Mr. Friend is a
crime analyst
with the Santa
Cruz, California,
Police
Department.
Predictive Policing: Using Technology to Reduce Crime
By Zach Friend, M.P.P.
4/9/2013
Nationwide law enforcement agencies face the problem
of doing more with less. Departments slash budgets
and implement furloughs, while management struggles
to meet the public safety needs of the community. The
Santa Cruz, California, Police Department handles the
same issues with increasing property crimes and
service calls and diminishing staff. Unable to hire more
officers, the department searched for a nontraditional
solution.
In late 2010 researchers published a paper that the
department believed might hold the answer. They
proposed that it was possible to predict certain crimes,
much like scientists forecast earthquake aftershocks.
An “aftercrime” often follows an initial crime. The time and location of previous criminal activity helps to
determine future offenses. These researchers developed an algorithm (mathematical procedure) that
calculates future crime locations.1
Equalizing Resources
The Santa Cruz Police Department has 94 sworn officers and serves a population of 60,000. A
university, amusement park, and beach push the seasonal population to 150,000. Department personnel
contacted a Santa Clara University professor to apply the algorithm, hoping that leveraging technology
would improve their efforts. The police chief indicated that the department could not hire more officers.
He felt that the program could allocate dwindling resources more efficiently.
Santa Cruz police envisioned deploying officers by shift to the most targeted locations in the city. The
predictive policing model helped to alert officers to targeted locations in real time, a significant
improvement over traditional tactics.
Making it Work
The algorithm is a culmination of anthropological and criminological behavior research. It uses complex
mathematics to estimate crime and predict future hot spots. Researchers based these studies on
In Depth
Featured Articles
- IAFIS Identifies Suspect from 1978 Murder Case
- Predictive Policing: Using Technology to Reduce
Crime
- Legal Digest Part 1 - Part 2
Search Warrant Execution: When Does Detention Rise to
Custody?
- Perspective
Public Safety Consolidation: Does it Make Sense?
- Leadership Spotlight
Leadership Lessons from Home
Archive
- Web and Print
Departments
- Bulletin Notes - Bulletin Honors
- ViCAP Alerts - Unusual Weapons
- Bulletin Reports
Topics in the News
See previous LEB content on:
- Hostage Situations - Crisis Management
- School Violence - Psychopathy
About LEB
- History - Author Guidelines (pdf)
- Editorial Staff - Editorial Release Form (pdf)
Patch Call
Known locally as the
“Gateway to the Summit,”
which references the city’s
proximity to the Bechtel Family
National Scout Reserve. More
The patch of the Miamisburg,
Ohio, Police Department
prominently displays the city
seal surroun.
Mr. Friend is acrime analystwith the SantaCruz, Califo.docx
1. Mr. Friend is a
crime analyst
with the Santa
Cruz, California,
Police
Department.
Predictive Policing: Using Technology to Reduce Crime
By Zach Friend, M.P.P.
4/9/2013
Nationwide law enforcement agencies face the problem
of doing more with less. Departments slash budgets
and implement furloughs, while management struggles
to meet the public safety needs of the community. The
Santa Cruz, California, Police Department handles the
same issues with increasing property crimes and
service calls and diminishing staff. Unable to hire more
officers, the department searched for a nontraditional
solution.
In late 2010 researchers published a paper that the
department believed might hold the answer. They
proposed that it was possible to predict certain crimes,
much like scientists forecast earthquake aftershocks.
An “aftercrime” often follows an initial crime. The time and
location of previous criminal activity helps to
determine future offenses. These researchers developed an
2. algorithm (mathematical procedure) that
calculates future crime locations.1
Equalizing Resources
The Santa Cruz Police Department has 94 sworn officers and
serves a population of 60,000. A
university, amusement park, and beach push the seasonal
population to 150,000. Department personnel
contacted a Santa Clara University professor to apply the
algorithm, hoping that leveraging technology
would improve their efforts. The police chief indicated that the
department could not hire more officers.
He felt that the program could allocate dwindling resources
more efficiently.
Santa Cruz police envisioned deploying officers by shift to the
most targeted locations in the city. The
predictive policing model helped to alert officers to targeted
locations in real time, a significant
improvement over traditional tactics.
Making it Work
The algorithm is a culmination of anthropological and
criminological behavior research. It uses complex
mathematics to estimate crime and predict future hot spots.
Researchers based these studies on
In Depth
Featured Articles
- IAFIS Identifies Suspect from 1978 Murder Case
3. - Predictive Policing: Using Technology to Reduce
Crime
- Legal Digest Part 1 - Part 2
Search Warrant Execution: When Does Detention Rise to
Custody?
- Perspective
Public Safety Consolidation: Does it Make Sense?
- Leadership Spotlight
Leadership Lessons from Home
Archive
- Web and Print
Departments
- Bulletin Notes - Bulletin Honors
- ViCAP Alerts - Unusual Weapons
- Bulletin Reports
Topics in the News
See previous LEB content on:
- Hostage Situations - Crisis Management
4. - School Violence - Psychopathy
About LEB
- History - Author Guidelines (pdf)
- Editorial Staff - Editorial Release Form (pdf)
Patch Call
Known locally as the
“Gateway to the Summit,”
which references the city’s
proximity to the Bechtel Family
National Scout Reserve. More
The patch of the Miamisburg,
Ohio, Police Department
prominently displays the city
seal surrounded by an outline
of the state. More
LEB Updates
Subscribe to receive monthly updates listing articles featured on
this website. Details
Home • Stats & Services • Reports and Publications • LEB •
2013 • April • Predictive Policing: Using Technology to Reduce
Crime
FBI — Predictive Policing: Using Technology to Reduce Crime
http://www.fbi.gov/stats-services/publications/law-enforcement-
bulletin/...
1 of 4 6/3/2013 3:38 PM
5. information that officers inherently know. For example, when
people are victims, the chance that they or
their neighbors will be victimized again increases. Offenders
criminalize familiar areas. There are
detectable patterns associated with the times and locations of
their crimes.
Using an earthquake aftershock algorithm, the system employs
verified crime data to predict future
offenses in 500-square-foot locations. The program uses
historical information combined with current
data to determine patterns. The system needs between 1,200 and
2,000 data points, including
burglaries, batteries, assaults, or other crimes, for the most
accuracy. Santa Cruz, averaging between
400 and 600 burglaries per year, used 5 years of data.
Throughout the experiment the Santa Cruz Police Department
focused on burglaries— vehicular,
residential, and commercial—and motor vehicle thefts.2 The
system works on gang violence, batteries,
aggravated assaults, drug crimes, and bike thefts. It functions
on all property crimes and violent crimes
that have enough data points and are not crimes of passion, such
as domestic violence. Homicides
generally do not provide enough data points to produce accurate
predictions.
Using the Data
To add an extra layer of security, employees transfer the data on
designated crime types from the
6. records management system (RMS) to the secure Web-based
system. The algorithm requires the date,
time, type, and location of a crime. This is public data, which
adds another level of security in case
someone intercepts the information. No one submits, collects, or
uses personal data. The system
processes the information through the algorithm and combines it
with historical crime data to make
predictions.
Staff members log in, just like they do on an e-mail account,
and the system generates hot spot maps.
For Santa Cruz police, there are 15 hot spot maps per shift.
Distributed through roll calls, these maps
indicate 500-square-foot locations. Officers pass through these
areas when they are not obligated to
address other calls. No one dispatches or requires them to patrol
the sites; they do it as part of their
routine extra checks. Some agencies using the program
designate predictive policing units to run patrols,
while others use unmarked cars to traverse hotspots.
The Santa Cruz Police Department decided not to mandate the
patrols. Personnel thought it would
eliminate the feeling of an administrative directive and
empower officers to be as proactive as their call
levels allowed. When police enter and clear the hot spot
locations, they notify dispatch with a designated
clearing code. This enables dispatch personnel to collect data on
the frequency of the officers’ presence
in the hot spots.
