Requesting and using STATS19 collision data

Collision data is the one part of a transport chapter that cannot be modelled, estimated or argued into existence. It has to be requested, and it comes from the local highway authority's road safety team, drawn from the STATS19 records the police complete after a collision involving personal injury.

Ask for it early. Authorities vary in how quickly they respond, some charge a fee, and the data shapes the access design rather than merely describing it. A consultant who receives the records a fortnight before submission has lost the chance to do anything useful with them.

What the records contain and what they leave out

Each STATS19 record covers a personal injury collision reported to the police. The fields include the date and time, a grid reference for the location, the severity classified as fatal, serious or slight, the number and type of vehicles, junction detail, road surface and lighting conditions, casualty details in anonymised form, and contributory factors recorded by the attending officer.

The omissions matter as much as the contents. Damage-only collisions are not recorded. Collisions not reported to the police do not appear, and under-reporting is understood to be higher for cyclists and pedestrians than for vehicle occupants. Near misses are invisible. Contributory factors are the attending officer's assessment made at the scene, not a finding after investigation, so they indicate rather than prove.

One technical point that changes numbers. Police forces moved to injury-based severity reporting systems during the second half of the 2010s, and the change increased the proportion of injuries recorded as serious. Comparing a period before a force's change with a period after it will show an apparent rise in serious collisions that reflects the recording method. The Department for Transport publishes adjusted figures to deal with this, and any analysis spanning the transition should say which basis it uses.

Requesting data from the highway authority

The request goes to the road safety or traffic data team at the county council, unitary authority or, in London, the borough alongside Transport for London for its road network. For sites on or near the strategic road network, National Highways holds the relevant records.

A request that gets a quick, usable answer specifies:

1. The area, defined as a plan with a boundary rather than a description, or as a list of named junctions and link lengths with grid references.
2. The period, normally the most recent five full calendar years, stated by date.
3. The format required, ideally a spreadsheet with grid references rather than a PDF plot, so the records can be mapped and filtered.
4. Confirmation that anonymised personal injury records are sought, which avoids a data protection objection that would otherwise delay the response.
5. Any contributory factor data available, which some authorities supply as standard and others on request.

Keep the covering email. Where an authority later disputes the study area, the original scoping correspondence settles it.

Choosing the study area and the period

Five years is the convention rather than a rule, and the reason is statistical. Collisions are rare events, so a shorter period produces noise and a longer one includes network conditions that no longer exist. Where a junction has been altered during the period, split the analysis at the date of the change and say so.

The study area should cover the site access, the junctions the development traffic will use in meaningful numbers, and the links between them. Drawing it too narrowly invites a consultee to ask about the junction just outside the boundary. Drawing it too widely dilutes any local pattern into a regional average and makes the analysis harder to read.

Agree the area and period with the highway authority in writing at the scoping stage, as part of the wider scope for a transport assessment. It takes one email and removes an entire category of argument later.

Separating the collisions relevant to your access

Raw counts tell you almost nothing. The analysis that carries weight sorts the records into those bearing on the proposal and those that do not.

Work through the records individually rather than in aggregate. For each one, note the movement involved, the direction of travel, the vehicle types, the time of day and the recorded factors. Then group them: collisions at the junction the development will use, collisions on the link past the site frontage, collisions involving turning movements the development will add to, and collisions with no relationship to the proposal.

A rear-end shunt on the approach to a signal junction 300 metres away, caused by a driver looking at a phone, is not evidence about your access. Three collisions in five years involving right-turning vehicles at the junction where the development proposes a new right turn very much are. Say which is which, and do not bury the second category in a total.

Mapping helps. A plot with each collision numbered and keyed to a schedule lets the highway engineer check your reasoning in a few minutes, which is exactly what you want.

Writing the analysis honestly

The temptation is to present a clean record as proof of safety and a poor record as a pre-existing condition that the development will not worsen. Both arguments are weaker than they look, and experienced highway officers have heard each many times.

Better to write the analysis as a sequence: here are the records, here is what they show, here is which of them relate to the movements the development affects, here is what the development changes, and here is the mitigation proposed where a pattern exists. Where a pattern points at a design fault such as poor visibility at a bend or an unclear priority, saying so and proposing works is far more persuasive than explaining it away.

Where the development can address an existing problem through a section 278 agreement, the improvement becomes a benefit in the planning balance rather than a cost. That reframing is worth the analysis on its own. The approach to using collision records properly applies whether the record is clean or not.

Where a clean record does not help you

A five-year period with no recorded collisions at an access does not establish that the access is safe for a different level or type of use. It establishes that the existing use has not generated recorded personal injury collisions.

The distinction matters because the question before the decision maker concerns the proposed development. An agricultural access used twice a day may have a clean record and still be unsuitable for 40 dwellings, because the movements change in number, in timing and in the driver's familiarity with the junction. Visibility splays measured against the recorded speed of traffic are the relevant evidence there, not history. That is particularly true for an access onto a national speed limit road, where speed surveys rather than collision counts do the work.

National policy sets a high bar for refusal on highway grounds, requiring an unacceptable impact on highway safety or a severe residual cumulative impact on the network. Collision data contributes to that judgement without deciding it.

Using the data to answer an objection

Objections citing accidents are common and are often based on damage-only incidents, local recollection or events outside the study period. None of those appear in STATS19, which is a fact worth stating plainly rather than dismissively.

The response that works acknowledges the local knowledge, sets out what the formal records show for the period and area, explains what the records exclude, and addresses the substance of the concern through design. If residents describe near misses at a bend, the answer is a visibility assessment at that bend, not a sentence noting that no collisions are recorded.

Where an objection raises a specific incident, ask the highway authority whether it appears in the records and on what date. Sometimes it does and was missed. Sometimes it was damage-only. Either answer is better than a guess.

Transport assessments are normally expected to be prepared by a qualified transport planner, and a draft prepared with AI assistance is a starting point for that professional rather than a substitute.

PlanScope drafts transport statements and supporting documents structured to UK planning policy, with survey and collision data flagged for you to supply. Browse the report types and start a draft. Reports should be reviewed before submission, and we do not guarantee planning approval.