1. technology and computing
2. hardware
3. computer
4. portable computer
5. palmtops and pdas

7th International Conference
on Evidence-Based Policing
The Birmingham Hot Spots Experiment
Operation Savvy
Dr Barak Ariel and Superintendent Jo Smallwood
With
Lawrence Sherman, Neil Wain, Cristobal Weinborn, Wendy Goodhill, Gabi Sosinski, Justice Tankebe and Orlee Yahalom
Background - Hotspots
• Most crime events concentrate in a small number of small areas called “Hotspots”
• "Law of concentrations of crime at place”
– (Weisburd, Telep, Braga & Groff 2010:167; Sherman et al 1989)
• Local deterrence (prevention) explains why crime would go down in hotspots
– (Ariel and Sherman, forthcoming)
• Most tests of hot spots policing report noteworthy crime and disorder reductions
– (Braga, Papachristos & Hureau 2012)
• The benefits of increased officer time spent in the hot spot plateau around 15 minutes
– (Koper, 1995; Telep , Mitchell & Weisburd, 2012)
What Don’t We Know?
• The replicability question
– Is the evidence extendable to England and Wales forces?
• The “how much” question
- Better dosage estimates in both treatment and control hotspots
- Tracking the number of visits rather then time-per-visit
• The “who” / “what” question
– Community policing
– Visible patrols
– Reassurance Policing
• The impact question
– What types of hotspots are (more) affected by hotspots?
– Are there more appropriate outcome variables than crime counts?
– How does hotspot policing affect community perceptions?
The Birmingham PCSO Hot Spots Experiments
Operations Savvy
Overall Research Design
• Randomised controlled trial over 12 months
• Assignment of all hotspots with 36 CFS or more to treatment vs. control conditions
• (intended) Intervention delivered by PCSOs only
• 3 X 15-minute patrols in all treatment hotspots, 3-10PM
• (intended) “business as usual” in control hotspots
• GPS locators on all front-line officers (radios), both PCSOs and PCs
• Surveys with residents and shop keepers
• Crime harm index as well as crime counts
Eligible Hotspots
• Birmingham South LPU
• 81 eligible hotspots defined as:
1. Maximum hotspot radius = 150 meters
2. Buffer zone/catchment area = 100 meters
3. Minimum distance between epicentres = 500 meters
4. “crimes” = street crimes, no shopping arcades / schools / hospitals / leisure centres
Treatment and Control Conditions
Police Community Support Officers
“PCSO are members of the support staff”
“Their primary purpose is to engage with the community and
offer greater public reassurance, for example:
1. Reporting and dealing with ASB
2. Reporting suspicious activity
3. Deterring juvenile nuisance
4. Visiting victims of crime
5. Reducing fear of crime
Source: www.communitysupportofficer.com
What Were Officers Asked to Do for Op Savvy?
COMMUNITY SUPPORT OFFICERS – in
Experimental Hotpots
• 3 visible patrols of 15 minutes each
during every shift (2PM-10PM), for 12
months
+
• Engage with the Community
COMMUNITY SUPPORT OFFICERS – in
Control Hotpots
• Engage with the Community
• Were not informed where hotspots are
POLICE CONSTABLES – in both
Experimental and Control hotspots
• Business as usual
 Patrol in vehicles
 Respond to CFS
 Fight crime
• Discretion as to ‘where’ and ‘when’ to
patrol (when not responding to CFS)
• NONE were informed where the
hotspots are, particularly the control
hotspots
Treatment Assigned – Community Officers to conduct
Directed Patrols of 15 minutes, 3 times per shift
1. “PCSO’s are directed by the relevant Problem Solving Sergeants
based on local knowledge, intelligence and dynamic risk
assessment” [NEW ROLE]
2. Assigned to engage, visit shops and talk to members of the
community [CLASSIC ROLE]
Tasking Sheet for each Treatment Patrol area
(example)
Tracking resources with personal GPS Locators
• GPS-enabled system
• “Sits” on Airways Systems
• Uses “Point in polygon” analysis
• Locates officers everywhere
• GPS-ping every 120 seconds
Results
Total Number of Crimes - before-after differences
Control
Exp't
0
-50
-100
-150
-200
-250
-238
-258
-300
8% Difference:
A failed experiment?
• Not at all
• The research shows that:
• All Crimes Are Not Created Equal
• All Hotspots Are Not Created Equal
• All Police Patrols Are Not Created Equal
The Devil is in the Detail…
Under The Bonnet in
Birmingham South
1. Crime Counts vs. Crime Harm Index
2. Effect is Conditional on the Type of
Hotspot
3. The Effect is Conditional on Dosage
4. Perceptions of Residents in
hotspots
1. Effect is Conditional on the Type of Hotspot
Effect of Police Patrol Depends on the Type of
Hotspot - Blocking
• As in previous hotspots experiments, we used a blockrandomisation design*
• “Blocking” is meant to create hotspot categories that are more
homogenous
• The blocking criterion often uses the number of events before
the experiment was conducted (the “Pre-RA stage”)
• This is mostly a statistical approach; the treatment ASSIGNED
is the same across all hotspots (WMP were not informed of the
hotspot blocks, to avoid treatment contamination)
• Is blocking justified? Were the blocks indeed different?
