Skip to main content

UK Crime Data: Recorded Crime vs the Crime Survey

Crime is measured two entirely different ways in England and Wales, and picking the wrong one is the single most common statistical error on this beat. This guide covers choosing the right source, calculating rates, reading outcome data, the data.police.uk mapping trap, and spotting recording changes that manufacture trends.

Last reviewed: Next review due:

Method, not numbers: this guide deliberately avoids quoting current crime figures, because they are superseded quarterly. Every claim here is about how the data is built. Take the figures themselves from the live release you are citing, and link to it.

The two-source problem

England and Wales measure crime twice, using methods that answer different questions and routinely move in opposite directions. Police recorded crime counts notifiable offences that someone reported and a force then recorded under the Home Office Counting Rules. The Crime Survey for England and Wales (CSEW) asks a large sample of household residents what has actually happened to them, regardless of whether they told anyone.

The ONS user guide puts the difference plainly: the CSEW “provides a better reflection of the true extent of crime experienced by the population resident in households in England and Wales than police recorded statistics, because the survey includes crimes that are not reported to, or recorded by, the police”, and it is “unaffected by changes in levels of reporting to the police or police recording practices”. The current ONS bulletin goes further, warning that police recorded crime “is not considered a reliable indicator of overall crime trends because of changes in police activity and recording practices”.

That is not a reason to ignore recorded crime. It covers offences and populations the survey cannot reach, and it is the only source that goes down to your town. It is a reason never to write “crime rose 12%” from a recorded-crime table without asking whether the recording changed.

Choosing the right source for the question

Is crime going up or down over the long run?
Crime Survey for England and Wales

The ONS bulletin states the CSEW is better for tracking long-term trends in violence with or without injury, theft offences, criminal damage, fraud and computer misuse. It is not moved by changes in how willing the public are to report or how forces choose to record.

How many homicides, robberies or knife offences were there?
Police recorded crime

The same bulletin says police data are preferred for homicide, offences involving weapons, and robbery. These are either too rare for a household survey to measure or, in the case of homicide, impossible to ask a victim about.

What happened in one town, ward or force area?
Police recorded crime (with caveats)

The CSEW cannot go there reliably: its target is around 31,000 interviews a year with a minimum target of 585 per police force area. Only recorded crime is published at Community Safety Partnership and street level — but you inherit every recording-practice problem with it.

How much crime never reaches the police at all?
Crime Survey for England and Wales

The gap between survey-estimated and recorded crime is the reporting rate, and it is one of the few numbers only a victimisation survey can give you. It varies enormously by offence type.

What happened to the cases the police did record?
Home Office crime outcomes data

Outcomes are published separately from recorded crime, by offence and force, in the Home Office open data tables. This is where charge, caution, and closed-without-suspect figures live.

Anything in Scotland
Recorded Crime in Scotland / Scottish Crime and Justice Survey

Scotland is a separate system with its own recording standard and its own survey. Neither the ONS crime bulletin nor data.police.uk covers it.

What the Crime Survey cannot see

The CSEW is the better measure of the crime it covers — theft, robbery, criminal damage, violence with and without injury, and since January 2017 fraud and computer misuse. But its scope is narrower than most people assume, and the gaps are where reporters get caught.

Homicide
A victimisation survey cannot interview a homicide victim. Homicide is measured through the Home Office Homicide Index, whose statistics are accredited official statistics, independently reviewed by OSR in 2016.
Crime against businesses and the public sector
The CSEW covers crime against people resident in households. Shoplifting, business burglary, and fraud against companies are outside its scope entirely — a common source of confusion in retail-crime stories.
"Victimless" offences
The ONS user guide is explicit: "The CSEW excludes those crimes often termed as ‘victimless’ (for example, possession of drugs)." Drug offence counts are almost pure measures of police activity.
People not living in households
"The CSEW does not cover the population living in group residences (for example, care homes or student halls of residence)" — nor prisons, hostels, or people sleeping rough.
Children, since April 2025
A survey of 10 to 15 year olds ran alongside the main CSEW from 2010 until March 2025, when it was discontinued during development of an online survey. Check what the current release actually covers before quoting child victimisation.
Two missing pandemic years
Standard CSEW datasets are not available for the years ending March 2021 and March 2022 because face-to-face fieldwork was suspended; a telephone survey ran instead. Any chart that draws a straight line through those years is smoothing over a methodological break.

