Exploring diverging trends between the Crime Survey for England and Wales and police recorded crime

Over the last ten years, trends in crime shown in the Crime Survey for England and Wales (CSEW) and police recorded crime have appeared to diverge. While the CSEW has seen a continued downward trend, there have been rises in police recorded crime, predominantly because of improvements in police recording practices. While a wide range of offences are covered by both data sources, adjustments need to be made to both series to account for differences in coverage.

We explored this divergence using two comparable subsets of crime types from both series. The data showed:

• a reduction in crime across both series from the year ending (YE) March 2003 to YE March 2014

• that the CSEW series generally continued to decrease over time, while the police recorded crime series saw rises between YE March 2015 and YE March 2019 because of improvements to police recording practices

• that improvements to police recording practices have had the largest impact on trends in violence with and without injury offences; when these offences were removed from both comparable subsets, trends across both time series were similar

• the latest police recorded crime data have seen a flatter trend and a small fall in YE March 2024 compared with YE March 2023, suggesting that the impact of inflationary changes because of recording improvements has fallen away

In theory, if all crimes from the CSEW subset were brought to the attention of and subsequently recorded by the police, our comparable crime ratio would be 1 (that is, the number of crimes in both datasets would be equal). However, in YE March 2024 there was 63% (a ratio of 1.63) more crime recorded by the police than crime estimated by the CSEW that respondents also informed us had come to the attention of the police.

Differences in coverage between the two data sources, such as the respondent age range covered by the survey, partly explain this disparity. Further explanations include accounting for known limitations and statistical adjustments made to the CSEW, which largely affect our main estimates of incidents of violence with and without injury. For example, the interviewer-administered parts of the survey underestimate domestic violence and a cap is applied to the number of repeat incidents in our main survey estimates to reduce year-to-year volatility and provide a better measure of trends over time, the primary objective of the survey.

Our two main sources of data for crime statistics in England and Wales are the Crime Survey for England and Wales (CSEW) and police recorded crime.

The CSEW is primarily an interviewer-administered face-to-face victimisation survey. People (aged 16 years and over) resident in households in England and Wales are asked about their experiences of crime in the 12 months prior to the interview. For the population and offence types it covers, the CSEW generally provides the better measure of trends on a consistent basis over time, because it is unaffected by changes in levels of reporting to the police or police recording practices.

Police recorded crime figures are restricted to a subset of notifiable offences that have been reported to and recorded by the police. Therefore, while the police recorded crime series covers a wider population and a broader set of offences than the CSEW (for example, residents of institutions, tourists, and crimes against commercial bodies), it does not include all crimes and especially those that do not come to the attention of the police or are not accurately recorded by them.

For further information on these two data sources, please see our User guide to crime statistics in England and Wales.

Comparable subset of crime

To compare crimes estimated by the CSEW against police recorded crime counts, a comparable subset of crimes has been created for a set of offences that are covered by both series. For more information, see our Analysis of variation in crime trends article (PDF, 175KB). The comparable subsets of crime exclude fraud and computer misuse because only a small proportion of these incidents are reported to the police. For example, 12% of these CSEW incidents were reported to the police or Action Fraud in YE March 2024. Over three-quarters of other CSEW offences reported via interviews in recent years fall into categories that can be compared with crimes recorded by the police.

CSEW estimates are further restricted to those incidents respondents indicate were reported to the police (either by the respondent themselves, someone else or by the police coming to know about it in another way). If we did not restrict CSEW figures in this way, numbers estimated by the CSEW should tend to include more crimes than those recorded by the police, all other things being equal.

Various adjustments are made to the recorded crime categories to maximise comparability with the CSEW. However, police recorded crime numbers are not adjusted to exclude victims of commercial offences and offences committed against children aged under 16 years, as we do not hold the information required to identify and exclude them. This means that, all other things being equal, police recorded crime numbers in the comparable subset should tend to include more crimes than those estimated by the CSEW.

In addition, the mapping between CSEW categories and police recorded offence codes are therefore approximate, and categories will not be directly equivalent in all cases. The categories used in the CSEW, and police recorded crime to produce comparable offence groups are presented here with their respective offence codes. One recent amendment to this comparable subset is the addition of "attempted assault" to the CSEW violence with and without injury category; time-series data (back to the YE March 2003) have been adjusted accordingly.