Mean n of CFS Per Hotspot
(baseline)^
70.00
60.00
60.29
50.00
42.60
40.00
27.52
30.00
20.00
10.00
High Level
Hotspots
Mid Level
Hotspots
Low Level
Hotspots
^ F=24.12(78,2); p≤.001
Ariel, B. & Sherman, L.W. (2014). Testing “local deterrence” by police patrols against a no-treatment control: a randomized, reverse-knockout experiment
Ariel, B., & Farrington, D. P. (2014). Randomized block designs. In Encyclopedia of Criminology and Criminal Justice (pp. 4273-4283). Springer New York.
Weisburd, D., & Gill, C. (2014). Block randomized trials at places: rethinking the limitations of small N experiments. Journal of Quantitative Criminology, 30(1), 97-112.
Systematic Observation with Google Street View
• The law of concentration of crime in place [and time] is heavily
supported, yet what makes hotspots hot is generally unclear
(cf. Brantingham & Brantingham et al.)
• Using Google Street view, we systematically observed the hotspots in
Birmingham (as well as Peterborough), and coded 18 variables, such as:
• crime attractors, physical disorder, bus stops, bars and nightclubs, youth centres,
main roads, shopping areas, high 'traffic' of people, schools, etc.
Ariel, B., Yahalm, O. and Weinborn, C. Systematic Observations of Hotspots of Crime and Anti-Social Behavior using Google Streetview: Application for Place Based criminology.
See also: Odgers, C. L., Caspi, A., Bates, C. J., Sampson, R. J., & Moffitt, T. E. (2012). Systematic social observation of children’s neighborhoods using Google Street View: a reliable and
cost‐effective method. Journal of Child Psychology and Psychiatry, 53(10), 1009-1017.
Index of Criminogenic Attributes (ICA)
Index of Criminogenic Attributes
(ICA)
2.64
High Level
Hotspots
1.84
1.93
Mid Level
Hotspots
Low Level
Hotspots
F(78,2) = 3.933; p<.05, Tukey’s HSD sig between H>M, H>L, but not M>L
What did this blocking procedure mean for
Op Savvy?
Before-After Differences of Crime –
Experimental and Control Hotspots Within Blocks
High
Medium
Low
120
100
80
60
40
20
0
-20
-40
-60
-80
-100
-120
94
Up by 157%
Down by 30%
Down by 53%
-63
-51
CONCLUSIONS SO FAR:
1. PCSOs foot patrol reduces overall crime recorded in hotspots by nearly 10%
2. But hotspots are not all the same; hottest hotspots are more criminogenic
3. Crime was reduced in High and Medium blocks by 40% but recorded crime “increased” in the Low block by 157%
2. The Dosage Question
Treatment Assigned vs. Treatment Delivered
• Can the “amount” of patrol explain the results?
• Identical dosage was prescribed (15 minutes per visit), yet dosage was not
applied consistently
• Officer-Level GPS Tracking NOW enables an enhanced capacity to track:
•
•
•
•
Total time spent within hotspots? PCs and PCSOs
N of visits to the hotpots? PCs and PCSOs
Who delivered the treatment? PCs and PCSOs
Dosage within types of hotspots? PCs and PCSOs
Wain, N. and Ariel. B. (2014) “The Tracking of Police Patrol”. Policing: A Journal of Policy and Practice.
Sherman et al (2014). “An Integrated Theory of Hot Spots Patrol Strategy: Implementing Prevention by Scaling Up and Feeding Back.” Journal of
Contemporary Criminal Justice: 1-28
Goddard, N. “A Direct Test of Patrols in Entertainment Hot Spots: Random Short Visits vs. Static Deployment”
Not just time per visit, but the number of visits
• 15 minutes as the optimal time per visit is intuitive and evidence-based (Koper 1995)
• Can the story also be about the number of visits per day?
• N of visits becomes important in a force that regularly patrols hotspots, often more
than 50 times per day
• Better observations must also take into account n of visits and/or minutes spent in
the no-treatment hotspots
• In fact, assuming the comparison hotspots are no-treatment areas is incorrect
ACROSS ALL HOTSPOTS OVER 12 MONTHS, what Did We
Learn From GPS-Trackers?
• PCSOs SPENT 54% more hours (3,421 vs. 1,562 hours)
• PCs SPENT 19% more hours (17,776 vs. 14,227 hours)
in Experimental Hotspots compared to Control Hotspots
• PCSOs VISITED 32% more times (4 vs. 3 patrols per day)
• PCs VISITED 7% more times (33 vs. 31 patrols per day)
the Experimental Hotspots compared to Control Hotspots
Linking Dosage to Crime Reductions
• Is overall more patrol associated with overall less crime 
this appears to be the case
• BUT is this pattern consistent across hotspot types? 