One further caveat matters for sensitive beats: ONS notes that sexual violence and domestic violence reported in the interviewer-led parts of the CSEW are prone to considerable under-reporting, which is why those topics are asked through separate self-completion modules. Quote the module, not the face-to-face headline.

The accreditation status nobody mentions

In 2014 the Office for Statistics Regulation removed the National Statistics accreditation from police recorded crime for England and Wales, after an assessment found evidence that the quality and consistency of the underlying data might not be reliable. More than a decade later that designation has not been restored: the ONS bulletin still labels police recorded crime as official statistics that are not accredited official statistics.

OSR published a systemic review of the quality of those statistics on 16 May 2024. It concluded that police forces are recording crime more accurately now than in 2014 — national recording accuracy having risen from 80.5% to 92.4% — and set out what would need to happen before reaccreditation could be reconsidered; fraud and computer misuse were excluded and will be looked at separately. Other parts of the same series fared differently: CSEW estimates regained accredited official statistics status in October 2024 after a data quality review, the CSEW fraud and computer misuse questions were accredited in March 2018, and statistics based on the Home Office Homicide Index have been accredited since an OSR review in 2016. A single ONS crime release can therefore contain accredited and non-accredited statistics side by side. If you are asserting authority for a number, know which one you have.

data.police.uk: what it is, and the snap-point trap

data.police.uk publishes monthly, record-level open data under the Open Government Licence v3.0: crime and anti-social behaviour incidents with subsequent police and court outcomes, and a separate stop and search dataset with date, time, street-level location, ethnicity, gender, age and outcome. The crime data is supplied by the 43 geographic forces of England and Wales, the British Transport Police, the Police Service of Northern Ireland and the Ministry of Justice, with geographic coverage of England, Wales and Northern Ireland. It is available as monthly CSV archives, a custom CSV builder, and a JSON API.

The trap: those are not the locations of the crimes

data.police.uk maintains a master list of anonymous map points. Each point sits over the centre of a street, or over a public place such as a park or airport, or a commercial premise such as a shopping centre or nightclub — and each must have a catchment area containing at least eight postal addresses, or no postal addresses at all. Points with between one and seven addresses were deliberately discarded to protect privacy.

Every crime’s real coordinates are compared against that list, replaced with the nearest point’s coordinates, and discarded. If the nearest point is more than 20km away, the coordinates are zeroed out. So a dense cluster of dots means a snap point is there — near a station, a nightclub, a supermarket car park — and tells you nothing about which building or which household. The list was built in 2012 from Ordnance Survey road centre points and points of interest, and refreshed in 2022, with over 679,000 street points surviving the privacy filter.

Three more limits follow from the same anonymisation. Dates are truncated to year and month only, so no day-of-week or time-of-day analysis is possible from the crime file. Home Office offence codes are collapsed into 14 broad crime types, so “violence and sexual offences” is one bucket covering a very wide range of seriousness. And the offence reference is replaced with a one-way hash, which is stable enough to join crimes to outcomes but useless for identifying a case.

The site is unusually candid about its own quality problems, and you should read them before publishing: estimates of geocoding accuracy vary between forces from 60% to 97%; court results are attached by fuzzy matching with success rates from 19% to 97% depending on where the crime happened; six forces are suspected of duplicating some anti-social behaviour incidents; and each month’s file is a snapshot that may later be reclassified or relocated in force systems without the site ever finding out. Neither the British Transport Police nor PSNI supply outcome data at all, and the stop and search dataset covers only a subset of forces in England and Wales — so a national stop and search comparison built from it is incomplete by construction. For mapping technique once you understand these limits, see our mapping UK data guide.

