Last Updated on September 28, 2026 by Karl Thompson
When people talk about official crime statistics for the UK, they usually mean the figures for England and Wales. Scotland and Northern Ireland have their own police forces and publish their own statistics. The England and Wales figures come from two main sources, and the first thing to understand about them is that they tell different stories:
- Police recorded crime counts the crimes that are reported to and recorded by the 43 police forces of England and Wales, plus the British Transport Police.
- The Crime Survey for England and Wales (CSEW) is a face-to-face victim survey that asks people about the crimes they experienced in the previous 12 months, whether or not they told the police.
- Neither source counts all crime. The gap between what happens and what gets counted is the dark figure of crime, and on the latest survey only around a third of crimes were reported to the police at all.
- Self-report studies and local victim surveys are the main alternatives, and each fills a different gap.
- Sociologists disagree about what crime statistics actually measure. Positivists treat them as a flawed but usable measure of crime; interactionists treat them as a record of what the police and courts do; Marxists and feminists focus on the crimes they leave out.
Crime statistics are usually taught at the very start of the crime and deviance module (Paper 3, 7192/3), because every theory you meet afterwards either relies on them or attacks them. They also turn up in theory and methods, where official statistics are a type of secondary data. This post covers crime statistics specifically; the general strengths and limitations of official statistics are in my post on official statistics in sociology.
Police recorded crime
Police recorded crime is exactly what it sounds like: the offences the police have written down. The Home Office collects the figures from every force and the ONS publishes them.
For a crime to appear in these statistics it has to get through two filters. Somebody has to report it, and then the police have to record it as a crime. Most of the criticism of police statistics is about one filter or the other.
Police figures do cover things the survey can’t. They include crimes against businesses, crimes against children, crimes against people living in institutions such as care homes, and crimes against society with no individual victim, such as drug possession. They are also the only realistic source for rare, serious offences.
In the year ending March 2026 the police recorded around 5.2 million crimes, excluding fraud and computer misuse. That was down 1% on the previous year, but up from 3.6 million in the year ending March 2015.
HOWEVER, that ten-year rise does not mean crime went up. The ONS says the increase mostly reflects improvements in how the police record crime, and it describes police figures as a better indicator of police activity than of crime trends. Police recorded crime lost its status as accredited official statistics in January 2014, after the Office for Statistics Regulation found that the underlying data might not be reliable. Remember that date — it comes up again below.
The ONS prefers police figures for homicide, offences involving weapons and robbery. Homicide fell to 499 offences in the year ending March 2026, the lowest since current recording practices began in 2003.
The Crime Survey for England and Wales
The Crime Survey began in 1981 as the British Crime Survey. Trained interviewers visit a sample of households and ask people aged 16 and over whether they have been a victim of crime in the last year, and whether they reported it.
Its great strength is that it doesn’t depend on the police. If the police change how they record crime, or people become more willing to report it, the survey isn’t affected. Its methods have stayed comparable since 1981, which is why the ONS treats it as the best measure of long-term trends in the crimes it covers. The survey’s estimates are accredited official statistics; police recorded crime is not.
It has real gaps, though. It leaves out:
- crimes against businesses and public bodies, including shoplifting
- people living in communal establishments such as student halls, care homes and prisons
- children under 16
- crimes with no specific victim, such as drug possession
- homicide, for the obvious reason that the victim can’t be interviewed
Fraud and computer misuse were only added in the year ending March 2017, so there is no long-run trend for either.
Because the survey interviews a sample rather than everyone, its figures are estimates with a margin of error. When the ONS says a change is “not statistically significant”, it means the change is too small to be sure it isn’t down to chance.
Questions about domestic abuse, sexual victimisation and stalking are asked in a separate self-completion section, so that respondents can answer privately. Victims are much less likely to disclose these crimes to an interviewer, particularly if the offender might be in the house.
Why the two sources give different figures
When you look at crime statistics, always check which source they come from. The two figures will differ, and the size of the gap varies from one type of crime to another.
For some crimes the gap is small. Vehicle theft figures tend to be similar in both sources, because an insurance claim needs a police crime reference. The ONS makes a similar point about residential burglary, which is well reported and well recorded.
For other crimes the gap is huge. Domestic abuse is the classic case: most incidents never reach the police, many more are disclosed to the survey, and some are never disclosed to anyone.