blocking provides an answer
• The differences seem to be centred on BOTH PCs AND PCSOs:
PCs patrolled the hottest treatment hotspots nearly 50% more times and patrol
the coldest treatment hotspots 50% fewer times, ccompared to control
hotspots
• PCSOs spent many more hours in Medium and Low Hotspots
Comparing Treatments at Experimental and
Control Hotspots with Officer-Based GPS trackers
(12 months of data, both PCs and PCSOs)
N of VISITS
TOTAL HOURS
T
C
Difference
PCSOs at High Level Hotspots
5.13
3.65
29%
PCSOs at High Level Hotspots
PCSOs at Medium Level Hotspots
4.81
3.35
PCSOs at Low Level Hotspots
3.75
PCs at High Level Hotspots
T
C
Difference
685
503
27%
30%
PCSOs at Medium Level Hotspots 1576 529
66%
2.46
34%
PCSOs at Low Level Hotspots
1151 530
54%
52.52
27.75
47%
PCs at High Level Hotspots
4321 3001
31%
PCs at Medium Level Hotspots
41.79
32.65
22%
PCs at Medium Level Hotspots
8750 4656
47%
PCs at Low Level Hotspots
21.12
30.43
-44%
PCs at Low Level Hotspots
4705 6570
-39%
CONCLUSIONS SO FAR:
1.
PCSOs foot patrol reduces overall crime recorded in hotspots by nearly 10%
2.
But hotspots are not all the same; hottest hotspots are more criminogenic
3.
Crime was reduced in High and Medium blocks by 40% but recorded crime “increased” in the Low
block by 157%
4.
PCSOs spent overall more time and visited more often Experimental Hotspots than Control Hotspots
5.
PCs patrolled the hottest treatment hotspots nearly 50% more times and patrol the coldest treatment
hotspots 50% fewer times AND 39% fewer hours, compared to control hotspots
2. Can Perceptions of Residents in Hotspots
Explain the results?
What Is the Perceived Impact of PCSOs Foot Patrol?
• Door-to-Door Survey in both Treatment and Control Hotspots
• N = 1,464 (Response rate = 74%)
• 90 items, four dimensions:
• Views on policing (satisfaction), Perceptions of legitimacy, Collective efficacy, Fear of crime
• Stratified Sample of:
 Residents
 Shop Keepers
• Experimental and Control Respondents are not statistically different in terms of socio-demographic and
other background characteristics
Ariel, B., Wain, N., Sosinski, G. and Tankebe, J. The Effect of Hotspot Policing on Collective Efficacy, Police Legitimacy and Fear of Crime: A Randomized Controlled Trial
Key Messages From surveys 1
• PERCEPTIONS ALONE DO NOT SEEM TO EXPLAIN THE EFFECT OF PCSOS
ON CRIME
• Satisfaction with police performance was generally consistent across experimental
and control hotspots
• perceptions about crime levels in experimental hotspots have NOT changed
(compared to control hotspots)
• Response time (especially to emergency CFS) is perceived to be the same
• In terms of Visibility of officers, respondents reported seeing more patrols in the
control hotspots (!)
Key Messages From surveys 2
• No overall differences in terms of neighbourhood cohesion
e.g. “This is a close-knit neighborhood”
• No overall differences in terms of legitimacy of police
e.g. “The police usually act in ways that are consistent with my own ideas about what is
right and wrong”
• Collective efficacy was lower in the Experimental Hotspots
e.g. “some children were spray-painting graffiti on a local building, how likely is it that
your neighbours would do something about it?”
CONCLUSIONS SO FAR:
1.
PCSOs foot patrol reduces overall crime recorded in hotspots by nearly 10%
2.
But hotspots are not all the same; hottest hotspots are more criminogenic
3.
Crime was reduced in High and Medium blocks by 40% but recorded crime “increased” in the Low
block by 157%
4.
PCSOs spent overall more time and visited more often Experimental Hotspots than Control Hotspots
5.
PCs patrolled the hottest treatment hotspots nearly 50% more times and patrol the coldest treatment
hotspots 50% fewer times AND 39% fewer hours, compared to control hotspots
6.
Police may increase visible patrols, but the public does not seem to be aware of these changes
7.
Satisfaction with police performance is not changed by additional PCSO patrols
8.
Hotspot Policing seems to reduce the Collective Efficacy of the neighbourhood
9.
Legitimacy of policing is not eroded with hotspot policing
3. Police-Generated vs. Victim-Reported Events
Outputs, outcomes and the source of call
• Summing up all crimes is misleading
• Some events are “police-generated” and these are largely dependant
on proactive policing strategies and tactics
• If PCSOs are directed to report shoplifting, shoplifting will go up
• If PCSOs are directed to report ASB, ASB will go up
• This is not the same for burglaries, homicide, assaults or robberies
Comparing Experimental and Control
Hotspots – Selected Crime Categories
Shoplifting
34
ASB
2
Against Person
-6
Against Property
-43
-60
-40
-20
FEWER CRIMES IN EXPERIMENTAL
0
20
MORE CRIMES IN EXPERIMENTAL
40
Victim
-Reported Crimes:
First thing first
–
separate
police
generated
Before-After Differences
events
from victim-reportedexperimental
crimes
control
0
-20
-40
-60
-80
-100
-96
-120
-140
-160
-180
62 fewer crimes -> 40% overall reduction
-158