Reading East MP Matt Rodda recently gave his opinion that violent crime was rising and that the government should end police cuts. Mr Rodda used Thames Valley Police performance figures for Reading as evidence for this. The Office for National Statistics (ONS) is wary of interpreting police crime report data to show trends in violent crime, and both Her Majesty’s Inspectorate of Constabulary and Fire & Rescue Services (HMICFRS) and Thames Valley Police (TVP) have suggested that recent rises in violent crime recorded by the police may be due to improvements in police recording procedures.
Estimates of violent crime levels taken from other sources, such as the Crime Survey of England and Wales (CSEW) and studies of hospital admissions for violent injuries in England and Wales, suggest a long-term national decline in violent crime or a recent stagnation in violent crime rates.
On his Facebook page, Matt Rodda said:
Mr Rodda’s office explained that the evidence he used of a rise in violent crime came from TVP’s 2017 crime summary between January and December :
|Offence||2015||2016||2017||% change between 2016 and 2017|
|Victim based crime||12047||12925||14618||+13.1%|
|Violence against the person||3214||3253||3494||+7.4%|
|Violence and sexual offences (combined)||3598||3699||3994||+7.9%|
This suggests an annual increase in violent crime of 7.4%, and in sexual offences of 12.1% between 2016 and 2017. Combining the two suggests an annual rise in violent crime and sexual offences (the category used by TVP in their other data, below) of 7.9% at the end of 2017.
TVP provide monthly breakdowns of crimes in the Thames Valley region; here they combine violence and sexual offences into one category (unlike the crime summary above). If you plot the number of violent crime and sexual offence reports in Reading Borough since December 2011, you get this graph:
The monthly data (dotted line) is very noisy, so a 12-month moving average has also been plotted (solid line) to clarify the apparent trend.
If you plot the same figures showing rolling 12-month percentage changes (for each month, comparing that month’s 12-month total with the same month the previous year), you get this:
In other words, there was a huge increase in the number of violent crime and sexual offence reports during 2015 and the early part of 2016, after which the rate of change flattened out.
But there is more to this data than is immediately apparent.
TVP data is not consistent.
Data may not be directly comparable from one year to another because police services in England and Wales changed how they recorded violent crime in 2015, following HMICFRS recommendations from 2014.
The effect of this change was to rebase police services’ violent crime report level throughout 2015 (and early 2016) so that figures before this time are not comparable with figures afterwards, and crime reporting figures during 2015 and 2016 are so badly affected by these changes that any interpretation is risky.
The ONS say (my italics):
Ongoing work by police forces over the last three years [to September 2017] to improve crime-recording practices are thought to be an important driver of the increase in all police recorded violence, but make interpreting trends in police recorded violence against the person offences difficult.
Nearly all police forces recorded a rise in violence in the latest year to September 2017 compared with the previous year… When interpreting these figures, it is important to bear in mind that these increases will reflect recording improvements and the extent of such effects differs across police forces.
These recording changes may not yet be fully implemented at TVP, and may continue to cause the number of recorded violent crimes in police data to rise. In February 2018, HMICFRS said:
I am disappointed with the quality of crime recording in Thames Valley. Although the force has implemented the recommendations from our previous crime recording inspection in 2014, we found that almost 1 in 5 crimes in Thames Valley are not being recorded properly…
If TVP has implemented some of HMICFRS’s recommendations from 2014, then comparing data between years should be treated with caution. Specifically, if the problem was a failure to record crimes, then implementation of HMICFRS recommendations means that the later the period, the more crime is being recorded; this would tend to increase the number of crime reports in later periods compared to earlier.
The ONS say (my italics):
Police recorded crime is restricted to violent offences that have been reported to and recorded by the police. In addition, due to the ensuing efforts of police forces to tighten recording practice and improve recording processes, this series is not currently believed to provide a reliable measure of trends.
The ONS adds that police recorded crime tends to be more accurate with low volume but high harm offences because these are more likely to be reported:
Police recorded crime showed continuing rises in a number of higher-harm violent offences that are not well-measured by the CSEW as they occur in relatively low volumes. This was most evident in offences of knife crime and gun crime; categories that are thought to be relatively well-recorded by the police. The occurrence of these offences tends to be disproportionately concentrated in London and other metropolitan areas.
The crime report numbers that TVP provide through police.uk are not the same as in their 2017 crime summary, even though both reports claim to use the same data. For example, if you add up violent and sexual offence reports for all of 2017 in all Reading police neighbourhoods (which equates to Reading Borough), you get 3675, but their crime summary says its 3994. Figures for other periods have similar discrepancies, so percentage changes from one year to the next are broadly similar to that in the crime summary.