FBI — Predictive Policing: Using Technology to Reduce Crime
http://www.fbi.gov/stats-services/publications/law-enforcement-
bulletin/...
7. 2 of 4 6/3/2013 3:38 PM
Evaluating the Process
During the first 6 months of the program, the department made
over 2 dozen arrests within the hot spot
locations. However, the true measure of the program’s success
is not apprehensions, but the reduction
of crime. Santa Cruz police officers indicated an initial 11
percent reduction in burglaries and a 4 percent
decrease in motor vehicle thefts. As time progresses, the
reductions increase. Over a 6-month period,
burglaries declined 19 percent.
The system requires 6 months of data to assess whether the
method actually is reducing the crime rate.
Because the Santa Cruz police did not introduce any additional
variables—no additional officers were
hired, shift lengths continued, patrol structure remained the
same—the department attributed the crime
reduction to the model.
The Los Angeles Police Department (LAPD) tested the method
under a controlled experiment. The
project scientifically proved the model’s effectiveness. The city
has a larger population and more
complex patrol needs than Santa Cruz. Researchers established
the experiment in the Foothill Division
with a population of 300,000 people. They compared the
predictive policing system with LAPD’s best
practices.
Similar to the Santa Cruz test, the department distributed maps
8. to officers at the beginning of roll call. On
some days analysts produced the maps using traditional LAPD
hot spot methods. On other days, they
used the algorithm. No one told the officers where the maps
came from. Graphically they looked the
same.
The algorithm provided twice the accuracy that LAPD’s current
practices produced. While property crime
was up .4 percent throughout Los Angeles, Foothill’s declined
by 12 percent. Foothill benefitted from the
largest crime reduction of any division during the experiment.
People found it hard to understand that an algorithm performed
similar to a crime analyst. Eventually,
even the most skeptical individuals realized that the method
worked. The LAPD expanded the program
to other divisions serving a total population of over 1.5 million
people. Each one that implemented the
predictive policing software achieved crime reduction. The
department recognized that the predictive
policing system is a large improvement over previously used
approaches. When looking at a map from 1
week, the assumption is that the next week will be the same.
The computer eliminates the bias that
people have.
Gaining Support
As with any new program, questions and concerns arise. People
resist change. The Santa Cruz Police
Department worked with officers to develop maps and solicit
feedback before implementation of the
program. The department emphasized that the program does not
replace officer intuition but
9. supplements it.
The Santa Cruz Police Department found that veteran officers
usually identify 8 or 9 of the 15 hot spot
locations. Newer officers discover 1 or 2 of the areas. This
validates skilled police officers’ intuition,
provides additional targeted locations, and imparts tactical
information for new officers. The maps
reinforce existing knowledge and inform about targeting
locations. They standardize information across
shifts and experience levels.
The algorithm combines historic and daily crime information,
produces real-time predictions of areas to
patrol, and normalizes information among shifts. It eliminates
the concern about adequate information
sharing. Officers obtaining the daily hot spot maps receive any
information they missed due to vacation,
illness, or regular days off.
The program shares information graphically. It does not replace
the value of senior officers teaching
younger ones or the need for roll calls to discuss crime trends.
It cannot replace police officers’
knowledge and skills and does not remove the officer from the
equation. It puts law enforcement in the
right time and place to prevent crime.
Conclusion
FBI — Predictive Policing: Using Technology to Reduce Crime
http://www.fbi.gov/stats-services/publications/law-enforcement-
bulletin/...
3 of 4 6/3/2013 3:38 PM
10. Accessibility | eRulemaking | Freedom of Information Act |
Legal Notices | Legal Policies and Disclaimers | Links | Privacy
Policy | USA.gov | White House
FBI.gov is an official site of the U.S. government, U.S.
Department of Justice
Close
Additional departments in the states of California, Washington,
South Carolina, Arizona, Tennessee, and
Illinois have implemented the program. In November 2011 Time
Magazine named predictive policing one
of 50 best inventions for 2011.3 The Santa Cruz police chief
acknowledged the recognition, but said the
accolades are less important than the crime reduction.
According to the chief, “Innovation is the key to
modern policing, and we’re proud to be leveraging technology
in a way that keeps our community
safer.”4
Endnotes
1 Dr. George Mohler, Santa Clara University, California, and
Dr. P. Jeffrey Brantingham, University of
California at Los Angeles, California, developed the algorithm
(mathematical model) for predictive
policing.
2 George Mohler, P. Jeffrey Brantingham, and Zach Friend
conducted the research in Santa Cruz,
California.
11. 3 Lev Grossman, Cleo Brock-Abraham, Nick Carbone, Eric
Dodds, Jeffrey Kluger, Alice Park, Nate
Rawlings, Claire Suddath, Feifei Sun, Mark Thompson, Bryan
Walsh, and Kayla Webley, “The 50 Best
Inventions of 2011,” Time Magazine, November 28, 2011.
www.time.com/time/magazine/article
/0,9171,2099708-13,00.html (accessed September 27, 2012).
4 Santa Cruz Police Chief Kevin Vogel.
FBI — Predictive Policing: Using Technology to Reduce Crime
http://www.fbi.gov/stats-services/publications/law-enforcement-
bulletin/...
4 of 4 6/3/2013 3:38 PM
Copyright of FBI Law Enforcement Bulletin is the property of
Superintendent of Documents
and its content may not be copied or emailed to multiple sites or
posted to a listserv without
the copyright holder's express written permission. However,
users may print, download, or
email articles for individual use.
Science and Justice 54 (2014) 494–501
Contents lists available at ScienceDirect
Science and Justice
14. closure of
some of the more established traditional laboratories. A
discipline
should be (re-)built around the study of the ‘trace’, the remnant
of a
unique criminal activity that occurred in the past. The
information it
conveys is not only restricted to serve the Court process but
should
also support the study of many types of crime activities,
following a
variety of objectives [35,36,52,53].
This paper aims at structuring this debate by stating twelve
inter-
connected propositions, at different levels of generality, to be
tested
by research. They should be considered as a preliminary
construct
open to evolution.
cole des Sciences Criminelles,
Elsevier Ireland Ltd. All rights reserv
1.1. Is this expansion necessary?
Proposition 1. Expanding R&D in forensic science beyond its
delineation
in the report is a necessity for providing a discipline with a
sufficient ambi-
tion to justify its existence, ensure its own coherency and
favour its sustain-
able development.
The need to consider the contribution of forensic science
beyond the
laboratory is already occasionally postulated by scholars [30],
15. but the
scientific literature on this issue remains rare [52]. There are
many pos-
sibilities to specify territories to be explored [37]. They have
overlapping
shapes. We propose one possible configuration that helps to
pinpoint
risks and opportunities for forensic science to engage in these
areas.
1.2. What should this expansion cover?
The most evident step for such an expansion consists of
adopting a
global view that goes from the crime scene to the presentation
of evi-
dence in Court. In this context, the traditional laboratory is
situated as
one possible structure responsible for performing specialised
operations.
Proposition 2. Research in forensic science covers the study of
its contri-
bution along the whole chain of the justice process, from the
crime scene,
to the presentation of forensic information in Court.
This elementary expansion for forensic science is not a given.
There
are many inhibitors that dissuade researchers to embark on such
a ven-
ture. Some commentators even deny, or strictly limit, an
expanded role
for forensic science along this chain:
ed.
16. http://crossmark.crossref.org/dialog/?doi=10.1016/j.scijus.2014.
05.001&domain=pdf
http://dx.doi.org/10.1016/j.scijus.2014.05.001
mailto:[email protected]
http://dx.doi.org/10.1016/j.scijus.2014.05.001
http://www.sciencedirect.com/science/journal/13550306
495O. Ribaux, B. Talbot Wright / Science and Justice 54 (2014)
494–501
‘Expanding the forensic scientists’ domain to the ‘activity
level’
destroys the line between their expertise in their specific
forensic
discipline and a more general (and dangerous) claim to general
investigative expertise ([46: 70]).
