Solvability Analysis:
Increasing the Likelihood of detection in Completed,
Attempted and In-Progress Burglaries
Colin Paine & Barak Ariel
Data Set I
• All residential burglaries in Thames Valley
from 1st March 2010 to 31st October 2011
•11,769 full offence residential burglaries
•Excludes distraction burglary, aggravated
burglary and non-residential burglaries.
Data Set II
• All residential burglaries in Thames Valley
from 1st March 2010 to 31st October 2011
•2,537 attempted residential burglaries
•Excludes distraction burglary, aggravated
burglary and non-residential burglaries.
Data Set III
• All in-progress residential burglaries in
Thames Valley from1st November 2009 and 31st
October 2012 (i.e. Burglary emergencies).
•N = 6,119
•Includes aggravated, distraction and non
residential burglaries.
Data Sources
Collapsed four databases:
• “Socrates” – the forensic information database
• CEDAR – the crime recording database
• IMM – the crime investigation management
application
• Command and control – the incident resourcing
system (only this system used for data on inprogress offences).
•
(Datasets aligned using crime numbers &
cleansed to remove duplicates)
Limitations
Unable to study ‘Suspect Named’ variable (Greenberg 1970).
Unable to study intelligence as a variable
Insufficient capacity to study ANPR as a variable
Full and Attempted Burglaries
Variables
129 variables in the data set
57 useful variables, such as:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Time range of offence,
time to first officer attendance,
offence witnessed,
suspect seen,
suspect disturbed,
fingerprints recovered,
DNA recovered,
items left at scene by offender,
rural/town location,
media appeal completed,
called in as burglary in progress.
Burglary Offences by Catagory
20000
18029
18000
16000
14306
14000
12000
10000
8000
6000
4000
2000
362
61
Distraction
Burglary
Aggravated
Burglary
0
Non-Dwelling
Burglary
Dwelling
Burglary
43.7% of all burglaries are residential (dwellings)
Number of Burglary Dwelling
Offences (Completed Offences and
Attempts)
14000
12000
11769
10000
8000
6000
4000
2537
2000
0
Burglary Dwelling (Completed
Offence)
17.7% of Residential Burglaries are Attempts.
Burglary Dwelling (Attempted
Offence)
Frequency of Burglary Offences by
Time of Day
900
800
700
600
500
400
300
200
100
0
23:00 -…
22:00 -…
21:00 -…
20:00 -…
19:00 -…
18:00 -…
17:00 -…
16:00 -…
15:00 -…
14:00 -…
13:00 -…
12:00 -…
11:00 -…
10:00 -…
09:00 -…
08:00 -…
07:00 -…
06:00 -…
05:00 -…
04:00 -…
03:00 -…
02:00 -…
01:00 -…
00:00 -…
Most burglaries do not occur at night
3000
2500
2000
1500
1000
500
0
Frequency of Burglary
Offences by Day of week
Sunday
Saturday
Friday
Thursday
Wednesday
Tuesday
Monday
Description of Dataset
(Solved Residential Burglary)
Burglary Detection by Type
1600
1400
1344
1200
1000
800
600
522
400
200
32
20
Adult
Cautions
Juvenille
Final
Warnings
5
2
1
0
Charges
TICs
69.8% of detections are charges.
Juvenille
Youth
Adult
Reprimands Restorative Restorative
Disposal
Disposal
Solved Crimes
•Detected crime selected as
the definition of solved crime.
1600
1926 detected crimes
1400
(12.99%)
1200
•Removed secondary
1000
detections – 525 removed
800
•1401 detected crimes
(9.79%)
600
•4 methods of disposal
400
included; charges,
200
reprimands, final warnings,
0
cautions.
•Majority are charges
Number of Disposals For Residential
Burglary 01/03/10 to 31/10/11
1344
5
Burglaries
Charged
20
34
Juvenille Juvenille final Adult Caution
Reprimand
warning
Detection rates of Residential
Burglaries 01/03/10 to 31/10/11
15.00%
10.00%
10.68%
9.79%
Full burglary
Total
5.68%
5.00%
0.00%
Attempted
burglary
Attempted burglary is harder to solve than the full offence.
No. of burglaries
Number of days to detect Burglary
(excluding 10% and 90% Percentile)
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
Mean 41 days
Median 25 days
Mode
3 days
0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190
No. of days to detect
Over 96% of all detections are obtained in 180 days. 1 Burglary took 5 and half
years to solve.
Solvability Analysis
Data Quality Assurance and
Cleansing
•Random selection and review of quality of data
capture in 100 burglaries from the sample
•Review of missing data to determine whether it
will undermine internal validity.
•Data was pretty good!
Percent of cases where a footwear
mark is recovered
6.00%
5.00%
4.00%
3.00%
2.00%
1.00%
0.00%
4.93%
1.39%
0.27%
0.12%
Detected cases %
Undetected Cases %
Footwear mark Footwear mark
Recovered (full Recovered
burglaries)
(attempt
burglaries)
Compared percentage of marks recovered in solved sample and unsolved sample.
Percent of cases where DNA is
recovered
14.00%
12.00%
10.00%
8.00%
6.00%
4.00%
2.00%
0.00%
13.29%
9.72%
0.90%
DNA
Recovered (full
burglaries)
1.00%
Detected cases %
Undetected Cases %
DNA
Recovered
(attempt
burglaries)
Compared percentage of marks recovered in solved sample and unsolved sample.
