Collision data for planning: getting it and using it well

Personal injury collision data is the strongest evidence available for a highway safety argument, and it is the evidence most often misused. A consultant who writes that five years of records show no collisions and therefore the access is safe has made a claim the data cannot support, and a competent highway officer will say so.

Used properly, the same dataset answers a narrower and far more useful question: does the recorded history show a pattern of the kind the objection predicts, and would the proposal add to it.

Where the data comes from and how to request it

Injury collisions attended or reported to the police are recorded on the national STATS19 return. Local highway authorities hold the records for their area, and the Department for Transport publishes the national road safety statistics as open data.

For a planning submission, request the records from the highway authority's road safety team rather than working from a public mapping tool. The authority's extract carries the full record set and can be cited.

Ask for, and state in the report:

• the study area, defined by a plan or by a list of links and junctions rather than a radius
• the period, with start and end dates
• the full record fields, including severity, vehicle manoeuvre, junction detail and control, light and weather conditions, and casualty class
• the date of extraction and the validation status of the most recent year

Allow time. Some authorities charge, some take several weeks, and the most recent year of data is usually provisional because validation runs about a year behind. Request early so that a delay does not sit on the critical path of a transport assessment.

Records are supplied anonymised. Do not reproduce anything that could identify an individual.

Choosing the study area and the period

Five years is the conventional period, and it is long enough to show a pattern without reaching back into a different highway layout.

Define the study area by what the proposal actually affects: the site frontage, the access itself, the junctions where development traffic joins the network, and the links between them. A circle drawn around the site captures collisions on roads your traffic will never use, which weakens the analysis rather than strengthening it.

Agree the extent in scoping wherever the scheme is large enough for a formal scoping exercise. An area agreed with the highway authority before the work starts cannot be criticised as self-serving afterwards.

Check whether the highway changed during the period. A new roundabout, a speed limit reduction, a signal upgrade or a junction realignment splits the data, and pre-change and post-change records should be reported separately.

Reading severity, causation and contributory factors

Severity is recorded as fatal, serious or slight. Two caveats travel with it.

Many police forces changed to injury-based severity reporting systems during the last decade, which altered the proportion of collisions classified as serious without any change on the road. Comparisons that straddle the change date are unreliable, and the report should say which years are affected.

Damage-only collisions are not recorded at all, and injury collisions are under-reported relative to hospital admissions data. So the record is a sample of incidents, not a census.

Contributory factors are recorded by the attending officer as an opinion formed at the scene. They are useful for spotting themes and they are not findings of fact. Write them as what they are: the attending officer recorded failed to look properly as a contributory factor in three of the seven records.

The fields that carry the most analytical weight are vehicle manoeuvre, junction detail and junction control, because those describe the conflict that occurred and can be compared directly with the conflict your access would introduce.

What the data can and cannot demonstrate

It cannot show that a location is safe. Absence of recorded injury collisions over five years at a low-flow rural access is close to statistically meaningless, and inspectors treat it that way.

It cannot establish causation. A collision recorded at a junction does not tell you the junction caused it.

It can show whether there is a cluster, and whether the authority's own criteria for a cluster site are met.

It can show whether the recorded conflicts match the type of conflict the objection predicts. If the objection is that vehicles emerging from the site will be struck by traffic travelling at speed, and the record contains three rear-end shunts in queuing traffic and one pedestrian incident at a crossing 200 metres away, the record does not support the objection.

And it can show whether comparable accesses on the same stretch of road feature in the record, which is often the most directly relevant comparison available.

Presenting a clean record without overclaiming

Structure the section so the conclusion is narrower than the data.

1. Source, period, extract date, study area and validation status, in that order, in two sentences.
2. A plan showing every record plotted, with severity distinguished, over the study area boundary.
3. A table by year and severity, and a second table by manoeuvre or conflict type.
4. A short commentary on any cluster, repeated manoeuvre or time-of-day pattern.
5. The caveats: under-reporting, the severity reporting change if relevant, and the provisional status of the most recent year.
6. A conclusion limited to what the data shows about the specific conflict at issue.

The wording that survives scrutiny is close to this: the records disclose no pattern of collisions involving vehicles turning into or out of accesses on this length of road, and the proposal would not introduce a conflict of a type represented in the record. That is defensible. A statement that the access is safe is not.

Illustrative example: a proposed access on a B road with an objection alleging a poor accident record at the adjacent junction. Five years of records show four collisions, three of them slight and all at the signalised junction 150 metres away, involving right-turning movements at the signals. None involves the site frontage. The technical note reports the four, distinguishes the location and manoeuvre, and concludes only that the records do not evidence a frontage safety problem. The figures are illustrative.

Using the data to answer a safety objection

Match the answer to the test in national policy. Development should be refused on highway grounds where there would be an unacceptable impact on highway safety, or where residual cumulative impacts on the network would be severe. Collision data speaks to the first of those.

Quote the objection's assertion in its own words, then set the data against it. Where the assertion is that the junction has a poor record, give the number, the severity split and the manoeuvres. Where the assertion is about speed, collision data is the wrong instrument and a speed survey is the right one, so say that rather than answering the wrong question.

Then offer mitigation that addresses any pattern the data does show. Improved visibility, a relocated access, a repositioned bus stop, a road marking change or a warning sign, secured by condition or by a highway agreement.

On smaller schemes the analysis usually sits inside a transport statement rather than a standalone note. Either way, keep the raw extract on file. Highway officers ask for it, and an analysis without the underlying records behind it is worth very little at appeal. If you are assembling a wider evidence base, browse the report types to see what else the authority is likely to expect.

This is general guidance, not legal advice. Check the position with your local planning authority or a qualified professional before relying on it. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval. Enter your postcode on PlanScope to see which transport documents your scheme is likely to need.