How the figures work

The headline rate

The rate shown on every area page is recorded crimes per 1,000 residents over a rolling 12 months. It is the total number of offences the police recorded in that area over the most recent twelve months of published data, divided by the area's resident population, multiplied by 1,000.

The rolling window is used rather than a calendar year because police data is published monthly with a lag of roughly two months. A calendar-year figure would be up to fourteen months out of date for most of the year.

How crimes are assigned to an area

Each crime record carries the Lower Layer Super Output Area (LSOA) it occurred in. LSOAs are small statistical areas of about 1,500 people. Every figure above LSOA level is built by summing LSOAs upward: LSOA to neighbourhood, neighbourhood to local authority district, district to county, county to region, region to country. A county total is therefore always the sum of its districts, and can never disagree with them.

Older data uses the 2011 LSOA boundaries and current data uses the 2021 ones. Where a code was retired, it is mapped onto its 2021 successor using the ONS lookup. Where a 2011 LSOA was split into several 2021 LSOAs, the mapping picks one of them: that is exact at district level and above, because splits never cross a district boundary, and approximate at neighbourhood level for those few areas.

The population denominator

Population comes from ONS mid-year population estimates at LSOA level, and is summed up the same hierarchy as the crimes, so the numerator and the denominator always describe the same set of small areas. The estimate year is shown on each page.

A resident population denominator systematically flatters residential suburbs and penalises town centres. A city centre with 8,000 residents may host 200,000 people on a Saturday, and every crime among them counts against those 8,000 residents. This is the standard measure used across the sector, and it is the right one for comparing similar places, but it is not a measure of personal risk.

What "recorded crime" means

These are offences the police recorded. They are not all offences committed. A crime that is never reported is never recorded, and reporting rates differ enormously by offence type: vehicle theft is nearly always reported because of insurance, while much anti-social behaviour and low-value theft is not.

A rise in recorded crime can therefore mean more offending, better reporting, a change in recording practice, or a police campaign that went looking for a particular offence. This site reports the direction and size of a change and does not tell you which of those it was.

Locations are approximate by design

Police open data does not publish exact crime locations. Each record is snapped to the nearest point on a published list of anonymised map points, chosen so that no point represents fewer than about eight addresses. A crime shown "on or near" a street may have happened on a neighbouring one. The list of snap points was itself revised in 2022, which moved some crimes between nearby points without anything having changed on the ground.

Neighbourhoods, and why they are not wards

City pages break the city into neighbourhoods. Those neighbourhoods are Middle Layer Super Output Areas, statistical areas of around 7,000 residents, and they are named using the House of Commons Library's MSOA names, which exist precisely so that statistical areas can be talked about in the names people use.

They are not electoral wards, and that is deliberate. Wards are named after churches, manors and old parishes rather than after the places people live. Lincoln's wards are Abbey, Carholme, Castle, Glebe and Minster; neither Ermine nor St Giles appears anywhere in the list, though both are places any resident could point to. Ward boundaries also move at every electoral review, which quietly breaks a time series, while these areas are fixed for a census decade.

Neighbourhoods do not get their own pages. Their figures appear on the page for the town or city they belong to, which is where the comparison between them is worth making.

The neighbourhood map and its colours

The map on a city page shades each neighbourhood by recorded crimes per 1,000 residents. Two choices there are worth stating, because both are places where a map can mislead without saying anything false.

The scale is fixed, not relative. Each neighbourhood is coloured by its rate as a multiple of the England and Wales average: under half, below, about average, above, or more than double. It is not split into five equal groups of what is on screen. Equal groups would shade a fifth of every city in the darkest colour no matter how much crime it actually records, which makes a quiet town look as alarming as a busy one. With a fixed scale, the same colour means the same thing in Guildford as it does in Middlesbrough.

The colours run teal to rust, not green to red. A green-to-red map tells the reader which places are good and which are bad, which is a judgement this site does not make. It is also the one colour pairing that disappears for the roughly one man in twelve with red-green colour blindness.

These areas average around 7,000 residents. Each figure is an average across a whole neighbourhood and says nothing about any particular street inside it.

Streets, kinds of place, and why neither is a worst-streets list

The police publish a location for each offence in one of two forms. Most carry a street name. The rest carry a kind of place: an offence at a supermarket is published as "Supermarket", with no address, because the snap point stands for the shop rather than the road it sits on. There are twenty of these categories, from Supermarket and Parking Area to Ferry Terminal and Race Track.

City pages keep the two apart. Kinds of place are totalled across the whole town, since a list of individual points would read as "Supermarket, Supermarket, Supermarket" with nothing to separate them. Streets are listed individually, so every entry names somewhere a reader can place.

