"Knife Crime" In The Media

Britain is in the grip of a knife crime epidemic. Isn't it? That is certainly the impression one gets from the media: every week seems to bring new stories of stabbing and murder among city youth. The latest victim is 16-year-old Joseph Lappin, who was set upon by a gang outside a Liverpool youth club on Monday evening, and knifed to death in an apparently motiveless attack. A friend was also stabbed twice and seriously wounded, while another friend received minor injuries in the attack by up to fifteen youths.

Then, yesterday, the Home Office published its latest quarterly update of the statistics for police-recorded crime and the British Crime Survey. The figures appear to show a big rise in violent crime compared to the same period last year: a 22% rise in "most serious violence against the person" and a 28% increase in the number of attempted murders with a knife. Some newspapers reported this as a clear rise in crime: the Telegraph gave the story these headlines:

Violent crime increases by a fifth as police fail to keep proper figures
Violent crime has jumped by fifth because police forces have failed to keep proper figures for more than a decade, the Government has admitted.

This is a very misleading headline, and the impression it gives is simply not true. You have to read further to find out what the figures really mean: violent crime has not jumped by a fifth, the number of recorded violent crimes has risen, but this is largely because many police forces have changed the way they categorise these offences. Similar headlines appeared in the Times ("Police fail to record crime properly, as violence rises 22%"), the Mail ("Violent crime up 22%") and the Express ("Violent crime soars"). The Guardian headlines give a truer picture:

Row over police statistics as recount leads to 22% 'rise' in worst violence
Apparent increase 'due to misinterpretation of rules'

The Independent too emphasises the confusion rather than the apparent rise ("Violent crime underestimated for 10 years"). In other words, your perception of the level of violence in Britain this morning is probably heavily influenced by your choice of newspaper. So what's the truth? More detailed analysis, reported in the Guardian but missing from the other papers, shows that the real rise in serious violence against the person was only around 5%.

Another problem with today's stories is common to all the papers I have seen: emphasis on the percentage change (or "relative risk") rather than the actual numbers (or "natural frequencies"). This has the effect of making the picture seem far worse than it really is, because a rise of 28% in the number of attempted murders by knife sounds far more serious than an increase from 50 incidents to 64, yet both are accurate ways of saying the same thing. We know from research on health statistics that people in general are very bad at understanding relative risk, yet the media persist in using whatever makes the story seem most dramatic.

There are many other reasons to be careful when trying to understand knife crime. The first point to bear in mind is that it is difficult to define "knife crime". The term encompasses a range of behaviours, from actual use (eg in stabbing) via threatened use (eg in mugging) to knife-carrying. None of these is simple to quantify, however. Some offences recorded as involving a "sharp instrument" might actually refer to a screwdriver, broken bottle or glass, not a knife. Similarly, offences listed as "threatening another person with a weapon" might involve sticks, rocks or other blunt objects as well as knives. Knife-carrying is even harder to define, since certain types of knife may be carried legally if the carrier has a good reason (eg for work, or a hobby such as fishing), adding a major subjective element to the definition.

The best attempt so far to disentangle the problem of knife crime in the UK has been conducted by the Centre for Crime and Justice Studies (CCJS) at King's College London. Their 2007 report ‘Knife Crime’ A Review of Evidence and Policy uses official police statistics as well as data from the British Crime Survey, the Offending, Crime and Justice Survey, and the MORI Youth Surveys carried out for the Youth Justice Board. All these surveys have strengths and weaknesses, but taken together they give little reason to believe that knife carrying has changed much since 2002 or that knife use has increased since 1997. How did the media report these reassuring findings? By now you can probably guess. The Times ran with the headline "Knife crime doubles in 2 years". This claim was justified in the story as follows

The full extent of Britain’s violent crime epidemic, which yesterday claimed the life of another teenager, is revealed in shocking new figures that show the number of street robberies involving knives has more than doubled in two years. Attacks in which a knife was used in a successful mugging have soared, from 25,500 in 2005 to 64,000 in the year to April 2007. The figures mean that each day last year saw, on average, 175 robberies at knife-point in England and Wales – up from 110 the year before and from 69 in 2004-5.

