11 Jun 2011

Crime in England and Wales

A review of the way crime statistics are compiled and presented in England and Wales, published last week by the National Statistician Jil Matheson, provides a good analysis of the reasons why crime trends are poorly understood by the general public.  However it is highly doubtful whether the recommendations in the report will do much to reduce the amount of conflicting information circulated by politicians and the media.

Sources of crime statistics

Policymakers and the public currently have two main sources of statistics on crime: recorded crime, based on incidents reported to the police, and the British Crime Survey (BCS), an annual household survey.

Both sources have value for specific purposes, but provide different information.  For example the BCS is most useful for measuring long-term trends and, unlike police recorded crime, includes anti-social behaviour.  On the other hand the BCS sampling is not sufficient to support a local breakdown of crime data -- that is available only based on recorded crime.

As the report notes, the existence of the two sources can lead to 'cherry-picking' in the use of crime statistics.  Selective use of crime statistics is a popular tool both for politicians with an agenda to push and for journalists with papers to sell.

Public understanding has been further complicated by the recent development of crime mapping websites, which present local police recorded crime data directly to the public without most of the analysis and commentary that support formal statistical releases from the Home Office.

Report recommendations

The National Statistician's report recognises that the public do not yet have a complete or coherent picture of crime, and makes several recommendations:

  • transfer of responsibility for publication of crime statistics, including processing and compilation of both the BCS and recorded crime estimates, from the Home Office to the Office for National Statistics (ONS);
  • improvements to presentation of crime statistics to provide a clearer understanding of the overall picture, bringing sources of crime statistics together;
  • an independent advisory committee on data requirements and methodology, with a refocusing on the statistical quality of recorded crime data;
  • addition of statistics on crimes against 10-15 year olds;
  • more attention to gathering of information on crimes against businesses.
Joined-up communication of statistics from different sources, and the use of an advisory committee to bring in learning points from other recent initiatives on data transparency, are both common sense measures.   Although the Home Office has a wealth of experience in analysing crime statistics and trends, transferring more oversight to the ONS -- the gold standard for statistical expertise in UK government -- will add credibility and impartiality to the production of crime statistics.

However the National Statistician also recommends that responsibility for the collection and validation of recorded crime data from the police should remain with the Home Office.  The Home Office has so far made little headway in tackling the highly variable quality of police recording of notifiable offences or the extent to which the compatibility of data produced by different police services is skewed by local policing initiatives and priorities.

There is a broad public and political consensus that the police should be spending their limited resources on visible police work rather than administration.  The ONS may find it has limited influence to improve the quality of crime statistics at the input stage.

Crimes against businesses

The report notes in particular a lack of information on crimes against businesses.  The Home Office is considering plans to carry out a new commercial victimisation survey, following previous surveys in 1994 and 2002.  The National Statistician further suggests that questions on crime should be incorporated in more general surveys of business carried out by government departments.

There is an acute need for more of this type of information in the private sector.  Hazard models used by UK business to judge the geographic risk of crime losses are too often merely extrapolated from residential crime statistics, and can be unreliable in outlying industrial areas.  The report also highlights the difficulty of capturing statistics on fraud and online crime against business, as these categories of crime tend to be undetected or under-reported.

Crime indices and modelling

The report considers in some detail, but ultimately declines to recommend, the option of developing a 'crime index' for England and Wales.  This is an approach taken in Canada, where the national statistical authority produces an index of serious crime by weighting the sentencing criteria attached to each category of offence.

While insight into crime trends can only really be derived from analysis of the detailed data, a crime index could be helpful for secondary uses.  In the public sphere a recognised standard measure of crime would take some of the controversy out of competing technical interpretations of the statistics.  In the private sector, and particularly for risk management and insurance purposes, companies would benefit from a baseline national picture of crime.

There is of course substantial scope for commercial modellers to continue developing crime indices for business applications, provided the ONS can put in place the culture necessary to produce robust and transparent statistics.  UK crime modelling will be given additional impetus by fresher geodemographic data once outputs from the 2011 Census become available.

Crime mapping

The re-launch of the NPIA's national crime mapping website earlier this year, with local police record crime statistics (including anti-social behaviour incidents) displayed at street level, clearly creates additional complications for communicating understanding of crime trends and relative levels of crime between locations.  The National Statistician's report is tactful but raises concerns:
There is a clear distinction between the provisional, real time management information provided via the crime mapper, and the fully validated and quality assured National Statistics on recorded crime that comply with the Code of Practice for Official Statistics.... Appropriate signposting should be introduced between the official crime statistics and the crime mapping website to provide clarity on their respective status as National Statistics and management information.
While popular and technically well-executed, the national crime mapping website reflects a compromise between a political imperative to provide public statistics at the greatest possible level of 'granularity' and the need to protect the privacy of victims of individual crimes (per advice from the Information Commissioner).  The resultant mapping, which aggregates crimes on a given street to a central point on that street, is of casual interest but rather unsatisfactory as a tool for proper analysis or visualisation.

Crime mapping still has some way to develop in the UK, and ideally as the technology becomes more widely available the NPIA (or its successor agency) should consider an approach based on heat-mapping (as in this example from Oliver O’Brien at the Centre for Advanced Spatial Analysis).  In the meantime the national crime mapping website includes a page listing numerous additional apps for accessing recorded crime data.