These discrepancies might be because the crime summary uses a different definition of ‘Reading’ from the police.uk website, or it could be because of issues TVP are having with data, or back fixes to historical data that appear in one dataset and not another.
… crime levels and classifications may change over time and this data may not match data previously published or data published in the future. For this reason, the data should only be used as an indicator of crime trends.
TVP data is noisy.
The range of rolling 12-month percentage changes in reported violent crime and sexual offence reports in Reading during 2017 is shown in this table.
|Month||% change between this 12 month period and the previous|
In case it’s not clear, this table shows a 5.86% increase between the year up to December 2016 and the year up to December 2017, not a 5.86% increase between November and December 2017.
The table suggests the number of violent crime and sexual offences shrank from January to August and then started growing again, but these fluctuations may just be ‘noise’ caused by (for example) weather, criminals passing through, newly trained police officers or members of the public who now don’t want to, or have now decided to, report crime to the police.
Crime data can be very seasonal, which adds to this ‘noise’. A 2016 study from the University of Cardiff [PDF download] into hospital admissions because of injuries due to violence noted:
… violence-related emergency department attendance was most frequent on Saturdays and Sundays and during the months of May, July and October.
The smaller the area under consideration (with fewer people) and the rarer the crime, the ‘noisier’ recorded crime data is likely to be. A single drunk in a car or an industrial accident can make a huge increase in annual violent crime reports for a region, but this may not mean that violent crime has really trended upwards.
The Thames Valley police and crime commissioner (PCC) Anthony Stansfeld explained how one incident can have a big effect on homicide rates during his presentation to Reading Borough Council in October 2017 comparing homicide rates across the TVP region between 2015/16 and 2016/2017 :
Homicide increased by 5.6%. Three of those – it’s a very low figure – were as a result of the Didcot Power Station collapse. If it hadn’t been for that, it would have gone down slightly.
The ONS discounted an apparent 11% decrease in violent crimes nationwide between September 2016 and September 2017 as statistically insignificant. This was based on Crime Survey of England and Wales (CSEW) data, which the ONS consider a “better measure of trends” than police data [note 1]. If annual fluctuations of 11% are insignificant ‘noise’ in the “better” dataset used by the ONS, then it may be reasonable to doubt if annual fluctuations of 11% are meaningful in TVP’s recorded crime data.
The only annual fluctuations in violent crime and sexual offences reports in Reading larger than 11% since 2012 have been an annual increase of up to 17% in late 2013 and early 2014, and an annual increase of up to 50% that started in early 2015 and lasted until summer 2016.
TVP have an explanation for the huge increase in violent crime and sexual offence reports during 2015/16, which matches what the ONS said about the same period:
The increase is reflected nationally and is largely attributed to improvements in compliance with National Crime Recording Standards, following recommendations made by HMICFRS in 2014.
… some of this is down to changes in crime recording and how we capture that information. However there are some increases in crime types, as well as increased victim confidence in reporting certain types of offences.
In other words, the only recent significant change in violent crime report numbers needs to be discounted because it is an artefact of changing police procedures and public attitudes towards reporting violence.
It may be little counter-intuitive, but improvements in crime recording and greater victim confidence in reporting violence is actually a good thing, even if it causes an increase in the number of violent crime reports, because this means that the police and society are taking previously hidden violence seriously.
TVP data may not mean what one thinks it does.
The ONS says:
Violent offences in police recorded data are referred to as ‘violence against the person’ and include homicide, death or serious injury caused by illegal driving, violence with injury, violence without injury, and stalking and harassment. Both actual and attempted offences are included in the figures.
Violent crime covers a wide range of offences including minor assaults (such as pushing and shoving), harassment and psychological abuse (that result in no physical harm), through to wounding and death.
In other words, verbal abuse is included in violent crime figures, as well as offences that may have been intended, but didn’t actually happen. A domestic shouting match may contribute to violent crime numbers (if it gets reported to the police) and, as reprehensible as such behaviour is, it’s not in the same league as physical assault or murder, even though they are all counted together.
What constitutes ‘violent crime’ also changes from year to year. The ONS again (comparing 2017 with 2016):
A new sub-category has been introduced within the main violence against the person offence group, covering offences related to ‘death or serious injury caused by illegal driving’. It contains offences previously counted under ‘violence with injury’.
… stalking and harassment offences have been moved out of the sub-category of ‘violence without injury’ and are now in a separate sub-category along with the new notifiable offence of malicious communications.