The two sources can even move in opposite directions. Between 2015 and 2026 police recorded crime rose by around 1.6 million offences, while crime measured by the survey fell. Almost all of the difference is down to recording practice, not to a change in how much crime was committed.
The latest figures: year ending March 2026
| Crime type | Crime Survey estimate | Police recorded |
|---|---|---|
| Fraud | 4.5 million incidents | Not published this year (see below) |
| Theft | 2.6 million incidents | 1.6 million offences |
| Violence with or without injury | 989,000 incidents | 507,169 with injury; 829,017 without |
| Computer misuse | 798,000 incidents | Not published this year |
| Criminal damage | 646,000 incidents | 433,251 offences |
| Robbery | 71,000 incidents | 87,362 offences |
| Homicide | Not covered | 499 offences |
Source: ONS, Crime in England and Wales: year ending March 2026, published 23 July 2026.
Don’t try to subtract one column from the other to find the dark figure. The columns count different people, over different time periods, using different rules. The fact that they can’t be lined up neatly is itself the methods point.
Long-term trends: the crime drop
The survey shows a long and steep fall in the traditional volume crimes since their peak in the mid-1990s. On the ONS figures for the year ending March 2026, theft is 77% below its 1995 peak, violence with or without injury is 78% below, and criminal damage is 81% below.
Fraud complicates the picture. It is now the largest single category of crime the survey measures, at 4.5 million incidents, and it is 32% higher than when it was first measured in the year ending March 2017. Computer misuse has gone the other way, falling by around 55% over the same period.
Even counting fraud and computer misuse, the survey’s headline figure of 9.6 million incidents is 14% lower than in the year ending March 2017. So crime has fallen even on the measure that includes the new crimes. What has really changed is the shape of crime: less of it happens in the street, and more of it happens through a bank account.
The dark figure of crime
The dark figure of crime is the crime that never appears in the official statistics, either because it was never reported or because it was reported and never recorded.
Reporting. In the year ending March 2026, only an estimated 36% of the crimes covered by the survey’s headline measure were reported to the police, excluding fraud and computer misuse. Victims may think an incident is too trivial to bother with, may not realise a crime has happened, may think the police couldn’t or wouldn’t do anything, may fear the offender, or may see it as a private matter. Some crimes have no victim to report them at all.
Recording. Reporting a crime doesn’t guarantee it gets recorded. In 2014 Her Majesty’s Inspectorate of Constabulary (HMIC) published Crime-recording: making the victim count, the most thorough inspection of police recording ever carried out in England and Wales. It found that 19% of crimes reported to the police were not recorded, which came to more than 800,000 crimes a year. For violent crime the under-recording rate was 33%, and for sexual offences 26%.
The same inspection surveyed police officers and staff. Of the 8,600 who said they were responsible for recording decisions, 39% said that performance and other pressures were distorting those decisions. That is about as clear an admission as you will ever get that the statistics are shaped by the organisation producing them.
Recording has improved a great deal since then. By the year ending March 2025 the inspectorate, now HMICFRS, estimated that the police were recording 94.8% of crime correctly, against 80.5% in 2014. That improvement is exactly why police recorded crime rose while the survey fell.
Shoplifting shows both filters at work. The police recorded 507,086 shoplifting offences in the year ending March 2026, down 4% after several years of increases. The ONS notes that reporting rates for shoplifting are known to be very low, and that recent rises partly reflect retailers and police working more closely together to report incidents. The survey can’t help here, because it doesn’t cover crimes against businesses.
Other ways of measuring crime
Self-report studies
Self-report studies ask people what crimes they have committed, rather than what they have suffered. They can reveal offending that never comes to the attention of the police, and show whether the people who get caught are typical of the people who offend.
The largest example in England and Wales was the Home Office’s Offending, Crime and Justice Survey, which ran every year from 2003 to 2006. It was the first national longitudinal self-report offending survey for England and Wales, and covered people aged 10 to 29 living in private households.
The limitations are fairly obvious. People may lie, forget, or exaggerate to impress. Questions tend to focus on minor offences. And the most serious and prolific offenders are the least likely to be sitting in a private household agreeing to fill in a questionnaire.
