The article on Crash Locations (link here) provides information on where you can download official, record-level data. Be aware that certain data fields will not be included if they are personal in nature (vehicle registrations), or sensitive (contributory factors). For many people, simply accessing summary data using the customisable DfT data download tool is sufficient. The DfT download tool can be accessed here.

Local authorities and police forces will often release this data on request subject to local polices on disclosure, GDPR, FoI and other considerations. This may include other information about a collision that does not form part of the official statistics including a narrative description of the collision.

If you are a data scientist you can use the ‘STATS-19’ R package for working with open road traffic casualty data from Great Britain, available at this link to the Comprehensive R Archive Network – CRAN: STATS19 package.

There is an excellent, short report on media reporting guidelines for road collisions which has been informed by a number of NGOs and professional organisations. It clearly sets out the type of language that should be used and why the term ‘accident’ should be avoided. Use of the term ‘crash’ or ‘collision’ is often left to personal preference. A link to the reporting guidelines is available here.

The central dataset uses information collected by police officers attending the scene of a collision and does not reflect any in-depth investigation that may subsequently take place. This means that any emerging evidence will not be reflected in the official statistics. The instructions for completing collision record forms (known as STATS20) are complex with police officers requiring training before undertaking the task. Take a look at the STATS20 instructions here.

Collision records do not seek to apportion blame or fault, although contributory factors may be recorded.

Simply counting the number of collisions along a road or in a specific area then sorting from highest to lowest is not an analysis of relative risk or frequency. Consideration must be given to length of road or the level of traffic for example; a road with double the number of collisions of another, but that is twice as long, or has twice the traffic will have exactly that same risk rate. The same applies when comparing areas that are very different in terms of population or road length.

Fortunately there are well-established methods for carrying out this analysis and there are published results that seek to answer some of the more common questions. 

The Road Safety Foundation analyse all major roads in Great Britain each year and produce an analysis of those that are ‘significantly improved’ or remain ‘high risk’.  The EuroRAP risk assessment dashboard contains their results and can be viewed here.

In 2021 a GB Road Safety Index was produced comparing the performance of local authorities and police forces across Great Britain. (To see the Index click here). This analysis allowed for complex reporting changes introduced in 2018 and looked at long-term trends to assess which areas had the most significant changes in collision numbers, as well as comparing rates. A guide to the severity adjustments for reported road casualties Great Britain is available here.

The Ministry of Justice releases a spreadsheet each year including summary data of all motoring offences. This was most recently updated in May 2022 for the year 2021 but only includes data by police force, not a local authority area or a constituency, unless the boundaries match exactly. Click here to open the latest spreadsheet on motoring offences.

Many road safety partnerships and police forces will publish their own data, often breaking down offences for individual speed cameras for example.

You can also review contributory factor information for collisions via the annual DfT publications. These will often link to a specific offence e.g. ‘Exceeding the speed limit’. Contributory factor information can be viewed here.