In other words, the violent crime figure in one year is unlikely to refer to the same set of criminal activities as another year.
Whether a violent crime is always recorded as such by a given police officer is also unclear. The ONS says:
It is known that violent offences are more prone than some other offences to subjective judgement about whether or not to record a crime.
Nationally, in 2014, an estimated one in three reports of violence that should have been logged as crimes were not recorded as such.
During his presentation to Reading Borough Council in October 2017, the Thames Valley PCC said (about the entire TVP area, comparing 2015/16 with 2016/17) :
Violence against the person went up 7.9%, compared to an increase nationally of 19%. I genuinely think this is because we record at a much lower level. Incidents in schools where somebody has had a punch-up are recorded in crime statistics today.
The ONS summarise their analysis of crime data up to September 2017 by saying that violent crime data in England and Wales “presents a complex picture”:
This fairly flat trend continues… in recent years, with no significant year-on-year change since the survey year ending March 2014.
More recent… inspections carried out by HMICFRS in the last year show that improvements in recording have been made since 2014 and this is likely to have been an important factor in the recorded increase in violent crime.
The ONS say that violent crime is on a long-term national decline from a peak in the early 1990s (using CSEW, not police data), even though the volume of violent crimes recorded by the police has trended upwards during this time. The ONS add that there also appears to be a real upward trend in knife and gun crime in metropolitan areas, although this doesn’t appear to include Reading, because during his presentation to Reading Borough Council on October 2017, the Thames Valley chief constable Francis Habgood said (about the Reading Borough area) :
Violence has been pretty stable for a long time here.
A 2016 study from the University of Cardiff [PDF download], based on admissions to hospitals in England and Wales for injuries sustained from violence, said that there were 40% fewer such admissions in 2016 compared to 2010, and 10% fewer in 2016 compared to 2015 (the period during which most police forces showed a huge spike in violent crime reports). Their report summarised:
This continues the overall steady reductions seen since 2002.
The ONS and the University of Cardiff may be incorrect in their analysis that violent crime rates in England and Wales are declining or flattening. The HMICFRS and the ONS may be unjustified in their criticisms of TVP crime report data. The Thames Valley PCC may be wrong that the volume of violent crime recorded by the TVP is increasing because recording is getting better, and the TVP chief constable may be wrong when he says violence in Reading is stable.
They may all be wrong, but one needs a really good argument to disagree with all of them.
I put these findings to Mr Rodda, who said:
While I understand the complexity of the statistics, I still believe the data you have provided demonstrates an increase in the violent crime rate.
Using your example of the decrease in violent crime of 5% between June 2016 and June 2017, when considered from June 2015 this still represents an increase in crime of 17% [note 2]. Furthermore, although you claim that growth less than 10% [note 3] can be explained by new methods of measurement or seasonal changes, 19 of 36 of the months you have provided show an increase in crime of more than this. This in addition to the fact that none of the decreases in violent crime in Thames Valley are below the 10% the rate you have suggested [note 4].
Regardless, according to the ONS, the general trend across the UK is one of increase in violent crime.
 There are issues with both CSEW and police recorded crime data. See table 5 here.
 I think Mr Rodda came to this 17% figure by averaging the 12-month rate of change from June 2015 until December 2017, a period of 2 and a half years. The dataset I sent him was this one (below) which was a rough draft of the shorter table included earlier the article.
|Month||% change between this 12 month period and the previous|
 Violent crime report numbers for the entire Thames Valley Area, and not just Reading, are also available from Thames Valley Police and the ONS. The ONS data for all of Thames Valley shows that reports of ‘violence against the person’ increased by 6% when comparing the year to September 2016 with September 2017, and TVP data shows the same metric as 8.9% when comparing the year to December 2016 with December 2017.
- Matt Rodda’s website, Facebook page and Twitter page
- Matt Rodda at parliament.uk and TheyWorkForYou
- Office for National Statistics (ONS) : Crime in England and Wales: year ending September 2017
- Thames Valley Police performance figures for Reading
- Thames Valley Police performance figures for the whole Thames Valley
- HM Inspectorate of Constabularies: Thames Valley Police performance
- HM Inspectorate of Constabularies: Too many crimes going unrecorded by Thames Valley Police
- Thames Valley Police crime statistics
- Violence in England and Wales in 2016: An Accident and Emergency Perspective – University of Cardiff [PDF download]
- Crime survey of England and Wales
- Presentation of the TV PCC Anthony Stansfeld and TVP chief constable Francis Habgood to Reading Borough Council on 30 October 2017 webcast and papers