Geographic Profiling in Australia

1
Geographic Profiling in Australia An
examination of the predictive potential of serial
armed robberies in the Australian Environment
The Seventh Annual International Crime Mapping
Research Conference March 31- April 3, 2004
Boston , Massachusetts
By Peter Branca
2
Introduction

• Outline of research in Australia
• Serial Armed Robbery Research
• Environment
• Data analysis
• Results
• Conclusions

3
Research In Australia
Few studies have been reported in relation to the
geography of serial crime in Australia.

• Kocsis Irwin (1997) examined serial rape, arson
  and burglary - support Canters Circle Theory
• Catalano (2001) examined the spatial behaviour
  patterns of serial robbery in Perth -
  criminological theories were helpful in
  predictions
• Kocsis et al (2002) assessed the Circle Theory
  for Geographic Psychological Profiling - rural
  town burglaries 50/50 Commuter and Marauder
• Spencer - (Unpublished) PhD research into spatial
  patterns of serial sex offences - both Commuter
  and Marauder behaviour exhibited by same
  offenders, utility of Dragnet and CrimeStat II
  investigated

4

• International Research -
• Focus on the home being central to serial crime
  locations.
• Centrography
• Journey to Crime
• Routine Activity Theory
• Circle Theory

5
Centrographic Analysis
Refers to the single location that is the
shortest distance to each crime site in the
series.
Centroid can be easily calculated using a GIS
6
Routine Activity and Journey to Crime (JTC)
Crime Activity
7
Prof. David Canters Circle Theory - Criminal
Range
Commuter
Marauder
87 of serial sexual offenders were found to be
Marauders
8
Computer Programs - JTC

• Rigel
• Dragnet
• Predator
• CrimeStat

9
Research Aims - Serial Armed Robbery in Victoria
To explore the predictive potential of Geographic
Profiling in relation to serial armed robbery in
Victoria, Australia. Utilising Journey to
Crime (JTC) and Centroid calculations to predict
the home location of serial offenders.
10
Australia - (a Quick Geography Lesson)
20 Million People
(25) 4.8 Million
11
(No Transcript)
12
Data
Source Victoria Police
28 serial armed robbers Total of 240 offences
(Mean of 8.6 crimes per series).
13
Data Analysis

• The vast majority of offenders were males 92.9.
• Average offenders age 31.5 years.
• The youngest offender was 20 and the oldest was
  47.
• Knife was the most common weapon used (34.6).
• Syringes exceeded firearms.

14
Data Analysis
Opportunity Theory of Travel
Distance v Amount Stolen
15
Data Analysis
Marauder or Commuter ?
12 of the 28 series, or 43, fitted the
description of the Marauder model (Group most
effective for Geo Profiling)
16
Data Analysis
Commuter - Street Offences 26 Marauder -
Milkbars (Convenience Store) 18, Other Shops
25, Service (GAS) Stations 17
17
Data Analysis
Commuter - Syringe 25 Marauder - Knife 47
18
Data Analysis
Other less significant comparisons were

• Offender Average Age (Mar - 33yrs, Com - 31yrs)
• Average Number of Offences (Mar - 11, Com - 7)
• Day of the Week (Mar - Sundays, Com - early in
  the week)
• Average Value Stolen (Mar - 1600, Com - 1200)
• Mode of Transport (Mar - Bicycle) Largely
  unknown

19
Conclusion

• It may be possible to differentiate between the
  Marauder and Commuter behaviour by examining the
  Offender and Offence Characteristics

20
Software used for analysis

• MapInfo
• CrimeStat
• MCi

21
Analysis Tools - MCi
MCi - MapInfo/CrimeStat Interface
Specially created for this research project
MapInfo
CrimeStat
CrimeStat
Analysis
View Results
22
Visual Analysis - Output
23
JTC Model Development
CrimeStat - Journey To Crime (JTC)

• Models can be based on either
• Mathematical function, or
• Empirically derived function

24
JTC Model Development
Mathematical function
The following methodology was used 1.
Calculate the Euclidean JTC distances using the
research data . 2. Group the distances into
appropriate distance intervals (ie range
bins). 3. Graphically display the data to assess
the central tendency and spread. 4. Generate
probability distributions using functions to
determine the model that best represents the
data. 5. Fit the frequency distribution model(s)
to calculate the appropriate parameters required
by CrimeStat. 6. Compare models to the original
data and select the best fit.
25
Culling the data for modeling

• To remove

• Commuters
• Outliers

26
Best Fit - Mathematical Models
27
Empirically Derived (Calibrated) Model Development
Developed with the research dataset
28
Prediction Analysis
JTC Predictions Models 1 2 (Mathematical
functions) Model 3 (Empirical Model -
Calibrated) Model 4 (Centrography)
29
Prediction Analysis
JTC Predictions (Model 1, 2 3)
30
Prediction Analysis
Model 4 (Centrography)
31
Prediction Analysis - Evaluation
Percentage of Activity Space (PAS) (Predicted
Area / Activity Space) 100 PAS (2.27/
23.27) 100 9.75
32
(No Transcript)
33
Analysis Results
Marauder Model
Entire Dataset
34
Conclusions - Research
Research Indicates

• Australian findings are consistent with
  international research
• The spatial behaviour of serial armed robbers is
  consistent to findings of other types of serial
  violent offenders
• It may be possible to identify a Marauder based
  upon offender/offence characteristics

35
Analysis Results - Further Investigation

• Identify a predictive relationship between
  offence characteristics and Marauder/Commuter
  behaviour patterns
• Research utilising larger data samples and other
  offence types should be investigated
• Need for JTC software programs with greater
  flexibility in relation to the mathematical
  models available

36
Thank You