Rates, denominators and small numbers

1
Never compare raw counts across areas
Birmingham records more offences than Rutland because more people live there. Convert to a rate per 1,000 or per 100,000 residents before any comparison, and state which you used.
2
Match the geography exactly
Home Office open data publishes at police force area and Community Safety Partnership level (usually a local authority). ONS also publishes police force area data tables. Use a population denominator for that same boundary and the same year — ONS mid-year population estimates.
3
Interrogate the denominator
Resident population understates the population at risk wherever the daytime, student or visitor population is large: city centres, transport hubs, resorts, university districts. If you cannot adjust for it, say so in the copy rather than crowning a seaside town Britain’s crime capital.
4
Respect small numbers
At ward, CSP or rare-offence level, counts are small enough that ordinary variation dominates. Two homicides one year and none the next is not a 100% fall. ONS says exactly this of homicide: because it is a low-volume offence "there tend to be fluctuations in numbers from year to year", especially in sub-group breakdowns.
5
Use several years, not two
A single year-on-year change in a small area is mostly noise. Plot five or ten years, or aggregate to a bigger geography, and only then decide whether there is a trend to report.
6
Do not build local estimates from the CSEW
The survey targets roughly 31,000 interviews a year with a minimum target of 585 per police force area. That supports national and force-level headline estimates, not wards — and ONS warns that estimates of less frequently occurring crime types can be subject to substantial variability.

Outcome rates — and what “charge rate” really means

Outcomes are recorded separately from crimes and published by the Home Office in the crime outcomes release and the open data tables. There are 22 outcome types — outcomes 1 to 22, with alternatives 1A, 2A and 3A added in April 2016 — covering charges, cautions, penalty notices, community resolutions, CPS decisions not to prosecute, evidential difficulties where the victim does not support action, and investigations closed with no suspect identified. “Charged or summonsed” is one category among those. It is not a synonym for solved, and the remainder are not all failures.

The bigger trap is the denominator. The Home Office publishes two different calculations. One traces outcomes for offences recorded in the period. The other counts outcomes for investigations closedin the period, where outcome volumes relate to outcomes recorded in that year regardless of when the associated crime was recorded — which can produce rates over 100% for low-volume crime types, and which the technical annex itself describes as “illustrative rather than ‘true’ rates”. If you compare a figure calculated one way against a figure calculated the other way, the comparison is meaningless.

The Home Office also warns directly that care is needed when comparing crime outcomes between police forces, because forces differ in how they implement outcome recording and in the mix of crime they deal with. Two further mechanical points: crimes recorded near the end of a period have had less time to reach an outcome, so recent periods look worse than they will once revised; and outcome data are drawn from the same non-accredited recorded crime collection, so they inherit its recording-quality caveats.

How recording changes manufacture trends

A change in what gets written down looks exactly like a change in what happened. These are the discontinuities you are most likely to run into.

  • The National Crime Recording Standard, 2002: Introducing the NCRS increased the number of crimes recorded. ONS bulletins therefore focus on police recorded crime trends from the year ending March 2003 onwards — comparisons that reach back past that point are comparing two different counting regimes.
  • Improved compliance after 2014: Forces have been pushed hard to record more of what is reported to them. OSR’s May 2024 review found recording accuracy had risen from 80.5% to 92.4% nationally over the decade. That improvement alone pushes recorded crime up without any change in actual offending.
  • Fraud moved to central recording, 2014: Changes to reporting and recording arrangements meant many fraud offences previously recorded by individual forces were recorded centrally instead. Force-level fraud series break at that point.
  • Conduct crimes, 2018: Stalking, harassment and controlling or coercive behaviour are now recorded as the principal offence where the conduct crime is judged the more impactful one — so offences that often occur alongside them drop out of the recorded series without any change in behaviour.
  • Any local operation, campaign or crackdown: Proactive policing generates recorded crime. A rise in drug possession, weapon possession or public order offences after a targeted operation measures the operation, not the neighbourhood.

The practical test before publishing any recorded-crime trend: read the release’s methodology and revisions notes for the offence you are writing about, and ask the force’s press office directly whether recording practice for that offence has changed in the period. Both answers belong in the story.

Four jurisdictions, not one

“UK crime figures” is almost always wrong. The ONS crime bulletin and the CSEW cover England and Wales only.