This confined view is mainly justified by the need to keep
scientific
independence, mitigating contextual bias and avoiding
encroachment
upon each other competencies.
Such statements are particularly stimulating for research. They
immediately lead to directly address the question of the
application of
forensic science. Would it truly deserve existence, by merely
bringing
small/simple pieces of evidence before the justice system
(source level
involvement only)? And in doing so endangering the fairness of
the
judicial process at such an increased cost?
The extremity of this confinement should be more carefully
studied,
17. because it is not immediately apparent when adjoined with
Polanyi's
statement:
‘Even the most strictly mechanized procedure leaves something
to
personal skill in the exercise of which an individual bias may
enter’
([41]: 19).
And combined with Rosenthal affirmation:
‘It costs something to reduce errors, and it costs more and more
to
get rid of each error as there are fewer of them left’ [49].
Reconstructing specific events that occurred in the past is
subject to
many forms of uncertainty. Whatever the level of sophistication
of pro-
cedures and models for making decisions, forensic failures will
continue
to occur unavoidably. Each high profile case will invariably put
a little
more pressure on the system with the effect of progressively
confining
scientific laboratories in the landscape of the justice system.
This reduction in the scope of laboratories may open space for
more
fragile information and biased forms of reasoning to prosper in
crime
investigation. This could not be more evident than in the
collection of
human information (interview) that is guided by forensic
results, in
18. the complete absence of forensic advice.
An alternative response to the collapse of some independent
labora-
tories is to rebuild forensic capacity within police
organisations. It may
be another natural evolution of systems searching to fill gaps
created.
Is this movement already a reality? Research could empirically
test
this hypothesis.
The reality of this confinement is often tempered by the
employ-
ment of a case manager in the laboratory. This still understudied
func-
tion focuses on mitigating risks of biases produced by observer
effects.
It proceeds by separating the management of the case from the
evalua-
tion of observations, and by filtering contextual information
about the
case through sequential unmasking procedures [56]. However,
this
defensive function provides little indication on how forensic
science
and crime investigation should logically be articulated to favour
the
resolution of investigative problems. This needs to be studied
also.
Whatever the viewpoint, forensic science cannot operate in
isola-
tion. Indeed, lack of research dedicated to expressing this
articulation al-
lows space for pervasive misunderstandings and tensions
19. between
organisations and individuals to prosper. It is also true that no
guarantee
can be made for forensic case data to be safely and
transparently
exploited to its full potential in the variety of processes it
serves.
The study of the whole chain brings into focus two of its
important
components: (a) the contribution of forensic science to crime
investiga-
tion and (b) crime scene investigation itself.
1.2.1. Studying how forensic science may integrate with crime
investigation
The proposition to adopt a global view that starts at the scene
and
ends in Court forces the study of different forms of articulation
of foren-
sic science within crime investigation, and their respective
conse-
quences on the whole process.
This is an area of many controversies. They occur in a judicial
context
that is itself poorly formalised [25], and which is the target of
many crit-
icisms. In particular, in his 1984 seminal paper, Egger [14]
pinpointed
the incapacity of police systems to connect dots, leading to
disastrous
failures in serial murder investigations. He denounced the
fragmenta-
tion of crime investigation as it causes linkage blindness.
20. This is where the fragmentation of forensic science, confined in
spe-
cialities and silos, certainly does not address these systemic
weaknesses
coined by Egger. Research may examine how the fluidity of the
treat-
ment of scientific information is inhibited by traditional
organisational
settings of forensic laboratories. Thus, the following statement
chal-
lenges the usual pathway designed for forensic science.
Proposition 3. Crime investigation is holistic, and forensic
science is a
significant contributor to it.
In shaping police organisations during the last decade, the focus
has
been on how crime analysts, investigators, forensic scientists
and other
contributors differ through their speciality, while they actually
partici-
pate collectively to the same process of crime investigation.
Digital
traces have added new dimensions to the picture. They are used
almost
systematically and are central to most of today's investigations.
A prom-
ising avenue for research would be to consider what the actors
(i.e. the
various contributors to the investigations bringing their own
knowl-
edge and expertise) share, and what kind of collaboration must
be
stimulated to favour and regulate problem solving. Indeed, the
term
21. investigation contains at its root vestige, which means in French
the
remnant of an activity, the mark, the ‘trace’ [10]; exactly what
forensic
science studies according to Margot [36]. Adopting this view
allows
the definition of stable concepts and frameworks.
A research programme could thus examine, as its object, a
system
composed of different kinds of investigators (e.g. police
investigator,
forensic investigator, criminal intelligence analysts) trying to
solve
problems through a collective approach, by bringing their
specific
knowledge and skills in treating specific types of information.
A lot of
empirical studies could be launched around this system, its
functioning,
its adaptation to the investigation of specific cases, its
transparency and
its effectiveness.
This kind of research will inevitably address the question of
organising
forensic science with respect to the fragmentation of the
investigative
process. Various forms of bias and their consequences have
been inten-
sively discussed in forensic literature. This catalysed the debate
of
marginalising the forensic scientist from the investigation.
However, the
consequences of this fragmentation and de-contextualisation
have been
22. far less considered. This opens an important consideration that
is directly
related with a more holistic view of the investigation.
Proposition 4. The fragmentation of processes in systems and
the distanc-
ing of scientists from other figures of the investigation might
contribute to a
variety of failures, not addressed by laboratory quality
management. Thus,
contextualisation and de-contextualisation must be studied in
mirror,
depending on needs and expectations of the criminal justice
process.
There are already many documented illustrations where such
fail-
ures have occurred. One significant example is the Byford's
report on
the Yorkshire Ripper inquiry in 1981 [6]. So called ‘Byford
scientists’
have since the mid 90s deployed good practices in the role of
forensic
investigators, contributing concretely to the resolution of many
serious
crimes [1,57].
It is their responsibility to generate a productive collaboration
with
the other ‘actors’ (i.e. a contributing figure), of the
investigation. Similar
models have since been developed in many laboratories. These
scien-
tists have a global view on forensic case data available in the
context
of a case. They provide advice on how to treat it by defining
23. sequences
of operations, as well as evaluating and integrating results with
other
parts of the investigation. Priorities are defined for optimising
informa-
tion gained and, at the same time, avoiding costly and
superfluous
496 O. Ribaux, B. Talbot Wright / Science and Justice 54
(2014) 494–501
operations. However, the nature of the contribution of this
generalist,
even if it is well admitted in the field, is very rarely debated in
the scien-
tific literature [1,54,55]. Consequently, the realisation of this
task
remains largely tacit and therefore needs to be formalised in
order to
ensure transparency (see proposition five). We experience this
same
modelling gap at the scene.
1.2.2. Studying forensic science at the crime scene
Empirical research has detected many unexplained discrepancies
when measuring performance at the crime scene. They have
been recur-
rently made evident not only across jurisdictions and
operational units,
but also when making comparisons within the same units [2].
Such
results are compounded by the disparity of views about the
character
of crime scene examination as shown by a survey conducted in
Scotland
24. [33]. As it turns out, a significant number of respondents,
among them
forensic scientists from the laboratory and crime scene
examiners
themselves, still consider it as substantially ‘mechanical’ (can
be carried
out by applying a set of predefined learned procedures that
allow little
space for inductive thinking).
Such unexplained disparities should raise many more questions
for
research. Standard operating procedures for crime scene
examination
and lab-on-a-chip technologies are on the way to
implementation be-
fore intensive research is performed to understand the
underlying
logic of the treatment. We risk putting the cart before the horse.
A set of studies should deserve more attention in this context.
For
instance, Kelty and Julian [24] have identified seven qualities
that char-
acterise high performance crime scene examiners selected by
peers:
knowledge, life experience, professionalism, approach to life,
cognitive
abilities, communication and stress management. These
dimensions
overlap with the recurrent themes identified in the literature
reviewed
by Ludwig and Fraser [32] on what factors seem to influence
forensic
science effectiveness (and weaknesses) in dealing with high
volume
25. crime.
