Thin pickings from the violent crime statistics

Just how likely is it that you – or your teenage son – will be attacked by a stranger in the street? And have the chances of this happening increased or diminished in the past 20 years?

Simple questions, but with no simple answers, as a charity that opposes the “walk on by” culture and stands up for the rights of witnesses has been discovering.

Witness Confident, set up last September, believes that if the value of eye-witness evidence was taken more seriously, more street crimes could be solved. Too much attention is paid to the rights of defendants, it argues, too little to encouraging people who observe crimes to bear witness.

Its report, Evidence Lost,  drew on data from the British Crime Survey to try to reach some conclusions about the frequency of stranger violence on Britain’s streets. In 2008-09 there were 852,000 incidents in which people were attacked by strangers, an increase of 11 per cent on the 766,000 recorded in 2007-08. If we add muggings, the total rises to 1,227,000 – considerably higher than the total of domestic and acquaintance violence in the same year (984,000).
Guy Dehn, a barrister and one of the leaders of Witness Confident, was immediately struck by a change in the figures from the same table in the previous year’s BCS, which recorded the figure for stranger violence as 744,000, not 766,000. No explanation was given for this change, but the Home Office explained that it resulted from a change in the estimated numbers of young males in the population.
The BCS measures the incident rates for different crimes from a sample, then multiplies them by the population estimates to come up with its total numbers of crimes. Young males are the most likely victims of stranger violence, so the change in the number of young males increased the multiplier for this category of crime, leading to the change from 744,000 to 766,000. This was not explained in the 2008-09 BCS.
The rise of 11 per cent between 2007-08 and 2008-09 (766,000 to 852,000) struck Mr Dehn as significant, but statistically speaking it is not. Because the sample is relatively small, the confidence intervals for the estimates are wide.  Even if the number for 2007-08 had remained at 744,000, the rise to 852,000 would not have been significant, because the confidence intervals would have overlapped – 744,000 (CI 652,000-837,000) and 852,000 (CI 735,000-969,000).
The Home Office therefore regards stranger violence as “stable”, with no significant change even when measured over the lengthy period between 1995 to 2008-09. Violence by acquaintances has, over that period, shown a decline of 62 per cent, and domestic violence a decline of 70 per cent – both statistically significant.
Mr Dehn felt the way the Home Office presented the data as lacking clarity, and asked the UK Statistics Authority for guidance.  As a result of the intervention of the authority and of the National Statistician, the Home Office undertook to signpost more clearly the effect of changes in population estimates on the BCS, and to clarify the presentation of a table in a 2006 report that Witness Confident had mistakenly concluded was about witness satisfaction with how incidents of violence had been handled by the police. (In fact, it did not represent satisfaction, but the percentage of such crimes witnessed.)
The figures for violent crime in the BCS show a downward trend, at odds with criminal statistics compiled by the police. But violence by strangers has not shown a decline, in contrast with the US, where stranger violence measured by the National Crime Victimisation Survey declined more rapidly than other forms of violent crime between 1980 and 2005.
Despite the deficiencies of the BCS in measuring violence against the person, there is still official reluctance to acknowledge that police recorded crime has any role to play. Changes in recording standards in 2002 are generally held to have made comparisons before and after that year invalid. But little serious attempt appears to have been made to estimate the effects of these changes beyond a single year. The Home Office concluded that the impact on violent crime was to increase it by 20 per cent in 2002-03. But using this adjustment, as the Conservatives did before the election to estimate violent crime trends, is frowned upon.
In February Philippa Stroud, Executive Director of the Centre for Social Justice, wrote to Sir Michael Scholar asking for clarification. Ms Stroud was the Conservative candidate for Sutton and Cheam in the General Election, and is now political adviser to Iain Duncan-Smith, the Work and Pensions Secretary. Was it valid, she asked, to use the 20 per cent figure to calculate longer-term trends in violent crime reported by the police?
The answer she got, from National Statistician Jil Matheson, was hardly helpful. The 20 per cent adjustment related, she said, only to a single year: “No similar estimate has been made for subsequent years as changes continued to be bedded in”. Why not, one wonders? The time and trouble spent on collecting and collating figures is largely wasted if no information about trends can be extracted. Ms Stroud’s estimate of a 40 per cent increase in violent crime between 1998-99 and 2008-09 was, she was told, invalid. This amounts to saying that you might as well chuck away the police recorded crime figures as worthless, since they cannot any longer be used to construct a time series.
Ms Matheson defends the BCS as an alternative and it is indeed a large and well-conducted survey. But in respect of crimes overwhelmingly suffered by a small minority, such as stranger violence, it has severe limitations. Young men aged between 16-24 are more than five times as likely to be the victims of stranger and acquaintance violence as the average person in the street. But the BCS sample in this age group is just 3,850. The BCS’ “youth boost”, designed to collect more data, increased the sample to almost 5,500 in 2008-09, but this larger sample was used for analysis of drug use, not crime, and has anyway been abandoned for 2009-10.
Until 2009 the BCS excluded under-16s, which probably meant that woundings and assault were being undercounted by about 380,000 among those aged 11-15, the Home Office estimated (Crime in England and Wales, 2004-05, table 3.01). All told, the crime statistics tell us much less than we would like to know about violent offences, particularly those carried out by strangers.
The public perception is that crime has increased, not diminished. Officials wring their hands at the public’s refusal to accept trends derived from the BCS, but in the circumstances such a refusal appears quite logical. If an increase of 11 per cent in a single year is not statistically significant, consumers of the statistics are entitled to ask what real value they are. And if they turn to crimes recorded by the police as an alternative, they are told that’s invalid. Small wonder they remain unpersuaded.
Update, June 2 2010
The UK Statistics Authority has recently published a review, Overcoming Barriers to Trust in Crime Statistics, to which I should have referred. In its press release it quotes Sir Michael Scholar, chair of the authority: "Having two different sources can undoubtedly cause confusion but the answer is not to change either of them fundamentally. The two sets of statistics throw different lights on the incidence and experience of crime,m and we need both of them."
But the trouble, surely, is that the light they shed is not different but contradictory. Until that is resolved, public confidence is unlikely to be regained, however many of the authority's sensible suggestions are adopted. The full list is available here.