Near-repeat mapping

Dr Andy Evans

[Fullscreen]

Repeat crimes

~31% to 76% criminals return to commit further crimes

Some rapidly return to the same victim (7-15% burglaries)

Some rapidly return to the same area (10-40% burglaries)

Some return to same area but over a longer time (hotspots)

Some crimes are 'slippery' - criminals repeat crimes slowly moving along streets

Townsley et al. (2003)

Near-repeats

Crimes clustered in space and time

Criminals return to an area over a few weeks until risk rises or reward drops

Might be thought of as an "optimal forager" strategy

If we can predict their return pattern we can prevent the crimes

Johnson & Bowers (2004); Bernasco (2008)

Analysis

Utilise Knox test

Work out distance between all event pairs in space and time

Put event pairs in a table where columns are time bins and rows space bins

 

 1 week2 weeks3 weeks4 weeks5 weeks
100m40301001
200m3520502
300m1010550
400m105020
500m05201

Johnson et al. (2007)

Significance

'Monte Carlo' distribution

Take dates and randomly attach to places of events

Gives distribution where these are independent

Build 1000 of these

Rank observed cell contents vs randomised to see if in top 5%

Johnson et al. (2007)

Results

Most near-repeat crimes fall with 400m of last crime

Most fall within a month of the last crime

Prediction : Prospective mapping

Divided area into cells (smallest possible given computing)

Events within 400m of cell picked

Each event given weight from distance in space from cell centre and in time from the present

Cell given aggregate weight for events

Bowers et al. (2004)

Prediction : Prospective mapping

Gave the locations of 78% of crimes within the next week compared with 51% based on previous models

Also identifying smaller potential target areas

Solution is statistically rigorous, but complicated for implementers

Johnson et al. (2007)

Prediction : Trafford method

Simple 400m buffers around crimes

Colour the buffers by weeks-elapsed for three weeks

Separate graph plots of crime densities for each area by time of day

Fielding & Jones (2012)

Prediction : Trafford method

Super-cocooning: target hardening and awareness raising in concentric forms around crimes

Trafford burglary down by up to 26%

Application

Determine critical scales

Prospective or Trafford methodology

Picks up repeat victimisation as well, but not general hotspots without time clustering

These are picked up by long term risk surfaces

Works best in high risk target- and environment-homogeneous areas