P1259090118ltjMq

1
Crime Risk Models Specifying Boundaries and
Environmental Backcloths Kate Bowers
2
Introduction

• Crime Risk Model specification
• Boundaries
• Units of Analysis
• Environmental backcloth
• Land use
• Housing
• Accessibility
• Crime Risk Model Accuracy
• Determining map accuracy and utility
• Testing against chance models
• Future Projects
• CA modelling of risk
• Area linking models
• Multi-level models

3
MAUP- The Modifiable Areal Unit Problem

• 'the areal units (zonal objects) used in many
  geographical studies are arbitrary, modifiable,
  and subject to the whims and fancies of whoever
  is doing, or did, the aggregating.' (Openshaw,
  1984 p.3).
• Staggering number of different options for
  aggregating data
• Administrative boundaries
• Automatic non-overlapping boundaries
• Grids and polygons
• Two problems exist
• Scale- variation which occurs when data from one
  scale of areal unit is aggregated into more or
  less areal units.
• Aggregation- wide variety of different possible
  areal units

4
Burglaries per 100 households
Burglaries per 100 households
5
Hot beats
6
Traditional Hotspot Map
Yellow burglaries within two days Green
burglaries within 7 days
7
Prospective Map
Yellow burglaries within two days Green
burglaries within 7 days
8
Map Evaluation

• Map accuracy
• Number of hits
• Search efficiency (hits per unit area)
• Map practicality
• Number of hot areas
• Size of hot areas

9
Map Evaluation accuracy
2 days (26) 1 week (70) Area covered Search efficiency (2 day per km2)
Prospective Map 62 64 5.4km2 2.96
Traditional Hotspot Map 46 56 5.4km2 2.22
Beat Map 12 24 5.1km2 0.59
10
Map evaluation practicality
Prospective Map Traditional Hotspot Map
Mean area 12778m2 56502m2
Mean perimeter 377 m 925 m
No. of hotspots 79 19
Mean AP ratio 10 51
11
Friction surfaces/opportunity structure

• Opportunity structure (Flow enablers)
• Land use, distribution of houses, house type and
  tenure (see Groff La Vigne, 2001)
• Friction
• distance, topology (water, railways etc), crime
  prevention activity, social factors (affluence
  and cohesion)
• Facilitators
• Proximity to bus stops and roads (see
  Brantinghams)

12
Accounting for Background Method

• GIS- vector grid mapping- 50 metre grid squares
• Housing- OS Land Line
• Number of houses in each square
• Average area of houses
• Physical area of square used covered by housing
• Roads
• Number of sections of roads running through grid
  square
• Length of road running through square
• Classification of road (Major, Minor)
• Weighting squares
• Housing alone
• Roads alone
• Combinations

13
Mapping Layers Land Use and Crime Risk
14
Accuracy concentration curve for the promap
algorithm and chance expectation
15
Accuracy concentration curve for the KDE
algorithm and chance expectation
16
Accuracy concentration curve for the Beat map
generated for the rate of burglary per 1000
households
17
Accuracy concentration curve for the promap
algorithm (including both opportunity surfaces)
and chance expectation
18
Median mapping algorithm accuracy
Percentage of burglaries identified Percentage of burglaries identified Percentage of burglaries identified Percentage of burglaries identified Percentage of burglaries identified
10 25 50 75 90
Prospective Promap 1.39 5.09 14.39 30.89 55.36 Percentage of cells searched
PromapHouses 1.59 5.09 14.39 28.39 48.88 Percentage of cells searched
PromapRDs 1.39 4.89 13.39 29.09 52.57 Percentage of cells searched
PromapHousesRDs 1.59 4.59 12.59 29.39 56.35 Percentage of cells searched
Chance Simulation 95th Percentile 3.8 11.5 27.3 44.8 56.8 Percentage of cells searched
Simulation Mean 7.0 17.0 34.3 51.3 61.3 Percentage of cells searched
Retrospective KDE 2.09 6.59 16.89 34.87 59.04 Percentage of cells searched
Choropleth (concentration) 4.03 15.50 35.40 49.12 63.02 Percentage of cells searched
Choropleth (rate per area) 3.34 10.85 23.47 42.55 58.82 Percentage of cells searched
Choropleth (rate per homes) 6.41 17.62 31.70 50.02 69.11 Percentage of cells searched
19
Relative vulnerability of different housing types
April 1995-2000 Households burgled Total number of houses of type Prevalence rate Total number of incidents Incidence rate
Semi-detached 24915 201918 24.68 26689 26.44
Detached 4122 53364 15.45 4428 16.60
Terraced 23824 214023 22.26 26490 24.75
Flats 12184 103199 23.61 13515 26.19


