Crime
Every small area of England and Wales shows how much crime the police record there, and whether it is rising or falling. Here is where the figures come from, how they were tested and what they cannot tell you.
How much crime
Small areas are drawn to hold about the same number of people, so a plain count compares fairly. Where the numbers are small it is steadied with similar places and the neighbours, so one bad year does not make a hotspot. Town centres are marked. Areas a force has not published are left blank.
Rising or falling
Crime trend shows how fast crime has moved in each area over the last few years. Small areas are pulled toward their council's figure, so chance is not read as a trend. Red is rising faster than the council. Blue is slower, or falling.
How it was tested
Each run predicts the past as it would have then, and is scored on what happened next. We also built a forecast for the next three years. It did no better than this year's figure, so the map does not show it. Use the map to plan prevention and ask questions, alongside what officers know.
Crime is a group in the Network panel, for authorities and police forces with an account. It has two layers. Each small area is drawn as its own shape. A small area is a census LSOA, home to 1,000 to 3,000 people. Only one layer is on at a time.
Pick the type of crime on the layer's options page (the › on its tile): all crime, which leaves out anti-social behaviour as it is not a crime, or one of the police's own groups, from burglary to shoplifting. Tap an area for its figures. A street's card shows the same for the small area it is in.
The map's area check (a circle round any point) shows crime too. There it counts what the police recorded, not our estimates. Each neighbourhood the circle covers counts by how much of its ground is inside. ‘For its size’ compares the circle's crimes per square km with every neighbourhood in the council, and anti-social behaviour is counted on its own.
Small areas are drawn to hold about the same number of people, so the number of crimes compares fairly from one area to the next. Per 1,000 residents is there too. It makes a town centre look extreme, because few people live there and many pass through. The count says plainly that a lot happens there.
A small area's count jumps about. Twelve after five is usually luck. So the figure leans on three things.
How much weight each part gets is not guessed. Each run tries the settings on past years and keeps the ones that predicted the following year best.
Town centres have few residents and many visitors, so a figure per resident overstates them. They are marked on the card. Some forces' data is missing from police.uk. Greater Manchester Police has published almost nothing for years, so about 1,700 areas there are left blank rather than shown as safe. Tap a street there and its card says why.
Each area's monthly counts give a trend, allowing for the time of year. A small area's trend is mostly noise, so it is pulled toward its council's. Against its council is the area's change a year minus the council's. The gaps are small: nine areas in ten are within about 1.5% a year of their council. So the colours go from the slowest areas to the fastest. Pale means it could be chance. The council is the local authority the area sits in.
We built a forecast for the next one, two and three years. We tested it on the last three years, using only what was known at the time. For all crime and seven of the types it came out the same as this year's figure. For the other seven it found next year's busiest areas no better than this year's figure did. So the map does not show it. If a later version does better, it will come back.
Each run repeats all of this as it would have been one, two and three years ago, using only what was known then, and scores it on what happened next. It reports three checks.
The run on 13 October 2026: the 5% of areas ranked highest held about a quarter of the next year's crime (chance is 5%). For all crime, last year's count did as well. The steadying helps most for the rarer types, such as weapons and bicycle theft. The range held the real figure about nine times in ten. Areas rising faster than their council kept doing so the next year about half the time (49% for all crime). So read Crime trend as what has happened, not what comes next.
One more limit. The figures begin in September 2021, while recorded crime was still recovering from the lockdowns in several categories. A rise in the early years can be that recovery and not a trend.
Crime is concentrated. A small number of places carry much of it, year after year. Researchers call this the law of crime concentration. It is why an area's level is the best guide to its next year, and why the trend is shown beside it as what has happened, not as a forecast.
For a traffic engineer, a rising area is a prompt to look at lighting, visibility, parking, walking and cycling routes and bus stops there. For a community safety team, it is a prompt to look at prevention. For policing or enforcement, a data protection impact assessment and an equality impact assessment come first, and a person decides anything that affects a person.
police.uk publishes each month about six weeks late. Every new month re-runs all of this, including the tests.