Scotland
A separate system. Recorded crime is collected under the Scottish Crime Recording Standard, introduced on 1 April 2004, rather than the National Crime Recording Standard and Home Office Counting Rules used in England and Wales. The Scottish Government’s user guide states flatly that "recorded crime statistics for England and Wales are not directly comparable with those in Scotland": although the main principles of the two standards are the same, the counting rules differ, and differences in legislation and common law also have to be taken into account. Scotland additionally distinguishes between "crimes" and "offences", a distinction not made in England and Wales. Recorded Crime in Scotland is accredited official statistics; the victimisation counterpart is the Scottish Crime and Justice Survey, covering adults aged 16 and over in private households.
Northern Ireland
Closer to England and Wales, because PSNI has adopted the NCRS and the Home Office Counting Rules. PSNI publishes its own recorded crime bulletins, and the victimisation counterpart is the Northern Ireland Safe Community Survey. NI crime records do appear on data.police.uk, but PSNI supplies no outcome data to it.
England and Wales
The ONS bulletin, the CSEW, Home Office open data tables and data.police.uk all sit here. Even within it, check whether a figure is England-only, Wales-only, or both — and whether British Transport Police, which polices the railways across Great Britain, is included in the total you are quoting.

Comparing forces fairly

  • Compare rates, not counts, and use the same population denominator and year for every force in the table.
  • Check recording accuracy before ranking. Forces differ in how completely they record what is reported to them — an improvement in recording pushes a force up a "worst for crime" league table it did not earn. HM Inspectorate of Constabulary and Fire & Rescue Services publishes crime data integrity findings force by force.
  • Do not rank forces on fraud. Fraud is routed to a national reporting centre and supplemented by industry referrals, so it does not sit in force recorded crime the way other offences do.
  • Hold the offence definition constant. Comparing one force’s "violence" against another’s only works if both are the same Home Office offence classification, not a data.police.uk broad category on one side and a specific offence code on the other.
  • Account for the crime mix when comparing outcome rates — the Home Office says explicitly that forces differ in crime mix and in how they implement outcome recording.
  • Give the force the numbers before publication and put its explanation in the piece. A recording change, a large historic-abuse investigation, or a single organised-crime operation can move a force-level series on its own.

Red flags in your own copy

  • The words "crime is up" attached to a recorded-crime figure, with no mention of recording practice.
  • A map of data.police.uk points presented as the locations where crimes happened.
  • "UK crime figures" sourced from an ONS release that covers England and Wales only.
  • A local rate built on fewer than about twenty offences, reported as a percentage change.
  • A "charge rate" quoted without saying whether the denominator is crimes recorded or investigations closed.
  • A long-run recorded-crime chart that crosses 2002, 2014 or the pandemic years without a break or an annotation.
  • A retail-crime or business-crime story citing the CSEW, which excludes crime against commercial bodies entirely.

Before you publish checklist

  • I have named the source in the copy — CSEW, police recorded crime, Home Office outcomes, or data.police.uk — and linked the release.
  • I have stated the geography and the exact period, and confirmed the release covers it.
  • I have checked whether the statistic is accredited official statistics or official statistics, and not overstated its authority.
  • I have converted counts to rates using a matching population denominator, and named the denominator.
  • I have read the methodology and revisions notes for the specific offence, and asked the force whether recording practice for it changed during the period.
  • If mapping, I have explained in the piece that data.police.uk locations are anonymised snap points.
  • If quoting an outcome rate, I have stated which calculation it uses.
  • I have put the numbers to the force or the relevant body and included their response.

When the published data stops short

Published crime data is aggregated, anonymised and capped at 14 broad categories. Ward-level breakdowns, offence sub-codes, repeat-victim counts and internal recording-audit results generally have to be requested. Forces are public authorities for FOI purposes.