Percent of cases where a fingerprint
is recovered
25.00%
20.00%
15.00%
10.00%
5.00%
0.00%
21.24%
13.89%
3.61%
1.00%
Fingerprint
Recovered (full
burglaries)
Detected cases %
Undetected Cases %
Fingerprint
Recovered
(attempt
burglaries)
Compared percentage of marks recovered in solved sample and unsolved sample.
Percent of cases with one or more
witnesses recorded
35
30
25
20
15
10
5
0
30.55
22.19
9.16
9.9
Detected cases %
Undetected Cases %
One or more
witnesses
recorded (full
burglaries)
One or more
witnesses
recorded
(attempt
burglaries)
Percent of cases initially reported as
a burglary in progress
25.00%
19.44%
20.00%
15.00%
10.00%
8.43%
5.97%
5.00%
Detected cases %
Undetected Cases %
3.86%
0.00%
Burglary in progress Burglary in progress
(full burglaries)
(attempt burglaries)
Compared percentage of burglaries in progress in solved sample and unsolved
sample.
Average (mean) time to
attendance of first officer
(minutes) excluding burglaries in
progress
300
260.51
250
200
150
178.17
144.94
116.22
100
Detected Cases
Undetected Cases
50
0
Time to attendance of Time to attendance of
first officer (excluding first officer (excluding
burglaries in progress) burglaries in progress)
Full Burglaries
Attempted burglaries
New solvability factors identified: media appeal, vehicle stolen in the crime,
anything left at scene by the offender, stolen property recovered, reported as a
burglary in progress, DNA and footwear marks recovered.
Percentage Frequency for Completed and Attempted
Burglaries
40.00%
35.00%
30.00%
25.00%
20.00%
15.00%
10.00%
5.00%
0.00%
The most powerful solvability factors occur the least often.
Completed Burglaries
Attempted Burglaries
Percentage of full burglaries by
number of solvability factors
70.00
60.00
57.69
50.00
40.00
30.00
24.88
20.00
9.41
10.00
4.53
2.35
0.88
0.20
0.04
4
5
6
7
0.00
0
1
2
3
Most burglaries have no solvability factors present.
Frequency of Factors Associated with Case Solution Where
No Solvability Factors are Present
25
21
20
15
12
10
8
7
6
5
5
2
1
0
Conspiracy
Charges
Suspect named CCTV Present
by victim
Unknown
Witness present Forensic Material Stolen Property
Present
recovered
372 solved residential burglaries had no solvability factors present.
Sample of 62 reviewed.
Intelligence
It is possible to predict the outcome of case dependent on presence of solvability
factors.
Predictive accuracy would be enhanced if suspect named variable were added,
conspiracy charges removed from sample and each variable were statistically
weighted.
Percentage of Burglaries with one or
more solvability factors by police area
70
65
60
55
50
45
The presence of solvability factors in burglaries is not uniform
Burglaries In Progress
• 5.59 Reported Burglaries in Progress in TVP
per day
• 6.11% result in an arrest at or near scene.
• 4% daytime and 7% nighttime.
Burglaries in Progress - Arrests Made
7000
6119
5745
6000
5000
4000
3000
2000
1000
374
0
Total number of
incidents opened
as burglary in
progress
No arrest made
Arrest Made
Just 6.11% result in an arrest in immediate response to the initial call
Percentage of Burglaries in Progress
with an arrrest by LPA
9
8
7
6
5
8.3
7.2
7.2
6.8
6.4
6.3
6.1
6.1
5.6
5.6
5.6
5.4
4.8
4.8
4
3
2
1
0.0
0
The arrest rate following burglaries in progress is not uniform across police
areas.
Percentage of Arrests Made at
Burglaries in Progress by the Minutes
to the Arrival of First Unit on Scene
25
20
15
10
5
0
The percentage likelihood of capture does not decrease with the passage of time.
Percentage of arrests made by
number of units attending
45
40
35
30
25
20
15
10
5
0
1 to 3
Units
4 to 6
units
7 to 9 10 to 13 13 to 15 16 to 18 19 to 21 More
units
units
units
units
units than 21
units
The optimum number of units to send to a burglary in progress is 19 to 21.
Predictor
Helicopter Attending
Odds of Making
an Arrest (OR)
66%*
Dog Unit Attending
Number of Units Attending
41%**
13%***
Number of Officers Attending
5%**
Sergeant Attending
Time to arrive
4%
0%
Conclusions
• Once interaction effects have been taken into account those factors that
most strongly predict arrest remain deployment of the helicopter and
the number of units deployed.
• For every additional officer the likelihood of making an arrest
increases by 6%.
• Deploying more units is more effective than just deploying officers –
better to send 2 single crewed units than 1 double crewed. For every
additional unit the odds of making an arrest increases by 13%.
• The attendance of dog units is correlated with arrest, but is masked by
the effect of number of officers.
• The impact of Sergeant attendance is not statistically significant.
• Burglary is hard to solve. Most burglaries have no solvability factors.
The most powerful solvability factors occur infrequently.
• The recovery of forensic material is more powerfully correlated than
expected, especially for completed burglaries.
• Solved attempted burglaries are more strongly correlated with burglaries
in progress, witnesses and the offender being disturbed.
• Solved full burglaries are more strongly correlated with forensic
recovery.
• It is possible to adopt a case screening model based on the presence of
solvability factors.
• This is now ready for an RCT to test a case screening approach.