Crime figures: dodgy statistics and dishonest politicians

Labour say that violent crime is down by 41% since 1997, and the Tories say that it is up by 44%. They can't both be right. This post looks at the reliability of the statistics used by both sides, and at how politicians manipulate them.


Broken Britain: Tory figures


Chris Grayling started this debate when he claimed that official statistics showed crime soaring under Labour, followed by a rare rebuke from the Chairman of the UK statistics authority, as the basis on which those figures were collected changed in 2002: in particular, a police discretion as to whether or not to record a reported crime was removed.


One might have expected such an obviously misleading comparison to be written off and forgotten by Tory activists as another Grayling Gaffe. Not so. Today the Telegraph, and others, unveiled new research claiming to show the "true scale of how violent crime has grown under Labour", heiled as "vindication" for Chris Grayling by Tim Montgomerie over at Conservative Home.


There are two lazy assumptions being made here. First, it is assumed that the release of these statistics justifies Grayling's comments. They do not, as Chris Grayling had no basis on which to found his claims when he made his previous comments. A misleading use of statistics is a misleading use of statistics whether or not a subsequent study agrees with the doctored result.


Second, there is a disturbing lack of discussion concerning the reliability of the 44% figure. Out of the Times, the Sun, the Mail and the Telegraph, only the Telegraph is clear about where this figure comes from. The key words are "Statiticians in the Commons Library have used a previous Home Office estimate on the effect of the change in counting rules to estimate the impact on previous figures, had those rules been in place then".


In short, these figures are based on an old estimate of the change in the accounting rules. Given that before 2002 the police had discretion as to whether or not to record a crime, it is difficult to see how that estimate could have been anything but completely speculative. It follows that data derived from that estimate is also highly speculative. It is difficult to see how this data can be the "robust statistic" Grayling claims it is.


The third lazy assumption is that these figures are to be preferred to those put forward by Labour, derived from the British Crime Survey. The figures are completely irreconcilable: Labour's claimed 41% drop in violent crime would put violent crime at 59% of 1997 levels, and the Tory's claimed 44% rise would put violent crime at 144% of 1997 levels. In other words, the BCS says there is over twice as much violent crime as Tory figures do. As with most things, the secret is to ask people who do crime statistics for a living, and it seems very clear that Crimonologists like to use the BCS.


The only possible conclusion is that Mr. Grayling is using unreliable statistics to try and justify a misleading use of statistics. In the short-term, the Tories might bolster poll ratingsfrom those who don't read into the headlines. But in the long-term, it will be public trust in politicians that pays the price.