The results are signs that crime scene examination is not
mechanical
in character. It rather belongs to an imaginative type of activity.
They
encourage associations of ideas, favour the use of a solid
scientific back-
ground to regulate reasoning, and allow the deployment of
abilities
Fig. 1. The reconstruction process. A criminal activity perturbs
the physical environment (a)
developed (d). The consequences are eventually tested, refuted,
discriminated and evaluated t
to intensively communicate and exchange information. Such
signs indi-
cate the types of frameworks to be elaborated or how to recruit
people.
But research must go far further in deciphering the underlying
logic
operating at the scene and in crime investigation.
1.2.3. The study of an investigative and crime scene logic
Let us use a very simplified model in order to sketch the
modelling
challenge (Fig. 1).
Forensic science deals with activities that are unique, and that
gener-
ate material (or digital) exchanges, according to Locard's
principle. After
a certain time (Δt), crime scene investigators attend the scene
and
collect specimens.
Crime reconstruction consists of developing hypotheses about
26. the
activity, objects and individuals. A series of cycles aim at
discriminating
and refuting hypotheses, as well as develop new hypotheses.
Beliefs
change as a function of newly available information.
The active development of hypotheses about what occurred in
the
specific situation is at the core of the process. Crime scene
investigators
intensively apply this ‘abductive’ logic at the scene [11]. At the
other
extremity of the judicial process, the Court, the expert evaluates
infor-
mation from the perspective of the remaining hypotheses of the
defence
and of the prosecutor. At this stage, a more deductive style of
reasoning
takes place for the forensic scientist, in a stabilised cycle
(probabilities of
observations knowing the proposition of the prosecutor, and
respec-
tively of the defence).
There is still much to be understood beyond this very simplified
model. Taking part in the expression of an investigative logic is
suggested as a priority for forensic science.
Proposition 5. Research in forensic science needs an
epistemological
component for elaborating upon the foundations of an
investigative logic
exploiting information conveyed by traces (forensic case data).
This very limited overview shows that the quality of crime
27. scene ex-
amination and forensic investigation is mainly determined by
the ability
to draw relevant hypotheses from observations in the specific
circum-
stances of the case, and to regulate their management.
Moreover, the
potential contribution of forensic science in crime investigation
is of a
. Relevant specimen is recognised, collected and measured (b).
Alternative hypothesis is
hrough experiments.
497O. Ribaux, B. Talbot Wright / Science and Justice 54 (2014)
494–501
broad variety, when viewed with such a clinical perspective
[58]. It is
felt to be significant, but this is still not demonstrated by
research.
An epistemological exploration can lead for instance to suggest
links
between works such as Peirce semiotic [13], the evidential
paradigm of
the historian Carlo Ginzburg [16], the method of historical
science that
searches for explanations to past singular events on the basis of
current-
ly observable marks [7,8], the scientific investigation of crime
defined
by Kind [25], the use of probabilities in the investigation
suggested by
Jackson et al. [22], and case-based reasoning when successful
solutions
28. to previous problems are used to deal with a new similar
situation [26].
Probably many other approaches can be integrated where traces
are
recognised as signs providing information on what occurred in
the past.
1.3. Questioning the effectiveness of forensic science
This background causes other concerns about how to
appropriately
integrate innovative and flourishing technologies. The
laboratory-
centric view of forensic science causes deviation from this
fundamental
question. It focuses mainly on technologies rather than on the
value of
information conveyed by forensic case data. Peter Deforest
wrote in
1999:
‘Has the field advanced?’ On the face of it there would seem to
be no
question that the field of criminalistics has advanced. But has
it?
While we have forged ahead technologically, in my view we
have
backslid with respect to our core activity — that of applying
science
and scientific reasoning to criminal and civil investigations
([12:
197]).
Technologies are aimed at supporting the resolution of
problems.
But the specific problem under scrutiny, as well as the global
29. impact
of forensic science on solving and reducing crime often remains
relegat-
ed behind the prescriptive application of procedures and
technologies.
The rapid evolution of technologies may provide an explanation
of
this deviation. DNA came about 80 years after the
implementation of
fingerprints. But the quantity of electronic data has since
exploded, at
a time scale that is equivalent to about half a professional
career. ‘Big
data’ has become an effective issue in less than one generation
of
students [51]. Efforts to assimilate these changes are
tremendous. The
Policing debate
Fig. 2. Order of magnitude of the DNA evi
market is stimulated to push new technologies at a rate that
exceeds
our own ability to integrate them. Lab-on-a-chip and big data
technolo-
gies are at the door, but the current paradigm provides no
solution for
avoiding a further destabilisation of the discipline. A solid
vision elabo-
rated through a strong focus on fundamental research has
become
indispensable in our rapidly evolving system.
Another exciting avenue for research exists here.
Proposition 6. Refocusing forensic science on problem solving
30. will
provide a more central position to the discipline in crime
investigation, as
well as greater stability and sustainability.
By focusing on means rather than ends and distancing from
other
actors, current forensic systems forget to consider their global
effective-
ness in policing [53]. Indeed, when it is measured, an
embarrassing
evidence funnel is made visible, i.e. there is a huge gap between
what
is collected at the scene and what is effectively used at the end
of the
judicial process. This gap is made always more explicit through
the so
called end-to-end studies on the use of DNA profiles (Fig. 2).
They show that the chance for a trace, generated by the activity
of
the offender, to be eventually presented at Court, remains very
low,
no matter how extensively technology is deployed.
Information is lost at various decision points throughout the
case
handling process. Registering the case and attending the scene,
process-
ing the crime scene and choosing if a specimen must be sent to
the lab-
oratory are examples of choices to be made. Many factors
influence the
extraction of the profile, the comparison of the profile with the
content
of databases, the interpretation of matches, the integration of
31. results
into the investigation, the use of this evidence in reporting, and
eventu-
ally the use of the piece of evidence at Court. Regardless, when
seen as a
whole, the process provides a questioning picture for the
effectiveness
of forensic science.
Such evaluations however are unfair. These indicators do not
grasp
the broad variety of useful information brought to the
investigation by
forensic science. The UK Parliamentary Select Committee
‘Forensic
Science on Trial’ acknowledged this in 2005:
‘The main contribution that forensic science makes to the
criminal
justice system is the generation of intelligence to assist
investiga-
tions’ ([20]: 8).
Justice / Forensic debate
dence funnel for high volume crime.
image of Fig.�2
498 O. Ribaux, B. Talbot Wright / Science and Justice 54
(2014) 494–501
There are unfortunately still very few empirical bases to ground
this
assertion.
Beyond investigation, forensic science is understood according
32. to
the lines debated in the scientific literature. Development
centres on
serving the Court process, yet it is evaluated on how it supports
crime
control or crime reduction strategies in policing (see Fig. 2).
There is a
tension here that needs urgently to be made visible by research,
because
this misunderstanding may have dramatic consequences in
deciding
the future of forensic science [2,5,48,57].
1.4. Forensic science in policing
There is again a lack of models and almost no scientific debate
to
express the potential of forensic science in policing.
Proposition 7. Forensic science can and should contribute to,
and actively
engage in policing.
Going into policing is another controversial possible expansion
for
forensic science. As Inman and Rudin state, forensic science is
required
to strictly stay focussed on the court's need:
‘The scientific analysis is only performed at the behest of
someone
seeking to introduce the evidence into a court of law’ ([21:15]).
In this context, can forensic science enter into the policing
debate,
which is a discipline that has acquired some autonomy from the
justice
system?
33. This is also a delicate question, but let's have a quick and
simplified
look at a contemporary policing debate in order to better
identify risks
and opportunities. It may actually be that the broad use of
separate fo-
rensic identification techniques in a court-oriented strategy is
already
contributing to overwhelm actors of the system. The order of
magnitude
of the number of matches obtained through biometric systems
has
changed. Estimations already indicate that the treatment of all
the
matches obtained at a national and international level through
auto-
matic data exchanges becomes intractable. At the same time, the
real
effect of this traditional intensive case-by-case identification
method
is controversial in terms of crime reduction [44].