Comparable offence groups for CSEW and police recorded crime

Violence with and without injury

Crime Survey for England and Wales

• Serious wounding (11)

• Other wounding (12)

• Common assault (13)

• Attempted assault (21)

Police recorded crime

• Assault with intent to cause serious harm (5D)

• Assault with injury​ (8N)

• Racially or religiously aggravated assault with injury (8P)​

• Assault with injury on a constable (8S)​

• Assault with injury on an emergency worker (8T)​

• Assault without injury on a constable (104)

• Assault without injury (105A)

• Racially or religiously aggravated assault without injury​ (105B)

Robbery

Crime Survey for England and Wales

• Robbery (41)

• Attempted robbery (42)

Police recorded crime

• Robbery of personal property (34B)

Theft from the person

Crime Survey for England and Wales

• Snatch theft from the person (43)

• Other theft from the person (44)

• Attempted theft from the person (45)

Police recorded crime

• Theft from the person (39)

Domestic burglary

Crime Survey for England and Wales

• Attempted burglary to non-connected domestic garage/outhouse (50)

• Burglary in a dwelling (nothing taken) (51)

• Burglary in a dwelling (something taken) (52)

• Attempted burglary in a dwelling (53)

• Burglary from non-connected domestic garage/outhouse (nothing taken) (57)

• Burglary from non-connected domestic garage/outhouse (something taken) (58)

Police recorded crime

• Residential burglary of a home (28I)

• Attempted residential burglary of a home (28J)

• Distraction burglary - residential (home) (28K)

• Attempted distraction burglary - residential (home) (28L)

• Residential burglary of unconnected building (28M)

• Attempted residential burglary of unconnected building (28N)

• Distraction burglary - residential (unconnected building) (28O)

• Attempted distraction burglary - residential (unconnected building) (28P)

• Aggravated burglary - residential (home) (29B)

• Aggravated burglary - residential (unconnected building) (29C)

Vehicle-related theft

Crime Survey for England and Wales

• Theft of car/van (60)

• Theft from car/van (61)

• Theft of motorbike motorscooter or moped (62)

• Theft from motorbike, motorscooter or moped (63)

• Attempted theft of/from car/van (71)

• Attempted theft of/from motorbike, motorscooter or moped ​(72)

Police recorded crime

• Aggravated vehicle taking (37.2)

• Theft from a vehicle (45)

• Theft or unauthorised taking of a motor vehicle (48)

• Interfering with a motor vehicle (126)

Bicycle theft

Crime Survey for England and Wales

• Theft of a pedal cycle (64)

Police recorded crime

• Theft or unauthorised taking of a pedal cycle (44)

Criminal damage to a vehicle

Crime Survey for England and Wales

• Criminal damage to a motor vehicle (£20 or under) (81)

• Criminal damage to a motor vehicle (over £20) ​(82)

Police recorded crime

• Criminal damage to a vehicle (58C)

Criminal damage to a dwelling

Crime Survey for England and Wales

• Criminal damage to the home (£20 or under) (83)

• Criminal damage to the home (over £20) ​(84)

Police recorded crime

• Criminal damage to a dwelling (58A)

Analysis of the comparable subset of crime in this section begins at the year ending (YE) March 2003, following the introduction of the National Crime Recording Standard (NCRS) for police forces in April 2002. This was an important change in recording standards and aimed to bring greater consistency to police recording of crime.

Since YE March 2003, estimates of incidents from the Crime Survey for England and Wales (CSEW) in the comparable subset of crime have been on a general downward trend (Figure 1). The number of incidents in the comparable subset was 65% lower in YE March 2024 (1.5 million incidents) compared with YE March 2003 (4.1 million incidents). The numbers of police recorded offences in the comparable subset of crime also fell between YE March 2003 and March 2014, falling by 47% from 3.2 million to 1.7 million offences. However, there were rises in the numbers of offences between YE March 2015 and YE March 2019, with volumes surpassing the volume of reported CSEW incidents during this period.

The divergence in trends between the two data sources can mainly be attributed to improvements in recording practices in police recorded crime. These improvements contributed to rises in police recorded crime between YE March 2015 and YE March 2019. Flattening trends over the last couple of years could suggest that the impact of inflationary changes because of recording improvements has levelled off.

As outlined in the HM Inspectorate of Constabulary and Fire & Rescue Services (HMICFRS) 2014 report, Crime recording: making the victim count, at the national level, an estimated four in five offences (81%) that were brought to the attention of the police and should have been recorded as crimes, actually were recorded. Following improvements to recording practices, the most recent data available from the HMICFRS 2021 report, Police performance: Getting a grip, showed that 92% of offences that were brought to the attention of the police and should have been recorded as crimes, were recorded.