Local victim surveys
National surveys produce national averages, and averages hide concentration. Local victim surveys focus on one area in much more depth.
The best-known is The Islington Crime Survey (Jones, MacLean and Young, 1986), based on around 2,000 door-to-door interviews. It was the first major study of crime, victimisation and policing in a British inner city. It showed that crime and the fear of crime were concentrated in particular streets and among particular groups, in a way the national survey couldn’t detect. Jock Young was one of the founders of left realism, and the survey was part of that approach’s argument that crime is a real problem for poor communities, not just a label applied to them. See my post on left realism.
Sociological perspectives on crime statistics
Positivism and consensus theories: take the figures at face value
Functionalist and subcultural theorists broadly accepted the official statistics as an accurate picture of who offends. If the figures showed that young working-class males committed more crime, then the job was to explain why, and theories like Merton’s strain theory and Cohen’s status frustration set out to do exactly that. Right realists take a similar view today.
The weakness of this approach is that if the statistics are biased, every theory built on them inherits the bias.
Interactionism: statistics measure the police, not crime
Kitsuse and Cicourel (1963) argued that sociologists had been asking the wrong question. Instead of asking what crime statistics tell us about criminals, we should ask what they tell us about the organisations that produce them. Their phrase for this was “rate-producing processes”: the everyday decisions by police, courts and other agencies about what counts as a crime and who counts as a criminal.
Cicourel’s own study, The Social Organization of Juvenile Justice (1968), applied this to two Californian cities. He found that police and juvenile officers worked with a picture of the “typical delinquent” as working-class and from a troubled home. Young people who fitted the picture were more likely to be arrested and charged. Middle-class young people who did the same things were more often dealt with informally, and never became a statistic.
This links directly to labelling theory and to the social construction of crime.
Criticism: the interactionist view can go too far. The survey is independent of the police and still finds real patterns, so the statistics are not purely a product of policing.
Marxism: the crimes of the powerful are missing
For Marxists the problem is less what the statistics include than what they leave out. Steven Box, in Power, Crime and Mystification (1983), argued that the crimes of the powerful — corporate crime, crimes committed by the police, and rape and sexual assault — are systematically under-counted and under-punished, while working-class street crime is over-represented.
The official sources bear some of this out by their very design. The survey only counts crimes against individuals and households, not against businesses or society, and police statistics depend on crimes being noticed and reported. Health and safety offences, environmental crime and fraud against organisations don’t appear in the survey at all, and reach the police statistics only if someone notices and reports them. See my post on the Marxist theory of crime.
Feminism: the hidden figure of violence against women
Feminists have long argued that domestic abuse and sexual violence are among the most under-reported crimes of all. The survey’s self-completion section exists largely because of this. On the ONS figures, 2.1% of people aged 16 and over experienced sexual victimisation in the year ending March 2026 (2.8% of women and 1.4% of men). The ONS stresses that these figures come from newly redeveloped questions and can’t be compared with earlier years.
The latest survey estimate for domestic abuse is for the year ending March 2025, when 7.8% of people aged 16 and over, around 3.8 million people, experienced it. The police recorded 834,256 domestic abuse-related offences in the year ending March 2026.
These two figures measure different things — one counts people affected, the other counts offences recorded — so don’t calculate a reporting rate from them. What they do show is the scale of the gap between what people experience and what reaches the police.
For how the statistics look by gender, see my post on gender and crime statistics.
Left realism: local surveys and the victim’s view
Left realists accept that the official statistics under-count, but argue that the answer is better measurement rather than dismissing the figures. Their local victim surveys, above, put the victim’s experience at the centre and showed that crime falls hardest on the poorest communities.