But what are the alternatives? Mostly they focus much more on
the
study of repetitive crimes. They started to emerge in the 1970s
along
many different lines. Goldstein has initiated one of the most
important
streams [17]. He defined ‘problem oriented policing’ (POP),
which
aims at providing a proactive component to the organisation.
The pro-
cess concentrates efforts on Scanning repetitive and persistent
criminal
or security problems encountered by the police in their daily
34. operations
(S). Their Analysis (A) encourages then inferring the causes of
the prob-
lem identified. Elaboration of possible Responses for mitigating
the
problem, choosing and implementing what is expected to be the
more
efficient solutions are the next steps of the approach (R). The
systematic
Assessment and adaptation of the chosen dispositive finalises
the pro-
cess (A). The focus is on mitigating the problem by preferring
preven-
tive approaches, not necessarily by arresting or identifying
individuals.
This method has become very popular in policing and known
under
the acronym SARA.
Does forensic science really have something to do with problem
ori-
ented policing? There are actually many examples of such
implications.
One such experience in Boston is particularly interesting and
was the
object of scientific papers written by Braga [3,4], a
distinguished special-
ist in problem oriented policing. It was eventually recognised
that the
systematic comparisons of bullets collected at crime scenes
‘helped
guide violence prevention efforts by establishing patterns in
particular
areas and among specific individuals’. Of importance is to
notice that
this programme was not led by the police, but by other
35. institutions
that had to deal with the problem. Policing does not mean that
the
police are the only actors, or even that they must lead such
projects or
to what degree they participate in the application of the method.
What is common in these modern strategies, beyond problem
solv-
ing policing, is the focus on repetitive crimes under all their
forms. Even
if different streams in policing exist, they tend to be regrouped
under
the umbrella of ‘intelligence-led policing’.
This opens the door to potential fears prompted by the
misinterpre-
tation of the term ‘intelligence’. Thus, has forensic science to
compro-
mise with intelligence by developing a forensic intelligence
branch?
Maybe the term itself ‘forensic intelligence’ which is
occasionally
regarded an oxymoron, is not an appropriate one. The term
seeks only
to express that forensic case data has the potential to contribute
in a
structured way to the detection and strategic approach towards
repeti-
tive crime, the support of preventative policing models, and the
disrup-
tion of criminal enterprises in crime reduction.
The first generation of intelligence-led policing systems was
much
more based on the results of research in criminology, that
36. clearly
showed that a few minority of offenders were responsible for a
high
proportion of crimes reported [42].
Targeting and neutralising these so-called prolific offenders
would
hopefully significantly reduce crime. Obviously, the problem
was how
to target prolific offenders. Traditional police solutions can be
intrusive,
and predicting recidivism is far from obvious. How forensic
science can
support this process in a more neutral way is thus a critical
question to
be evaluated within this framework.
This way of considering forensic science in policing changes
the
metrics, and even the econometrics, for evaluating its
efficiency. In
this framework, assessing the following proposition takes an
evident
importance.
Proposition 8. The adoption of intelligence-led policing changes
how to
consider the effectiveness of forensic science. Intelligence-led
crime scene
processing and crime scene linking become central.
Let us take one example to obtain an understanding of the
dimen-
sion of the challenge.
1.5. Forensic science and intelligence-led policing: an example
37. A crime scene investigator attends scenes of residential
burglaries
into the same …
RESEARCH ARTICLE
Rethinking the Compstat process to enhance problem-solving
responses: insights from a randomized field experiment
Brenda J. Bonda* and Anthony A. Bragab
aSawyer Business School, Suffolk University, Boston, MA
02108, USA; bSchool of Criminal
Justice, Rutgers University, Newark, NJ 07102, USA
Compstat is an important administrative innovation in policing
that provides a
much-needed mechanism for holding mid-level managers
accountable for controlling
crime rates. Research evidence suggests that Compstat is more
likely to generate
reactive crime control responses rather than more creative
problem-solving responses.
A randomized field experiment in Lowell, Massachusetts found
that ‘problem-solving
meetings’ produced more innovative responses to crime
problems and generated
stronger crime control gains when compared to the Compstat
process. Analysis of
qualitative data collected to monitor the implementation of the
experiment revealed
that important differences in meeting dynamics were associated
with observed differ-
ences and suggests promising avenues to improve Compstat
38. processes.
Keywords: Compstat; problem-solving; police management;
crime control
Introduction
Compstat has been hailed as one of the most prominent police
innovations of the past
20 years (see, e.g. McDonald, 2002; Silverman, 1999). While it
has many features,
Compstat can be generally viewed as a combined technical and
managerial system that
embeds the technical system for the collection and distribution
of police performance
data in a broader managerial system designed to focus the
organization on specific
objectives, usually involving crime reduction, by holding a
subset of managers account-
able for using organizational resources appropriately in pursuit
of these objectives
(Moore, 2003; Moore & Braga, 2003). Since the approach was
first implemented by the
New York City Police Department (NYPD) in 1994 (Bratton,
1998), Compstat has been
adopted by various police organizations, where police
executives attempt to improve
their performance by embracing data-based decision-making,
enhanced problem-solving,
and management accountability (Weisburd, Mastrofski,
McNally, Greenspan, & Willis,
2003).
Despite the extensive diffusion of Compstat across police
agencies in the USA and
other countries, the crime control benefits of Compstat remain
40. In this paper, we reflect upon the outcomes of a randomized
field experiment in
Lowell, Massachusetts that found ‘problem-solving meetings’
designed to ensure
adequate treatment dosage in treatment crime hot spots resulted
in far more problem-
solving responses and greater crime control gains when
compared to the crime-reduction
responses implemented at control hot spots naturally generated
by a citywide Compstat
process (Braga & Bond, 2008). Using qualitative data collection
and analysis methods,
we compare meeting dynamics to understand why the problem-
solving meetings yielded
more innovative and effective crime-reduction interventions
relative to Compstat
meetings. The results of our analyses suggest that the inputs and
processes of the prob-
lem-solving meetings accounted for the observed differences in
strategies and outcomes.
By integrating these dynamics into Compstat meetings, these
managerial systems could
be used to good effect in moving police departments towards
more robust community
problem-solving activities.
Compstat and problem-solving
Herman Goldstein (1979) argued that the crime control failures
of the standard model of
policing could be explained by the fact that police departments
were poorly organized
to do something about crime. Compstat sought to overcome this
organizational problem
by empowering the command structure to do something about
41. crime problems (Bratton,
1998; Weisburd et al., 2003). As originated by the NYPD,
Compstat operates as a stra-
tegic control system implemented to collect and disseminate
information on crime prob-
lems and track responses used to control them (McDonald,
2002; Silverman, 1999).
These elements are most visibly displayed in Compstat meetings
during which precinct
commanders appear before the department’s top brass to report
on crime problems in
their precincts and what they are doing about them (Bratton,
1998; Silverman, 1999;
Weisburd, Mastrofski, Willis, & Greenspan, 2006).
A recent Police Foundation study of the implementation of
Compstat by US police
departments defined Compstat as a ‘strategic problem-solving’
model that seeks to focus
police organizations on specific crime problems and to empower
police organizations to
identify and solve those problems (Weisburd et al., 2003). When
compared to common
conceptions of problem-oriented policing programs that tend to
focus on line-level prac-
tice (Eck & Spelman, 1987; Goldstein, 1990), Compstat focuses
on the ways in which
police agencies can be organized as problem-solving institutions
rather than on the spe-
cific problem-solving strategies that police use to address crime
problems (Weisburd
et al., 2006). The Police Foundation identified six key elements
of Compstat that form a
comprehensive approach for mobilizing police agencies to
identify, analyze, and solve
public safety problems: mission clarification; internal
42. accountability; geographic organi-
zation of command; organizational flexibility; data-driven
problem identification and
assessment; and innovative problem-solving (Weisburd et al.,
2003).
