Police Recorded Crime statistics need some honest effort

Nigel Williams, 16 January 2014

The UK Statistics Authority has withdrawn its designation of ‘National Statistics’ from statistics derived from Police Recorded Crime. It is an exaggeration to describe the designation as a ‘gold-standard’. The report itself explains :

Designation as National Statistics should not be interpreted to mean that the statistics are always correct.

Having the designation is no guarantee that every procedure has been correctly applied or that every survey respondent has answered the questionnaire honestly. It is however, a higher standard of verification than is applied to most areas of public discourse. The designation represents a substantial effort to audit whether the statistics have been produced in accordance with a stringent code of practice. As an area of high-profile public interest, crime statistics are subject to extra public scrutiny. Whistleblowers such as James Patrick and Rodger Patrick have drawn attention to specific areas of concern that allow auditors to pinpoint breaches of the code of practice. Elsewhere, where the public is less concerned, suspected flaws can last several years before being investigated. A BIS report in 2013 into non-sampling error in the Trade Union Membership statistics, raised as an issue by ONS in 2011, attracted no headlines at all. That series has retained its designation as a National Statistic.

The statistics based on Crime Survey of England and Wales escaped most of the censure. The Statistics Authority’s report required the ONS crime team to jump some hurdles regarding documentation of the uses to which people put the output. They are confidently expected to meet those requirements by April 2014. Indeed it is the reliability of the estimates from the Crime Survey that allowed ONS statisticians to remark on the divergence between the two sources of data. Survey respondents were saying that they were telling the police about more crimes than the force appeared to be recording.

A second exaggeration would be to conclude that no crime statistics deserve any notice at all. Only this sort of aggregation saves us from the mistaken belief that our own individual experience is universally true. The lazy cliché ‘Lies, Damned Lies and Statistics’ ignores the truth that statistics are a form of language. People willing to lie with statistics are equally capable of lying with ordinary words, which are seldom subject to the same standards that are applied to the production of statistics. Several agencies are involved in producing recorded crime statistics – recorded by the police, collated by the Home Office, analysed by the ONS, audited by the UK Statistics Authority – and all will now be at pains to demonstrate that they are not the cause of the inaccuracies. The College of Policing has quickly announced a national training and accreditation package for crime registrars. They will not be short of advice. The designation report had this idea:

We suggest that ONS produce a diagrammatic representation showing how police recorded crime data are subject to bias and error at each stage of the recording process.

Each stage of the production has the responsibility to pass the whole truth on to the next agency. Now that those responsible have a record, it is hoped that they will be deterred from offending again.