This effect has been most pronounced for violent crime, with recording rates improving from 67% in 2014 to 90% in 2021. When violence with and without injury offences are removed from the comparable crime time series, the impact of improvements to police recording practices is less pronounced and the trends across both time series are more similar (Figure 2).

Another important consideration for understanding shorter term variation in trends are the differences in reference periods for the CSEW and police recorded crime. For example, the latest annual CSEW figures are based on interviews conducted between April 2023 and March 2024. These measure respondents’ experiences of crime in the 12 months prior to their interview. This means crimes measured could have occurred at any time between April 2022 and February 2024 (a 23-month period). In contrast, the latest annual police recorded crime data are based on offences recorded between April 2023 and March 2024 (a 12-month period).

The impact of these differences is likely to be more pronounced at times when crime numbers are significantly rising or falling. For example, this may explain why the difference in the volume of crimes across the two comparable subsets is particularly high for YE March 2023 (Figure 1). The CSEW reference period (April 2021 to February 2023) included times where victimisation patterns would have been affected by social restrictions related to the coronavirus (COVID-19) pandemic, whereas the police recorded crime reference period did not.

The divergence ratio was developed to monitor comparability and contains volume measures of both the Crime Survey for England and Wales (CSEW), and police recorded crime, using the comparable subset of crimes. In theory, if all crimes from the CSEW subset were reported to the police and subsequently recorded by the police, the ratio would be 1. In reality, some variation in the ratio is to be expected because of the inherent variability owing to sampling associated with the survey, and the lack of precise mapping between the offences.

Between the year ending (YE) March 2004 and YE March 2013 the divergence ratio gradually decreased, suggesting a gradual erosion of police compliance with the Home Office counting rules over time (as previously reported in 'Why do the two data sources show differing trends?'). However, the ratio has since increased, rising above 1 in YE March 2017 (Figure 3).

In the latest year (YE March 2024), 63% more crime was recorded by the police than crime estimated by the CSEW that respondents also informed us had come to the attention of the police (a ratio of 1.63). However, this varied by individual crime types (Table 1), with violence with and without injury (2.64), robbery (1.92) and theft from the person (1.48) being substantially higher than 1.

Differences in coverage

Differences in coverage between the two data sources partly explain ratio disparities across a range of crime types as the CSEW only interviews people aged 16 years or over resident in households, whereas police recorded crime data include offences committed against all people of all ages.

Although not included in the comparable crime ratio, estimates of the number of incidents for violence with and without injury, robbery and theft from the person against children aged 10 to 15 years are available from the Children's Crime Survey for England and Wales (CCSEW) up to YE March 2020. An average of the CCSEW estimates for YE March 2015 to YE March 2020 suggest that if they were combined with CSEW estimates, they would account for approximately 24% of violence with and without injury, 21% of robbery, and 9% of theft from the person incidents.

While the omission of crimes against children from the CSEW can largely explain the ratio disparities for robbery and theft from the person offences, they only partially explain the differences for violence with and without injury offences (reducing the ratio from 2.64 to 2.01).

Offences committed against businesses are also not covered in the CSEW. In 2023, approximately 16% of licensed vehicles were registered as company kept (see Department for Transport's Vehicle licensing statistics data tables). This difference in coverage fully explains the ratio disparities for vehicle-related theft and criminal damage to a vehicle as police recorded crime also includes crimes against commercial vehicles.

Although differences in coverage largely explain the ratio disparities for robbery, theft from the person, vehicle theft and criminal damage to a vehicle in YE March 2024, it only partially explains the disparity for total comparable crime. This is because the disparity in the ratio for total comparable crime is predominantly explained by the large disparity for violence with and without injury offences.

Crime Survey for England and Wales limitations

Known limitations of the CSEW for estimating incidents of violence with and without injury are likely to be the most important factors for understanding the ratio disparity for these offences. These include the underestimation of domestic violence in the interviewer-administered parts of the survey and the capping of repeat incidents to reduce year-to-year volatility of survey estimates.