In practice, however, these core elements do not seem to
translate into enhanced
problem-solving responses in Compstat agencies. A Police
Foundation survey found that
Compstat police agencies were not any more likely to
implement innovative
Police Practice and Research: An International Journal 23
problem-solving responses than non-Compstat police agencies
(Weisburd et al., 2003).
The on-site observations documented a preponderance of
traditional responses, such as
saturation patrol and increasing arrests in problem areas. The
Police Foundation research
team also conducted deeper ethnographic assessments of three
‘model’ police depart-
ments – including the Lowell Police Department (LPD) – that
closely followed the
NYPD Compstat model (Willis, Mastrofski, Weisburd, &
Greenspan, 2004). Compstat
police agencies tended to place the greatest emphasis on
mission clarification and inter-
nal accountability elements. Unfortunately, the reinforcement of
the traditional hierarchi-
cal command structure interfered with the decision-making
authority of mid-level
managers and diminished creative problem-solving efforts.
43. The LPD and Compstat
Lowell, Massachusetts is a small city of some 105,000 residents
located about 30 miles
northeast of Boston and has a geographic expanse of 14.5
square miles. After his
appointment in 1995, Superintendent Edward F. Davis adopted
Compstat as a central
component of his departmental reforms and closely followed the
model developed by
the NYPD; as such, this makes Lowell an ideal research site as
the LPD engaged a
model that was widely replicated in police agencies across the
USA and many other
nations. As described by Willis et al. (2004), the LPD’s
Compstat goals included, ‘elicit-
ing collective input on crime patterns and problem-solving
strategies, encouraging infor-
mation sharing on crime locations, victims, and suspects, and
facilitating the
deployment of department resources’ (p. 474). LPD Compstat
meetings were held on a
bi-weekly basis, led by Superintendent Davis or one of the two
Deputy Superintendents,
and typically included 25–30 attendees.2 The captains who led
the LPD’s three policing
sectors were the focal point of the Compstat meetings and other
mid-level managers
would also be queried as crime problems intersected with their
locus of responsibility.
Citywide and sector-level crime trends and hot spot maps were
closely reviewed; the
presentation of the data was accompanied by questioning by
Davis and the Deputy
Superintendents on the nature of concerning crime trends and
44. concentrations and the
captains’ strategies to address recurring crime problems.
Willis et al. (2004) found that the LPD’s Compstat fostered a
clear sense of mission
and accountability relative to crime control. However, they also
reported that the process
did not realign decision-making toward street supervisors, nor
did it provide sector cap-
tains with greater flexibility in using organizational resources.
Although the use of crime
data and analyses for problem and hot spot identification were a
centerpiece of the
Compstat process, the captains rarely referenced data beyond
personal perceptions and
police reports to understand the conditions that cultivate crime
patterns and hot spots.
Further, the Willis et al. (2004) study revealed that captains
habitually relied on tradi-
tional crime control tactics rather than engage in more creative
and innovative strategies
to deal with crime hot spots. While the LPD’s Compstat was
established, in part, to
encourage information sharing and innovative problem-solving,
Willis et al. (2004) sug-
gested that Lowell fell short of achieving these goals because
the meeting reinforced tra-
ditional, hierarchical command and control constitutions that
impeded the reforms and
innovation that they had hoped to achieve.
The current study
This research analyzes process data collected during the
implementation of a
randomized controlled trial to test the impact of problem-
45. oriented policing on crime and
24 B.J. Bond and A.A. Braga
disorder hot spots in Lowell (Braga & Bond, 2008). Using
computerized mapping tech-
nology, 34 crime and disorder hot spots were identified and
subsequently matched into
17 like pairs; one member of each pair was randomly allocated
to treatment and control
conditions. The intervention period lasted for one year (1
September 2005 through 31
August 2006). Superintendent Davis assigned ultimate
responsibility for the implementa-
tion of the problem-oriented policing intervention at the
treatment places to the captains
that managed the LPD three policing sectors. The problem-
oriented policing intervention
was managed through monthly ‘problem-solving meetings’
designed to ensure adequate
treatment dosage in treatment crime hot spots. The control hot
spot areas experienced
the routine amount and kinds of police strategies that such areas
in Lowell would expe-
rience without focused intervention – arbitrary patrol
interventions, routine follow-up
investigations by detectives, and ad hoc community problem-
solving attention.
Policing actions at the control hot spots were managed through
the LPD’s bi-weekly
Compstat meetings. These meetings were attended by the
research team to document
possible contamination issues – whether the control hot spots
46. were receiving concen-
trated police actions that were similar to the interventions
implemented at the treatment
hot spots. While control hot spots were not acknowledged as
such, these locations were
routine subjects of Compstat meetings as persistent problem
places that required ongo-
ing police attention. When control hot spots were addressed, the
research team took
careful notes on the discussion that followed. It is worth noting
here that treatment hot
spots would occasionally be mentioned during the Compstat
process.
Analytical framework and data collection
The randomized controlled trial design facilitated a structured
comparison between the
problem-solving meetings that guided the implementation of the
problem-oriented polic-
ing intervention at the treatment hot spots and the Compstat
meetings that governed
problem-solving attention in the comparison hot spots.
Qualitative data collected to
monitor treatment and control conditions were analyzed to
understand the key character-
istics and dynamics of the two sets of management
accountability meetings. The analyti-
cal framework draws on group/team3 effectiveness research to
examine whether meeting
characteristics and dynamics had an influence on the observed
differences. The
Inputs-Process-Outcome (IPO) framework is applied to the LPD
meetings to explore the
connection among the inputs into group work, group process
behaviors to develop
47. strategies and tactics to address crime and disorder, and their
impact on working group
outcomes (Campion, Medsker & Higgs, 1993; Gladstein, 1984).
The application of the
IPO framework to this research is appropriate for two reasons.
First, since the LPD was
the site for both sets of meetings, organizational factors were
constant. Variations in
organizational culture, hierarchy, policies, and procedures can
have strong impacts on
group dynamics. Second, the Compstat and the treatment
problem-solving meetings had
very similar crime-reduction and accountability objectives.
Qualitative data were collected using two basic techniques: (1)
overt participant
observation; and (2) intensive interviewing of LPD
commanders, managers, and officers.
As described by Lofland and Lofland (1984), participant
observation is the process in
which a researcher gains a close and intimate familiarity with a
given group of individu-
als and their practices through an intensive involvement with
people in their natural
environment, usually over an extended period of time. In this
study, the LPD was
overtly observed by the research team in two sets of crime
control meetings and field
settings as researchers monitoring the implementation of an
experiment. Intensive
Police Practice and Research: An International Journal 25
interviewing, also described by Lofland and Lofland (1984) as
48. ‘unstructured
interviewing,’ involves recurring guided conversations to
discover the subject’s experi-
ence of a particular topic or situation.
Over the course of the experiment, N = 14 monthly problem-
solving meetings and
N = 22 bi-monthly Compstat meetings were observed.4 In both
meetings, the number of
participants and their level and nature of participation, patterns
of participation, partici-
pant roles within the agency, and the content of conversations
involving strategies to
address hot spot locations were carefully documented. The
research team also
interviewed key LPD staff at scheduled meetings and through
informal conversations.5
Qualitative data on the genesis and implementation of crime
control strategies were also
collected during N = 52 weekly ride-alongs to monitor progress
at treatment hot spots
and N = 12 monthly researcher site observations to document
actions taken at control
hot spots.6 We sought to improve reliability of the qualitative
data by cross-checking
and probing study participants’ responses to the interview
questions. Interview and
meeting observation data were recorded in the form of
handwritten notes, transcribed,
and analyzed by the authors.
In the analysis, we selected statements that illustrated themes
consistently found
throughout the data. The quotes used were not atypical, with the
exception of a few
49. issues that we indicate a small number of respondents
mentioned. We were also careful
throughout the data analysis to ensure that the emerging themes
correctly reflected
respondents’ descriptions. Thus, the research team utilized
grounded theory methods to
identify recurrent topics in addition to less common but salient
issues (Strauss, 1987).