Neither list ranks danger. The locations are not addresses: every crime is snapped to the nearest point on a fixed published list, so one point collects everything recorded across a cluster of addresses, and an offence shown on one street may have happened on the next. The busiest points sit where the people are, which is why stations, shopping parades and nightlife streets head the list in almost every town.

There is also no denominator. A location has no resident population, so no rate can be calculated for it, only a count, and a count measures how busy a place is at least as much as how much crime happens there.

Known gaps

• Missing months. Individual forces miss a monthly return regularly. Where the force covering an area has not published up to the current month, that area's page says so at the top, with the number of missing months. A gap is never shown as a fall in crime.
• Greater Manchester. Greater Manchester Police has not supplied street-level data since July 2019, following the failure of its new record management system. Pages for Manchester and the rest of Greater Manchester carry that warning.
• Scotland. Police Scotland does not publish to data.police.uk, so Scotland is not covered by this data at all. Scottish figures come from a different source on a different basis and are not shown here rather than being silently mixed in.
• Northern Ireland. Police Service of Northern Ireland records carry no LSOA code, so Northern Ireland crimes can only be counted force-wide, not by town or neighbourhood.
• Revisions. Months inside the current three-year window are revised between releases. Each monthly update reloads the whole window rather than only the newest month, so a revised figure here always matches the current published figure.

Which areas get a page

An area only gets its own page if it has enough data to fill one: at least 100 recorded crimes in the latest twelve months and at least 1,000 residents. Everything below that threshold is covered on its parent area's page instead.

Who is in a ranking, and who is not

The ranked lists, the rank quoted on each area's page, and the head-to-head comparisons all cover the same set: every area with a page whose police force filed all twelve of the months being counted. Nothing else is in them.

That last condition does real work. A force that has missed five monthly returns produces a twelve-month total short by five months, and a short total is indistinguishable from a fall in crime. At the time of writing Gloucestershire Constabulary has filed nothing since January 2026; ranked as published, four of its six districts took the first four places in "the safest places to live" on a 45% drop that was five missing returns and nothing else. They are named on the ranking pages, with the reason and the last month their force filed, rather than dropped without explanation. Their own pages still carry their figures, with the shortfall stated at the top.

Greater Manchester's districts are excluded from rankings on the same principle, and more severely: their force has filed nothing since 2019, so there is no rate to rank.

Offence, year and comparison pages

Each area also has a page per offence type, a page per calendar year back to 2011, and a head-to-head page against other areas. All of them are built from the same monthly counts as the area page, on the same rules, and each says what it is doing differently:

• Offence pages use the Home Office categories exactly as forces publish them. They are comparable between areas and across months, and not comparable with court figures or with any other classification.
• Year pages report the counts exactly as recorded, and say how many of the twelve monthly returns the force actually filed. The rate per 1,000 on a year page uses the current population estimate rather than that year's, so it compares places rather than reconstructing a historical rate; the counts do not depend on it.
• Comparison pages plot a rolling twelve-month rate rather than monthly counts, so two places of different sizes sit on the same scale. A line breaks wherever the twelve months behind it contain an unfiled return.

Every area page offers its whole monthly series as a CSV download, one row per month per offence, with a column saying whether the force filed that month. Reuse it under the Open Government Licence.

What is here besides crime

Four datasets sit alongside the crime figures. None of them is crime data, and each is here because a crime rate on its own invites a conclusion it cannot support.

• Workday population (ONS Census 2021). Every rate here divides crime by residents, and for a district that is mostly town centre that is the wrong denominator: the City of London has around 15,000 residents and a workday population of 92,154, so crime against the people who are actually there is divided by a sixth of them. The resident rate stays the headline and the rankings stay ordered by it, because that is what the rest of the field publishes; the workday rate is shown beside it. Census day fell in March 2021 under Covid restrictions with widespread home working, so central London workday counts are lower than a normal year would show.
• Deprivation (English Indices of Deprivation 2025). The strongest known correlate of recorded crime, so an area's crime rank set against its deprivation rank says far more than the crime rank alone. Blackpool is 1st most deprived and 4th for crime; Westminster is 95th most deprived and 2nd for crime. Those are two different stories that the crime rate alone tells identically. England only, because Wales publishes a different index on a different basis. A correlation is not a cause, and nothing here claims deprivation produces crime.
  