Strengths and weaknesses of each source
Police recorded crime
| Strengths | Weaknesses |
|---|---|
| Covers crimes the survey misses: crimes against businesses, children and people in institutions, and crimes against society such as drug possession | Only counts crimes that are reported and then recorded. Around 36% of the crimes the survey covers are reported, and HMIC found 19% under-recording in 2014 |
| The best source for rare, serious crimes: homicide, weapons offences and robbery | Shaped by police decisions, priorities and performance pressures |
| Local and up to date, down to street level on police.uk | Changes in recording rules move the figures, as with the 2025 robbery clarification, and most of the rise since 2015 is better recording, not more crime |
| Recording accuracy has improved a great deal, to 94.8% by 2025 | Lost its accredited official statistics status in 2014. For crimes with no victim, it measures police activity almost by definition |
The Crime Survey for England and Wales (CSEW)
| Strengths | Weaknesses |
|---|---|
| Independent of police reporting and recording, so it counts crimes victims never report | Leaves out businesses, people in institutions, under-16s, crimes with no victim, and homicide |
| Methods comparable since 1981, which is how we know the crime drop is real | Based on a sample, so its figures are estimates with a margin of error, and many year-on-year changes are not statistically significant |
| Accredited official statistics, and changes to the survey are documented by the ONS | Victims under-disclose domestic and sexual violence, even with the self-completion section |
| The private self-completion section improves measurement of sensitive crimes | Changes to questions break the trend, as with sexual victimisation (not comparable) and anti-social behaviour (inflated) in 2025–26 |
Positive evaluations of official crime statistics
They are free, published regularly and easy to get hold of. Anyone can read the ONS bulletin, and police.uk lets you look up recorded crime in your own area on a map. Remember that the map only shows police recorded crime, and many crimes are never recorded. (When I was teaching in Surrey, the local data made my teacher’s salary feel even more measly, given the conspicuous wealth all around. The figures are a lot less interesting than the relative deprivation.)
They are large-scale. The survey interviews a big national sample, and police statistics cover every force in England and Wales, so both are far more representative than any study a sociologist could run alone.
The survey allows comparisons over a very long period. Its methods have been consistent since 1981, which is the only reason we can say with confidence that the crime drop is real.
The two sources complement each other. The survey is better for common crimes against individuals; police figures are better for rare and serious crimes and for crimes against businesses. Used together, they give a fuller picture than either alone.
They show patterns that need explaining. Even their critics use them — Box’s argument depends on comparing what the statistics show with what they leave out.
Criticisms of official crime statistics
The dark figure means neither source counts all crime, and the missing crime isn’t random. Some crimes, such as domestic abuse, sexual offences and crimes against businesses, are much more likely to go uncounted than others.
Police statistics are shaped by police decisions. Recording practice, performance pressures and priorities all affect the figures. Crimes with no victim, such as drug possession, only appear when the police go looking for them, so they measure police activity almost by definition.
Changes in rules create changes in figures. When the counting rules or survey questions change, the statistics move, even if crime hasn’t. Several examples from 2025–26 are below.
The survey has blind spots. It misses businesses, institutions, under-16s, victimless crime and homicide, and as a sample it only produces estimates.
The categories don’t always match what sociologists want to study. Crime statistics contain no measure of social class at all, as I explain in my post on social class and crime, so any claim about class and crime rests on substitutes such as area deprivation.
For interactionists the problem is deeper: the statistics are socially constructed, and tell us as much about the people who produce them as about crime itself.
How crime statistics are made: four examples (2026 update)
The latest ONS bulletin shows how rules, systems and question wording produce movements in the figures that have nothing to do with how much crime was committed.
Recorded fraud has temporarily disappeared
A new national system for reporting fraud and cybercrime, Report Fraud, replaced Action Fraud on 4 December 2025. While the City of London Police move across to it, the Home Office and the ONS have agreed to suspend the supply of recorded fraud and computer misuse data. So the largest category of crime in England and Wales has no police recorded figure in the latest bulletin. The survey estimates are unaffected, which is a good illustration of why the ONS prefers the survey for fraud.
A rule change almost doubled business robbery
In April 2025 the Home Office clarified the counting rules with police forces. If an offender steals from a shop and uses or threatens violence against a shop worker, security guard or anyone else, including to escape, it should be recorded as robbery of business property rather than as shoplifting. Recorded business robbery rose by 91% to 29,669 offences, and overall robbery rose by 10%. Meanwhile robbery of personal property actually fell by 9%. The ONS also thinks the clarification slightly reduced recorded shoplifting.
Same behaviour, different box on the form, very different headline.
A question change inflated anti-social behaviour
The survey found that 40.9% of people had experienced or witnessed anti-social behaviour in the year ending March 2026, up from 35.5% the year before. The ONS says some of the increase was genuine, but that it was inflated by changes to how the questions were asked from April 2025.