Results
Meeting inputs
While the problem-solving meetings were based on the LPD’s
Compstat process, there
were some obvious differences. The problem-solving meetings
were held on a monthly
rather than bi-weekly basis. Compstat centered on the
identification of crime patterns and
hot spots across the city and within sectors, while problem-
solving meetings focused on
17 persistent crime and disorder hot spots. The problem-solving
meetings were also
smaller, with a mean of 13 participants that included Davis,
then-Deputy Superintendent
Kenneth Lavallee, the three sector captains and selected officers
from their commands,
the Community Liaison, the Director of Research and
Development, and a civilian ana-
lyst from the Crime Analysis Unit. Compstat meetings averaged
27 participants; the addi-
tional meeting members included other members of the LPD
command staff and external
agency representatives (such as probation officers, prosecutors,
and others).
50. It is also important to note here that the Compstat meetings
were held in a large
room with the LPD command staff seated around a U-shaped
table facing a screen with
projected computerized crime statistics for presentation
purposes. Lowering ranking offi-
cers, civilian staff, and members of outside agencies were
seated around the table.
Superintendent Davis or one of the Deputy Superintendents led
the meeting as a formal
question-and-answer session for particular sector captains with
occasional input from
other high-ranking LPD officers on specific issues. In contrast,
the problem-solving
meetings were held in a smaller room with all participants
(regardless of rank, civilian/
sworn status, or external observer status) sat around a single
rectangular table. While
Superintendent Davis or Deputy Superintendent Lavallee led the
meetings, the format
was much less formal and there was an expectation that
everyone around the table
needed to participate in the discussion.
26 B.J. Bond and A.A. Braga
At each problem-solving meeting, the crime analyst presented
simple trend analyses
of citizen calls for service and crime incidents in each of the
treatment hot spots to
determine whether crime and disorder problems were being
positively impacted. If the
data revealed that calls for service were decreasing,
Superintendent Davis praised the
51. captains and their officers and asked them to explain why they
believed their actions
were producing the desired effects, and what else could be done
to keep calls for ser-
vice decreasing. If the analysis revealed that the number of
citizen calls for service had
remained the same or increased, Superintendent Davis peppered
the captains with ques-
tions about their plans for dealing with recurring problems in
the hot spot areas. The
problem-solving meetings also served as a venue for the
command staff, captains, offi-
cers, and other LPD staff to explore and share ideas on
plausibly effective prevention
strategies for persistent problems in the treatment places.
While the performance measurement accountability principles
were borrowed from
Compstat, the activities at the problem-solving meetings
represented an ongoing scan-
ning, analysis, response, and assessment process (see Eck &
Spelman, 1987). The routine
measurement and review of strategies in the treatment places
served as an important
mechanism to ensure that there was a strong treatment dosage
for the experiment. The
problem-solving meetings were designed to ensure that the
captains and their officers
were implementing the problem-oriented policing program and
adhering to the require-
ments of the experimental research design. These meetings were
explicitly focused on
implementing the approach by addressing local community
concerns as measured by
trends in citizen calls for service in the treatment hot spot areas,
holding police managers
52. accountable for dealing with identified problems, and serving as
a venue to enhance the
creativity of implemented responses through open discussion
and idea sharing.
Research on group work suggests that inputs such as group size,
group composition,
and task design influence the outcomes of working groups
(Campion et al., 1993;
Stewart, 2006). Our analysis documented that the Compstat
meetings were larger and
group composition was slightly more varied than the problem-
solving meetings. Our
analyses of the qualitative data also suggested neither group
size nor composition seem
to exert noteworthy independent effects on the meetings. The
smaller size of the
problem-solving meetings may have made it easier to conduct
intimate conversations of
particular problems. However, as will be discussed further
below, the meeting processes
were much more influential in explaining variations in outcomes
between the Compstat
and problem-solving meetings.
Task design represents what the group is attempting to
accomplish (Hackman &
Oldham, 1980). In the Compstat meetings, Superintendent Davis
tended to guide con-
versations towards clusters of crime incidents within a
particular sector, specifically
including the 17 comparison hot spots. In this way, the task
designs of the two meetings
were broadly similar as both meetings were intended to produce
the same outcomes –
increased problem-solving responses and reduced crime at
53. specific problem places.
However, the geography of targeted crime problems were
clearly more variable in the
Compstat meetings as discussions, at times, ranged from
specific addresses to neighbor-
hoods. Another important difference was the explicit
identification of specific problems
within each targeted treatment hot spot in the problem-solving
meetings. On average,
the sector captains identified four problems per treatment hot
spot that seemed to
generate recurring crimes. These problems became a ‘caseload’
of work: underlying
conditions associated with these problems were dissected,
appropriate responses were
discussed, and response performance was measured repeatedly
over the course of the
experiment.
Police Practice and Research: An International Journal 27
Meeting process
Meeting process elements include a variety of group dynamics,
including communica-
tion and cooperation behaviors, constructive feedback systems,
collaborative planning
practices, the level of interdependence and trust among
participants, and the functional
diversity and roles participants represent in the organization and
group (Williams &
Allen, 2008). Analysis of our qualitative data from the
Compstat and problem-solving
meetings suggested that the two working groups had the greatest
54. differences in three
key areas: communication and information sharing,
collaborative planning, and the
appreciation of the roles and skill sets of a broader range of
group members in
achieving group goals.
Communication and information sharing
Communication and information sharing are recognized as
influencing group
effectiveness (Woolley, Gerbasi, Chabris, Kosslyn, & Hackman,
2008). Woolley et al.
(2008) asserts that it is more than just information sharing that
matters, but rather the
revealing, organizing, and pooling of information that matters
most to outcomes. The
LPD Compstat meetings almost always involved one-to-one
communication between
the lead executive officer and the sector captains in a
mechanical and, often, superficial
manner. Similar to the observations made by Willis et al.
(2004), communication
exchanges were usually perfunctory, involving quick reviews of
identified problems and
brief reports on responses by the sector captains. While we did
not observe any
instances where sector captains were publicly humiliated, the
communications by the
lead executive officer were highly authoritarian in nature and
occasionally dismissive
when the sector captains provided answers that were believed to
be uninformed or
incorrect.
Unless the Superintendent prompted another meeting member to
55. contribute, the
captain of the reporting sector was the only person discussing
problems at these varying
high-crime places. Moreover, these discussions often did not
lead to conversations on
appropriate problem-solving responses that were linked to
underlying conditions and
dynamics. Over the course of the study, we documented N = 170
distinct discussions of
spatial crime problems at Compstat meetings. Only 25.8% (N =
44) of these discussions
were followed by a problem-oriented conversation about the
varied ways LPD officers
could impact these persistent spatial crime problems.
Unfortunately, the resulting
responses usually included a preponderance of traditional
enforcement actions.
Many instances were noted where deficient problem-solving was
driven by a lack of
engagement of low-ranking officers and civilian staff. For
instance, in Compstat meeting
1, the Superintendent asked the North Sector captain to report
on his plans for address-
ing community concerns over an assault hot spot in the
downtown area. The captain
responded by simply stating that he was ‘redeploying officers to
address community
concerns and complaints.’ The community liaison was not asked
to report on commu-
nity concerns and substance of the LPD’s response to the
problem. In fact, across the
observed Compstat Meetings, this important LPD staff member
was rarely asked to
comment on any police-community communication issues.
56. In Compstat meeting 19, the Superintendent requested the East
Sector captain to
explain some concerning increases in sector-wide property and
violent crime trends by
reflecting on the content of the underlying crime incident
reports. The captain responded
by stating, ‘Chief, I don’t know how these statistics are
generated, and I’m not sure I
28 B.J. Bond and A.A. Braga
ever will.’ The crime analyst who was managing the
presentation of crime statistics for
the Compstat meetings was not asked to provide any insights on
the underlying crime
incident reports. Rather than engage in what could have been a
valuable conversation
about the data, the conversation moved to the next topic. After
each of these Compstat
meetings, we asked the community liaison and crime analyst
why they did not offer any
insights. Both expressed concerns that it would be detrimental
to their careers if they
offered information that contradicted their superiors. This
parallels previous research
findings that most Compstat participants do not contribute
information to avoid the
perception of questioning authority (Weisburd et al., 2006;
Willis et al., 2004).