Domestic violence collected in the interviewer-administered parts of the CSEW is under-reported and, therefore, domestic violence incidents are substantially under-estimated by the CSEW. Because of the sensitive nature of the subject matter, domestic abuse (including use of force) is covered by the CSEW in a self-completion section of the survey, as victims may be reluctant to disclose these experiences in the presence of an interviewer. Prevalence rates for domestic violence derived from the self-completion section are significantly higher. For example, in YE March 2023 estimates from the self-completion module showed that there were approximately 491,000 victims of domestic abuse where force was used. This was more than six times higher than the number of victims of domestic violence estimated using data from the interviewer-administered parts of the CSEW (72,000 victims).

While we can estimate the number of victims using self-completion data, we are unable to estimate the number of incidents. The divergence ratio for violence with and without injury for YE March 2024 is considerably lower if domestic violence is excluded from the CSEW comparable subset. This is also the case if violence against the person offences that were flagged as domestic abuse related are excluded from police recorded crime subset. After making these exclusions, the ratio for violence with and without injury reduces from 2.64 to 1.89.

Estimates from the CSEW are capped at the 98th percentile of victim incident counts for each crime type to reduce year-to-year volatility in survey estimates and to provide a better measure of trend over time, which is the main purpose of the survey. Therefore, for some crime types, most notably violence with and without injury, the numbers of incidents may be under-estimated. For example, in YE March 2024, the CSEW main estimate was 1.1 million violence with and without injury offences, but the uncapped estimate was 2.0 million (see our Uncapped CSEW dataset). If we used the uncapped number of incidents instead of the capped, the divergence ratio for violence with and without injury would reduce from 2.64 to 1.44. It is important to note that there is both a high level of volatility and uncertainty around uncapped estimates (2.0 million incidents in YE March 2024; lower bound 1.1 million and upper bound 3.0 million). For further information on the capping of incidents in the CSEW see our Improving victimisation estimates derived from the Crime Survey for England and Wales article.

Figure 4 shows how a combination of differences in coverage between the two data sources and limitations of the CSEW are likely to account for the ratio being above 1 for violence with and without injury. A series of adjustments were made across CSEW estimates and police recorded crime counts for YE March 2016, YE March 2020 and YE March 2024:

• Ratio 2 excluded domestic violence from both time series

• Ratio 3 excluded domestic violence and made additional adjustments by using uncapped CSEW incidents

• Ratio 4 excluded domestic violence and made additional adjustments by using uncapped CSEW incidents and accounting for children’s violence estimates from the 2015-2020 CCSEW

The combination of these factors can fully explain the ratio disparity for violence with and without injury for YE March 2020 and YE March 2024. They are also important for understanding how improvements to police recording practices have affected the ratio, with Ratio 4 increasing from 0.51 in YE March 2016 to 1.01 in YE March 2020. This complements findings from the HMICFRS Police performance report, which showed that the proportion of these crimes that were recorded increased from 67% in 2014 to 90% in 2021.

The Crime Survey for England and Wales (CSEW) gathers information from a sample rather than from the whole population. The sample is designed to be as accurate as possible given practical limitations such as time and cost constraints. Therefore, the estimates contain some uncertainty and are not precise figures. This can have an impact on how changes in the estimates should be interpreted. Where crime is relatively rare, such as robbery, the small number of offences estimated by the CSEW from year-to-year mean estimates have large margins of error and can result in large fluctuations from year to year. Read more information in our how we measure and communicate uncertainty for our surveys article.

Prior to the coronavirus (COVID-19) pandemic, CSEW response rates remained relatively consistent from year to year, and target sample sizes were always achieved. While the CSEW response rate fell from 64% for the year ending (YE) March 2020 to 42% in YE March 2023, our Crime Survey for England and Wales data quality review: June 2024 showed that quality and accuracy of YE March 2023 wave 1 survey data were broadly similar to YE March 2020 data, when these statistics last had accredited official statistics status.

The police recorded crime statistics for England and Wales are currently published as official statistics, not accredited official statistics. Their accreditation was removed in 2014 following an assessment by the Office for Statistics Regulation (OSR), which found evidence that the quality and consistency of the underlying data may not be reliable. While there have been improvements to the quality of crime recording, there are still several areas that require further work before police recorded crime data can regain their accredited official statistics designation, as outlined in OSR's most recent review, The quality of police recorded crime statistics for England and Wales (PDF, 711KB). This also includes the possibility that several police forces are currently over-recording crime in some way. However, this is likely to be at a small scale and have had a minimal impact on the ratio disparity.

Further details on the quality of the CSEW data and police recorded crime can be found in our Crime in England and Wales QMI.