New questions on sexual victimisation
The survey’s questions on sexual victimisation were redeveloped to measure it better. The ONS says the latest estimates can’t be compared with previously published ones, so there is no trend for this year.
Each of these supports the interactionist argument that statistics are produced, not simply collected. HOWEVER, notice that in each case the ONS tells you about the change. That transparency is one of the strongest defences of the official statistics: their flaws are documented, which is more than can be said for most data.
Crime statistics in the exam
Crime statistics come up in three ways on the AQA specification:
- As evidence in the crime and deviance section of Paper 3. Any question on patterns of crime by class, gender, ethnicity or age depends on them, and good answers say which source a figure comes from.
- As an evaluation point against theories. Criticising a theory for relying on official statistics is standard AO3, especially against functionalist, strain and subcultural theories.
- In theory and methods. Official statistics are a type of secondary data, and crime statistics are the best-known example in the positivist versus interpretivist debate.
Two practice questions to try:
Outline two reasons why police recorded crime may underestimate the true extent of crime. (4 marks)
Pointers: victims may not report crimes, for example because they think them too trivial or fear the offender; the police may not record crimes that are reported, as HMIC found in 2014.
Outline and explain two reasons why official crime statistics may not give a valid picture of crime. (10 marks)
Pointers: the dark figure (non-reporting and non-recording, with the 36% reporting figure and HMIC’s findings); and the social construction of the statistics (Kitsuse and Cicourel’s rate-producing processes, Cicourel’s typifications, counting rule changes such as the 2025 robbery clarification). Remember that on the no-item 10-markers there are no marks for evaluation — explain each reason fully instead.
Official crime statistics: FAQs
What are the two main sources of crime statistics in England and Wales? Police recorded crime, which counts crimes reported to and recorded by the police, and the Crime Survey for England and Wales, a face-to-face victim survey of people aged 16 and over.
What is the Crime Survey for England and Wales? A face-to-face survey, run by the ONS, that asks a sample of people aged 16 and over in households whether they have been a victim of crime in the last 12 months, whether or not they reported it to the police. It began in 1981 as the British Crime Survey.
What is the dark figure of crime? The crime that never appears in official statistics, because it isn’t reported or isn’t recorded. In the year ending March 2026 only around 36% of the crimes covered by the survey’s headline measure (excluding fraud and computer misuse) were reported to the police.
Which source is more reliable? The ONS treats the Crime Survey as the better measure of long-term trends for the crimes it covers, because it isn’t affected by changes in reporting or recording. Police figures are preferred for rare, serious crimes such as homicide, weapons offences and robbery.
Revision bundle
If you like this sort of thing, then you might like my Crime and Deviance revision notes bundle, written by an AQA examiner of eight years’ standing.
[IMAGE: Crime and Deviance revision bundle cover, linked to https://revisesociology.sellfy.store/p/juqd/]
Signposting / Related Posts
This post has been written primarily for students studying the crime and deviance module, part of Paper 3 of the AQA’s A-level sociology course.
The social construction of crime takes the interactionist argument further, showing how changes in the law create and remove crimes.
Official statistics in sociology covers the general strengths and limitations of official statistics as a research method.
The patterns in crime statistics are covered in more depth in social class and crime, gender and crime statistics and official statistics on ethnicity and crime.
You might also like this crime statistics revision video.
Please click here to return to the crime and deviance hub, or to the homepage — ReviseSociology.com.
Sources used to write this post
Box, S. (1983) Power, Crime and Mystification. London: Tavistock. Cicourel, A. V. (1968) The Social Organization of Juvenile Justice. New York: Wiley. HMIC (2014) Crime-recording: making the victim count. HMICFRS (2025) How effectively do the police record crime? Jones, T., MacLean, B. and Young, J. (1986) The Islington Crime Survey: Crime, Victimization and Policing in Inner-City London. Aldershot: Gower. Kitsuse, J. I. and Cicourel, A. V. (1963) ‘A note on the uses of official statistics’, Social Problems, 11(2), pp. 131–139. ONS (2026) Crime in England and Wales: year ending March 2026. ONS, User guide to crime statistics for England and Wales.
This post has been written primarily for A-level sociology students. This topic is part of the crime and deviance module, usually studied in the second year.