Deceptive accuracy of crime and disorder point data

I write this entry in my blog for my own benefit as much as I am writing it for anyone else. I am grappling with the problem of collating and mapping my (ok the MPS) crime and disorder data in a way that it can be analysed with other data, such as census data, geodemographic data and composite indices like the Index of Multiple Deprivation.

Let me try to explain my ideas (which are yet to be fully rounded) about the spatial nature of crime and disorder. Police information systems are becoming increasingly precise in the recording of location which is of course a good thing. The MPS CRIS system that records reported crime provides northing and easting co-ordinates for the location of the crime, address of victim, witnesses, suspects, offenders and other relevant locations. I use these rather than postal addresses (which are also recorded) because as co-ordinates are not related to people in the same way addresses are this helps me with privacy and data protection issues. Even though the location is precisely recorded by a point one metre by one metre the means by which that location is derive is not necessarily so precise. Firstly, depending on the type of crime the exact location of its occurrence has varying degrees of accuracy. A crime that occurs in a residential premises can be reliably pin pointed but what if it happens in a tower block? A number of flats which are on top of each other will have the same point. A crime that happens out in the street, may have an imprecise location due to the memory of of the victim. And then there is the problem of crimes that are continuous, these are mainly crimes relating to drugs, weapons and stolen goods. The location of the offence is normally recorded at the location it is discovered by police, often in the police station after arrest creating unhelpful hotspots. Lastly on this first point, anti-social and disorder type crime tend to be spread over a series of locations because if recorded as a crime the full series will be recorded as one event. The second issue is; even if the crime location is unique and precisely known the way in which the CRIS system allocates location may only be as accurate as the post code (perhaps with street number, I need to enquire) which if this is a business or academic address like UCL may cover a large area of offices and land. So if a bike is stolen in the quad at UCL the co-ordinate location may be the University reception office.

All this creates a slightly deceptive accuracy of crime and disorder point data. I have no real problem with this fuzziness because the research regarding fear of crime, in particular in reference to people's perception of crime and disorder suggests that a crime event has a far greater spatial influence than a point one metre squared.