Measuring the dark figure of crime

This research, undertaken by researchers at the University of Manchester, used data made available via the Office for National Statistics (ONS) Secure Research Service (SRS), which is being expanded and improved with ADR UK funding.

Research summary

The ‘dark figure’ of crime (the number of offences which go unrecorded in crime statistics) has ‘severe consequences’ on crime prevention strategies. This research, using the Office for National Statistics’ (ONS) Crime Survey for England and Wales, explores the geographical inequalities between the dark figure and official police statistics. By using small area estimations, this research presents the first dark figure of crime map for England and Wales.

The dark figure is compared across local authority districts and middle layer super output areas – small areas designed to improve reporting of statistics. The results show the dark figure is larger in deprived and wealthy municipalities than in middle-class areas. The dark figure is also larger in low-cost housing areas with greater concentrations of unqualified citizens, immigrants, and non-Asian minorities. Identifying the areas where crimes are being unreported can inform policy makers and police forces to help design better interventions aimed at improving crime reporting rates.

Data used

The ONS Crime Survey for England and Wales (CSEW) has measured crime since 1981. Alongside police recorded crime data (collected by the Home Office), the survey is a valuable source of information for the government about the extent and nature of crime. The nationally representative survey measures crime by asking members of the public about their experiences of crime over the previous 12 months. In this way, the survey records all types of crimes experienced, including those that may not have been reported to the police. It is important that the voices of those people who have and have not experienced crime are recorded so that an accurate picture can be captured.

Methods used

The aim of this research was to trial several model-based small area estimation methods and evaluate the reliability of their dark figure of crime estimates for local authority districts and middle layer super output areas. The small area estimation methods evaluated include direct estimators, and traditional, spatial, temporal and spatial-temporal model-based approaches.

The research used the auxiliary variables available for all years of the CSEW. These were: ratio of males or females, average age, unemployment, house prices, income, population density, urban/rural classification (urban conurbations, small urban areas and rural areas), ethnic groups, qualifications, occupations and workday population density. A parametric bootstrap approach was used to assess the uncertainty of each SAE approach.

Research findings

Two small area estimation models were found to best estimate the dark figure of crime – the first for local authority districts and the second for middle layer super output areas. For local authority district estimates this was a spatial-temporal empirical best linear unbiased predictor (STEBLUP). For middle layer super output areas estimates this was a spatial empirical best linear unbiased predictor (SEBLUP). In both cases, the predictors were chosen for their ability to handle small and large data variance and to compute estimates of adequate precision. The chosen small area estimation models were used to create dark figure of crime maps of England and Wales for 2011 to 2017.