  One caveat that is easy to miss and rarely stated: the index is not independent of crime. Recorded crime is one of its seven domains, weighted at 9.3% of the score alongside income and employment at 22.5% each. So an area with a high crime rate scores as slightly more deprived partly because of that crime, and any comparison of crime against deprivation, including the predicted rate shown on area pages, is to that extent circular. The effect is bounded by that 9.3% and it is not nothing. It is also why somewhere as wealthy as the City of London is not the least deprived place in England: it does very well on income and employment, and badly on the crime and living environment domains, which between them cover air quality, road casualties, housing condition and recorded offences.
• Police officer numbers (Home Office police workforce return). Recorded crime partly measures how much policing there is to record it, and a force that lost a quarter of its officers and got them back is not the same force across the series.
• Stop and search (data.police.uk, published alongside the crime files). Force level, because the records carry coordinates but no area code, and that is also the level at which it is officially analysed. The figure worth reading is the share of searches that found something: a force searching a great many people and finding almost nothing is describing its own practice rather than its area. These are not broken down by ethnicity here, because doing that responsibly needs the population being measured against, and a composition without a denominator looks like a finding and is not one.

Greater Manchester, and the second source

Greater Manchester Police stopped supplying street-level data in 2019, so for its ten districts the extracts everything else here is built on contain almost nothing. Those pages used to say only that there were no figures. The figures existed the whole time, in a different Home Office release, and they now appear.

That release is police recorded crime at Community Safety Partnership level, which is a local authority in almost every case. It is a genuinely different measurement and the pages built on it say so at the top:

• counted to the Home Office Counting Rules, rather than being whatever a force uploaded;
• quarterly, by financial year, rather than monthly;
• nine offence groups rather than the fourteen categories used elsewhere, and they do not map onto each other: "Theft offences" spans burglary, vehicle crime, shoplifting and bicycle theft, which are four separate categories in the street-level data;
• located to the district and no finer, so those pages carry no map, no neighbourhood table and no street data.

The two sources are never added together or drawn on one axis. They answer the same question by different rules and would disagree even if both were complete, so an area served from this source is compared only against an England and Wales figure built the same way, and is left out of every ranking and head-to-head on the site. Which areas fall back to it is worked out from the data on each refresh rather than kept as a list, so a force that resumes filing stops being an exception without anyone editing anything.

Who oversees policing, and why it is not the council

Every area page names the office accountable for policing there. It is worth being precise about this, because the common assumption is wrong: councils in England and Wales have no policing powers. A police force answers to a directly elected Police and Crime Commissioner, or to the combined authority mayor where that office has absorbed the role, as in Greater Manchester, London, West Yorkshire, South Yorkshire, and York and North Yorkshire. Three forces have no elected commissioner at all: the City of London Police answers to the City of London Corporation, the British Transport Police to an appointed authority, and the Police Service of Northern Ireland to the Northern Ireland Policing Board.

What is recorded here is election results, not terms of office. People resign, die and are disqualified mid-term, so a page saying someone "is" the commissioner goes wrong the day that happens and nobody notices. Saying that a named person won a named election cannot go stale. Wiltshire's May 2021 winner was disqualified before taking office; Norfolk's 2024 winner was replaced at a by-election in July 2026. Both are listed as what they are. For who holds an office today, the force's own website is the authority.

These elections are marked on the long trend chart so the term boundaries can be seen against the line. They are context, not an explanation. Recorded crime moves with how offences are counted, how a force records what is reported to it, national policy and a pandemic, and those effects are far larger than anything one commissioner does in a four-year term. With 37 commissioner areas and four elections there is nothing here that could support a claim about a party either way, and this site does not make one. If you see a rise or a fall either side of a marker, the honest reading is that you are looking at two facts printed on the same axis.

Sources and licence

• Recorded crime: data.police.uk, Home Office, Open Government Licence v3.0.
• Population estimates: Office for National Statistics, mid-year small area estimates, Open Government Licence v3.0.
• Area boundaries and lookups: ONS Open Geography Portal. Contains OS data © Crown copyright and database right.
• Workday population: ONS, Census 2021 table WD013, Open Government Licence v3.0.
• Deprivation: Ministry of Housing, Communities and Local Government, English Indices of Deprivation 2025, Open Government Licence v3.0.
• Police officer numbers: Home Office, Police workforce open data tables, Open Government Licence v3.0.
• Stop and search: Home Office, data.police.uk, Open Government Licence v3.0.
• Greater Manchester and the second source: Home Office, "Police recorded crime and outcomes open data tables", Community Safety Partnership level, Open Government Licence v3.0.
• Police and crime commissioner, mayoral and by-election results: the declared results of the 2012, 2016, 2021 and 2024 elections in England and Wales, together with the London and Greater Manchester mayoral elections and the Northumbria (2019), Wiltshire (2021) and Norfolk (2026) by-elections.

You are free to reuse anything on this site under the same Open Government Licence, with attribution. If you spot an error, the About page says how to reach us.