917 | 2,878 | 18.753 | 1,916 |
| England | 2025 | 33,755 | 91,696 | 292.565 | 32,927 |
| Wales | 2025 | 1,917 | 2,951 | 19.185 | 1,916 |
All observed panel collisions are shown, including rows withheld from modelling. Mileage is allocated independently of collisions. Rates must use matching eligible numerators and denominators; national totals must not be averaged from LSOA rates.
Major-road AADF is converted to annual vehicle miles, with route, authority and distance checks. Failed routes retain an explicit road-number or road-class fallback. The remaining authority mileage is allocated to minor roads using fitted flow, shrunk group means and road length alone.
The selected traffic candidate is model, based only on geographic out-of-fold error at counted sites in 2022–23. Training uses 3,739 distinct sites across 421 20 km blocks. Separate 2024/2025 diagnostics distinguish new and repeated sites.
The selected model’s 2024 geographic holdout weighted absolute percentage error is 59.2%. Among 77 new sites, its mileage-flow bias is -20.3%. These diagnostics do not establish LSOA exposure accuracy.
Density, resident urban status and road attributes enter fitted exposure. Later tests of these features must compare the alternative denominators on identical rows. The scenario range is an allocation sensitivity range, not a statistical confidence interval.
Enter at least two characters. Up to 30 matches are shown; refine the name or enter the complete code. This filters the lookup and never changes feature bands.
Loading the national area index…
The joined modelling file repeats the exact frozen feature values, bands and statuses alongside the annual observations. The smaller panel omits those repeated feature columns.
| Download | Purpose | Size |
|---|---|---|
| Joined modelling panel | 142,688 rows with all frozen value/band/status columns | 82.17 MB |
| Annual collision–mileage panel | Counts, mileage alternatives, rate gates and quality flags | 22.01 MB |
| Panel field dictionary | Units, calculation roles and aggregation warnings | 0.01 MB |
| Analysis eligibility register | Every national row and its inclusion decision | 0.54 MB |
| Collision assignment ledger | GB source partition, E&W polygon and road-alignment diagnostics | 17.95 MB |
| Unmatched and unresolved collisions | No silent polygon fallback or motor reclassification | 0.11 MB |
| Authority mileage conservation | Every control area/year under all four named mileage columns | 0.42 MB |
| Official collision comparison | All-collision source totals against RAS0403 | 0.03 MB |
| Regional road-class comparison | Independent publication totals and AADF discrepancies | 0.01 MB |
| Count-site validation | Spatial folds, new sites and repeated sites | 0.04 MB |
| Major section audit | Route gates, failures and Counted/Estimated source flags | 3.48 MB |
| Minor network support | Traffic training coverage by region, class and function | 0.00 MB |
| Minor fragment predictions | Inspect flow predictions and support flags | 50.63 MB |
| Road identifier crosswalk | Stable OS identifiers for the internal road-link references | 67.44 MB |
| Compact reproduction kit | Offline panel reconstruction with exact output checksums | 46.70 MB |
ChecksumsSource registerFrozen assembly designReproduction instructionsExecuted checks
Open the implemented Phase 4 baseline: collision-frequency models with log(miles / 100,000) as the exposure offset. The results compare exposure scenarios and assess spatial and temporal performance, with calibration limits documented before testing composite indices.
Browser interaction checks were unavailable in this build environment. Data, source preservation, application build and download checks are documented separately.