skip to primary navigation skip to content

Department of Geography


Combining police perceptions with police records of serious crime areas: a modelling approach

Research Team: Robert Haining and Jane Law


This paper investigates the location of serious crime neighbourhoods in Sheffield, England in 1998 based on two sources of data: senior police officer perceptions of where such neighbourhoods are located and the evidence contained in the Police’s own database of recorded crime. We report the results of modelling these two spatial distributions using 2001 census data on output areas. We also demonstrate how expert knowledge might be combined with the evidence contained in large geo-referenced databases. We conclude with an evaluation of the use of model-based approaches to identifying high crime areas. The purpose of the paper is to go beyond descriptive mapping of crime data and to explore how to combine in a formal way, different types of knowledge in the analysis of crime and disorder maps.


policeoerception fig 1

Figure 1: (a) Map of the 1998 police defined HIAs (PHIAs): 47 output areas; (b) Map of the empirically defined HIAs (EHIAs): 47 output areas; (c) Overlay of figures 1(a) (PHIAs) and 1(b) (EHIAs).

police perception fig 2

Figure 2: (a) Map of the count of serious offences in 1998 by output area; (b) Map of the count of offenders (all offences) in 1998 by output area; (c) Map showing the rank order of output areas using the methodology for defining EHIAs (1= highest score); (d) Cumulative frequency distribution of the combined standardized scores used to specify which output areas are EHIAs.

police perception fig 3

Figure 3: (a) Map of the 47 HIA output areas obtained from the repeated measurements model (four covariates retained); (b) Model probabilities (posterior means) for output areas to be classified as in an HIA by the repeated measurements model.

Further reading

  • Haining, R.P. and Law, J. (2006). “Combining police perceptions with police records of serious crime areas: a modelling approach.” Journal of the Royal Statistical Society, Series A (to appear in 2007).