Police FOI RequestsFOI Request Builder

Frequently asked questions

What is the difference between police recorded crime and the Crime Survey for England and Wales?
Police recorded crime counts notifiable offences that were reported to the police and then recorded by them under the Home Office Counting Rules. The Crime Survey for England and Wales (CSEW) asks a large sample of household residents what has happened to them, whether or not they told the police. The ONS user guide states the CSEW "provides a better reflection of the true extent of crime experienced by the population resident in households in England and Wales than police recorded statistics, because the survey includes crimes that are not reported to, or recorded by, the police", and that it is "unaffected by changes in levels of reporting to the police or police recording practices". Police recorded crime, by contrast, covers a wider range of offences and a wider population — including businesses, institutions and visitors — but only the fraction that reaches a crime record.
Can I describe police recorded crime as an official statistic?
Yes, but not as an accredited one. The Office for Statistics Regulation removed the National Statistics accreditation from police recorded crime for England and Wales in 2014 after an assessment found evidence that the quality and consistency of the underlying data might not be reliable. It has not been restored. The ONS crime bulletin labels police recorded crime as official statistics that are not accredited official statistics. OSR published a systemic review of the quality of those statistics on 16 May 2024 which found forces are recording crime substantially more accurately than in 2014, and framed the review as a first step towards reconsidering accreditation. Statistics based on the Home Office Homicide Index, and CSEW estimates, are accredited.
Why should I not map data.police.uk crimes to an exact address?
Because the coordinates are not the coordinates of the crime. data.police.uk maintains a master list of anonymous map points — street centre points and public or commercial locations — and each point must have a catchment area containing at least eight postal addresses or no postal addresses at all; points with between one and seven addresses were discarded. Every crime is moved to the nearest point on that list, and its real coordinates are thrown away. If the nearest point is more than 20km away, the coordinates are zeroed out entirely. A cluster of dots on a street corner therefore tells you a snap point is there, not that crimes happened there.
Does data.police.uk cover the whole UK?
No. Its geographic coverage is England, Wales and Northern Ireland. The crime and outcomes data comes from the 43 geographic police forces of England and Wales, the British Transport Police, the Police Service of Northern Ireland and the Ministry of Justice. Scotland is not included — for Scotland you need the Scottish Government’s Recorded Crime in Scotland series and Police Scotland’s own publications. Note also that neither the British Transport Police nor PSNI supply outcome data, and the stop and search dataset covers only a subset of forces in England and Wales.
How should I work out a crime rate for a local area?
Divide recorded offences by the resident population of exactly the same geography and the same period, then express it per 1,000 or per 100,000 residents. Use ONS mid-year population estimates for the denominator, and make sure the boundary matches: Home Office open data publishes by police force area and by Community Safety Partnership area, which is usually a local authority. Two warnings. Resident population is the wrong denominator where the daytime or visitor population dwarfs it — city centres, seaside towns, university districts — so say so rather than reporting a resort as Britain’s crime capital. And a rate built on a handful of offences will swing wildly year to year for reasons that have nothing to do with policing.
What does a force’s "charge rate" actually mean?
It depends which of two calculations the publisher used, and they are not interchangeable. The Home Office publishes outcomes both for offences recorded in the period — tracing each recorded crime through to whatever outcome it eventually received — and for investigations closed in the period, where "outcome volumes relate to outcomes recorded in that year regardless of when the associated crime was recorded". The second measure can exceed 100% for low-volume crime types, because outcomes assigned this year may attach to crimes recorded in earlier years; the technical annex describes those rates as "illustrative rather than ‘true’ rates". There are 22 outcome types, so "charged" is one specific category, not a synonym for "solved".
Why do fraud figures behave so strangely at local level?
Because fraud is not recorded the way other crime is. Police forces in England and Wales direct fraud victims to the national reporting centre run by the City of London Police as national lead force for fraud, and the recorded fraud series also incorporates referrals sent to the Home Office directly by Cifas and UK Finance. So a fraud committed against someone in your patch does not necessarily appear in your force’s recorded crime the way a burglary would. The routing has changed twice: many fraud offences moved from force recording to central recording in 2014, and a new national reporting service, Report Fraud, replaced Action Fraud on 4 December 2025. As of this guide’s last review, ONS and the Home Office had agreed the City of London Police could temporarily suspend supplying recorded fraud and cybercrime data during that transition, and those figures were not presented in the latest crime bulletin. Check the current release before writing about recorded fraud trends.

Related guides

Data JournalismData Journalism HubRead →Data JournalismUK Statistical Bodies: ONS, NRS, NISRA and UKSARead →Data JournalismMapping UK DataRead →BeatsCrime Reporting BeatRead →InvestigativePolice Misconduct InvestigationsRead →FOIFOI Requests to Police ForcesRead →

Primary sources

Check the live release. Crime statistics are revised and republished on a quarterly cycle, and methodology changes are announced inside the releases themselves. Verify every figure against the current publication before you file.