I have scoured the report but have been unable to find the source of these figures. Perhaps they were found in an early draft or press release, but they certainly do not appear in the final report. Here's what it does say about knife use and mugging:

Mugging figures have been ignored in the analysis because of low sample sizes and recent changes in the definition of mugging.

A footnote explains this in more detail: in 2003-4 there were 19 knife-point muggings reported in the BCS, and in 2006-7 there were 45. Remember that the BCS has in excess of 40,000 participants per year, so these frequencies are a tiny proportion of the total sample. To extrapolate from these figures to the whole population is simply ridiculous, yet that seems to be what the Times did. They cherry-picked the worst figures from a generally positive report and spun them to create the scariest possible headline.

The reason all this matters is that sensationalist reporting is actually making the problem worse. The CCJS report shows that those young people who do carry weapons almost always say they do so for self-defence. If they believe that others are likely to be carrying, they are more likely to do so themselves. When the Mail screams that "Shock figures reveal no part of Britain is safe as knife violence spreads EVERYWHERE" they help to create the problem they purport to abhor. Here's another (even crazier) example: "Britain on alert for deadly new knife with exploding tip that freezes victims' organs". It's about a knife that is sold in America, designed to kill sharks and bears. Is there any reason to believe such knives are being carried in Britain? Any evidence at all? No.

Last year, one of my undergraduate students conducted a research project on weapon-carrying among the youth of Liverpool. It was a small study, which is why I am blogging about it rather than trying to publish it, but it was well conducted and had some very interesting findings. She managed to recruit 57 young people, most of whom were excluded from school, in an area notorious for crime (including the murder of Rhys Jones a few weeks earlier). Because we were concerned that self-report questionnaires may lead to under- or over-reporting, we used a randomising element to make it very clear to the participants that their responses would be anonymous. Here's how it works. The questions are phrased as YES/NO items, for example "Have you ever carried a knife on the streets" or "Have you ever used a knife at school?". Before responding to each question, the participants are asked to flip a coin. We tell them "If the coin comes down heads, say YES to the question. If it comes down tails, tell us the truth". This was explained very carefully so that they all understood that nobody, not even us, would be able to tell if they had actually carried out the behaviour or not. We then asked about knife and gun carrying and use, at school and on the streets ("use" in this context could mean use to threaten as well as use to stab or shoot). What we wanted to know was whether we would get more YES responses than the 50% that would be predicted by chance (see footnote for an example). The answer was no. A series of binomial tests revealed no significant deviations from chance.

We also asked for the participants' perceptions of weapon carrying. We gave them a visual analogue scale, labelled from 0% to 100% in 10% steps and anchored at either end with "Nobody" and "Everybody", and asked them to rate what proportion of people their age in their area they believed to have carried or used knives or guns at school and on the street. Here is what we found:

Weapon type	Location	Mean % (SD)
Guns	Carried at school	17.54% (23.16)
	Carried on the street	29.82% (20.39)
	Used in school	3.68% (7.22)
	Used on the street	19.29% (17.71)
Knives	Carried at school	30.17% (23.33)
	Carried on the street	46.31% (26.56)
	Used in school	12.80% (16.55)
	Used on the street	34.38% (25.35)

In other words, these young people believe that nearly half of their peers have carried knives and that nearly a third have carried guns. They believe that over a third have used a knife on the street and that a fifth have used a gun. These figures are far in excess of the actual percentages, which we have seen were not significantly different from zero. Why do they have such a negative view of the world? I believe the media must take a large part of the blame.

Here is the take-home message. Maybe you believe you need to carry a knife, because you think nowhere is safe. You think that everyone else has a knife, so you'd better have one too. You are wrong. Leave the knife at home.



Footnote: Imagine we had surveyed 100 people and obtained a 60% yes rate for kinfe carrying. 50 of these would have been directed to say yes by the coin, while 10 of the remaining 50 people were giving us a true yes. This would suggest that 20% of respondents had actually carried a knife. This is significantly higher than chance, according to a binomial test (p=0.0284). In our study, with 57 participants, we would have needed 36 yes responses to obtain a p value below 0.05, which would have meant approximately 30% of the sample had given a real yes. This is close to the population rate of knife carrying according to the MORI Youth Surveys, but because of the nature of our sample we expected higher frequencies than this.