20
Prevalence rates for different types of housing
in each quintile
April 95-00 Housing Type Housing Type Housing Type Housing Type
Prevalence rate Semi Detached Terraced Flat
Quintile 1 16.37 (6176) 10.32 (1793) 18.87 (498) 12.29 (318)
Quintile 2 20.39 (6179) 17.85 (1038) 18.44 (2485) 15.87 (1018)
Quintile 3 29.56 (5206) 27.46 (579) 21.31 (6150) 20.26 (1838)
Quintile 4 44.16 (3965) 57.83 (336) 21.95 (7751) 25.69 (2701)
Quintile 5 53.21 (3377) 71.29 (391) 25.91 (6924) 27.31 (6285)
21
Where next?- Modelling Street Network

• Examples of the accessibility measure used by
  Beavon et al. (1994)

• Quickest path analysis (connectivity of grid
  squares)

22
Where next?- Multi-level models

• Individuals Victims vs repeat victims
• Housing type
• MO of offence
• Victim characteristics
• Small area Cell or neighbourhood
• Accessibility
• Housing details
• Crime risk levels
• Larger area Census tract
• Social and demographic information

23
Where Next?- FCA Local density-dependent
transmission
Susceptible Infected Immune Unoccupied

• Possible outcomes
• Pathogen extinction (short infectious period)

prevalence
time
Slide by Joanne Turner (University of Liverpool)
24
Where Next?- CA Model Parameters

• Re-infection rates
• Different levels and lengths of immunity possible
• Target hardening/ Police patrolling
• Greater susceptibility in some than others
• Random short lived susceptibility
• Infection beginning from and re-occurring in
  different areas
• Random sparks
• Weak infectious models are possible
• Non-uniformity of contiguous cells

25
References

• Johnson, S.D., and Bowers, K.J. (forthcoming
  2007). Burglary Prediction Theory, Flow and
  Friction. In Graham Farrell, Kate Bowers, Shane
  Johnson and Michael Townsley (Eds.), Crime
  Prevention studies Volume 21, Monsey NY Criminal
  Justice Press
• Johnson, S.D., Bowers, K.J., Birks, D.J. Pease,
  K. (forthcoming 2007). Micro-Level Forecasting of
  Burglary The Role of Environmental Factors. In
  W. Bernasco and D. Weisburd (Eds) Crime and
  Place, in preparation.
• Johnson, S.D., McLaughlin, L., Birks, D.J.,
  Bowers, K.J. Pease, K. (forthcoming 2007)
  Prospective crime mapping in operational context.
  Home Office On-Line Report
• Bowers, K.J., Johnson, S.D., Pease, K. (2005).
  (Re)Victimisation risk, housing type and area a
  study of interactions Crime Prevention and
  Community Safety An International Journal 7(1),
  7-17
• Bowers, K.J., Johnson, S. and Pease, K. (2004)
  Prospective Hotspotting The Future of Crime
  Mapping? British Journal of Criminology 44 (5),
  641-658.
• Hirschfield, A.F.G., Yarwood, D. Bowers,
  K.(2001) Spatial Targeting and GIS The
  Development of New Approaches for Use in
  Evaluating Community Safety Initiatives in M.
  Madden and G. Clarke, (eds) Regional Science in
  Business, Springer-Verlag.

26
Nearest Neighbour Index Retrospective and
Prospective Methods