Problem-solving meetings fostered a more participatory and
democratic problem-
solving approach, without minimizing the accountability of
captains. The communica-
57. tion tone of the lead executive officer was less formal, more
engaging, and inclusive.
Captains were encouraged to discuss crime problems and their
problem-solving ideas,
without a fear of being embarrassed if an idea was incomplete
or failed. All meeting
members were expected to contribute and share information. In
problem-solving meet-
ing 5, the crime analyst began the conversation on treatment hot
spots in the West Sec-
tor by observing, ‘Three hot spot areas seem to be driving the
numbers, so maybe we
should focus on those areas.’ The West Sector captain and two
patrol officers then fol-
lowed this observation by identifying the underlying problems
that might be driving the
increases and the entire group contributed to the search for
potential responses. In prob-
lem-solving meeting 5, Superintendent Davis suggested, ‘Let’s
look at the hot spots
with crime reductions so we can try to understand why other hot
spots are not showing
the same decreases.’ Instead of offering the same perfunctory
replies used in Compstat
meetings, the sector captains engaged in a thoughtful discourse
on the observed differ-
ences. Other participants, such as the community liaison,
enriched the discussion by
contributing information based on what they knew from their
work in the hot spots.
Problem-solving meeting members were encouraged to
participate with the intent of
offering feedback in order to improve upon, rather than simply
criticize, ideas. In orga-
nizational research, this is known as ‘voice behavior’ (LePine &
58. Dyne, 1998 and this
manner of risk-taking communication has been found to
influence group outcomes posi-
tively (Edmondson, 1999). For instance, in problem-solving
meeting 3, Deputy Superin-
tendent Lavallee asked participating officers if their strategies
at their various treatment
hot spots included serving warrants. The East Sector lieutenant
responded by stating
‘we are focusing on car break problems in this hot spot because
that is our biggest
problem, and while warrant checks are important, they are not
of greatest concern to the
police or the community at the moment.’ Deputy Lavallee
accepted this explanation
without questioning the lieutenant’s judgment, suggesting a
dynamic in support of open
and honest communication without fear of reprimand. Risk-
taking communication was
rarely observed at Compstat Meetings. Captains …
THE NUMBERS DILEMMA: THE CHIMERA OF MODERN
POLICE
ACCOUNTABILITY SYSTEMS
JAMES F. GILSINAN*
INTRODUCTION
In the lexicon of dyadic American folk sayings, "motherhood
and apple
pie" is. being joined by "transparency and accotmtability."
Similarly, like
"motherhood and apple pie," one takes on the iconic nature of
59. the saying at
considerable risk. After all, who could be against transparency
and
accountability? Management gum Stephen R. Covey argues that,
"[ajccountability breeds response-ability."^ This formulation
neatly captures,
the assumed relationship between accountability and
performance. There is
also an assumed relationship between transparency and a robust
democratic
system.̂ Joumalist Peter Finn notes, a "basic tenet of a healthy
democracy is
open dialogue and transparency." Finally, accountability and
transparency are
seen as reinforcing each other." Being held accountable
assvunes that one's
actions are available for review and critique.' Indeed, the Dodd-
Fratik Wall
Street Reform and Consumer Protection Act, passed in the wake
of the recent
financial meltdown, attempts to regulate the disclosure
requirements of
fmancial firms on the asstimption that such mandated
transparency will result
in better performance on the part of banks and other financial
institutions.^
Police departments in the United States and elsewhere have
been quick to
jump on the transparency and accountability bandwagon.^ The
increasing use
* E. Desmond Lee Professor in Collaborative Regional
Education, Department of Public Policy
Studies, Saint Louis University.
60. 1. STEPHEN R. COVEY, PRINCIPLE-CENTERED
LEADERSHIP 49 (1990).
2. Elizabeth Garrett, Accountability and Restraint: The Federal
Budget Process and the
Line Item Veto Act, 20 CARDOZO L. REV. 871, 924 (1999).
3. Peter Fenn, POLITICO: THE ARENA (Apr. 23, 2009),
http://www.politico.com/arena/
perm/Peter_Fenn_l 49A409A-FDA2-41EA-ACA1-
41133DF86F66.html.
4. Joshua N. Auerbach, Police Accountability in Kenya, 3 AFR.
HUM. RTS. L. J. 275, 282
(2003).
5. Ruth W. Grant & Robert O. Keohane, Accountability and
Abuses of Power in World
Politics, 99 AM. POL. SCI. R. 29, 29 (2005). .
6. Cody Vitello, The Wall Street Reform Act of 2010 & What it
Means for Joe & Jane
Consumer, 23 LOY. CONSUMER L. REV. 99, 101 (2010).
7. Erik Luna, Transparent Policing, 85 IOWA L. REV. .1107,
1163 (2000).
93
94 SAINT LOUIS UNIVERSITY PUBLIC LAW REVIEW [Vol.
XXX11:93
of accessible, web-based, real time crime data, using geographic
information
61. system (GIS) technology to display neighborhood crime
pattems, represents
the move toward transparency on the part of major city police
departments.^
Similarly, the rapid adoption of COMPSTAT-like programs in
mid to large
size departments speaks to a willingness to be held accountable
for crime
occurrences and their control.' Unfortunately, there are a
multitude of reasons
why systems designed to increase transparency and
accountability will not
work and, in fact, may make the very goals sought by programs
of
organizational reform less likely to be achieved.
In this Article, I argue that there are five obstacles facing police
reformers
seeking to increase transparency and accountability in law
enforcement
organizations.' Moreover, these obstacles are nearly
insurmountable." The
phrase "nearly insurmountable" means that while change can
and will occur,
the changes will, at best, be at the margins of the organization—
and, at worst,
such changes may make situations in need of correction more
problematic.'^
This pessimistic assessment is due to the nature of the numbers
themselves; the
nature of organizations, particularly those that do not produce
an objective
product; the culture of policing; the institutional environment in
which the
police operate; and the larger cultures' failure to distinguish
among the
62. concepts of data, information, and knowledge.'^
I. THE SIREN SONG OF COUNTING AND NUMBERS
When presented with a table of numbers that purport to
objectively
measure or describe a phenomenon, in keeping with the dictates
of
accountability or transparency, or both, the reader is often
lulled into a state of
process amnesia. Numbers are always the end product of a
series of decisions,
many of which are subjective and somewhat arbitrary.''' There
are at least six
decision points that affect the "objective" nature of the numbers
being
reported.'^ The first is obviously the decision to count one thing
rather than
another.'^
8. W. at 1175.
9. John R. Firman, Deconstructing COMPSTAT to Clarify its
Intent, 2 CRIMINOLOGY &
PUB. POL'Y 457, 458 (2003).
10. See m/ra Parts I-V.
11. See Gary W. Sykes, The Functional Nature of Police
Reform: The 'Myth' of Controlling
the Police, 2 JUST. Q. 51, 53 (1985).
12. Id
13. See infra nott 111.
14. Craig D. Uchida, Carol Bridgeforth & Charles F. Wellford,
Law Enforcement Statistics:
63. The State of the Art, 5 AM. J. POLICE 23, 29 (1986).
15. Id
16. Id
2012] THE NUMBERS DILEMMA 95
The simple act of deciding to count or not count something
confers or
denies a certain importance to an object or outcome. A number
of criminal
justice examples illustrate the point.
Until the early 1970s, domestic assault did not "coiinf as a
serious
offense.'̂ On the other hand, homosexual activity between
consenting adults
did "count" as a crime. In fact, why do we bother to count crime
at all? Those
familiar with the history and the development of the FBI
Uniform Crime
Reports (UCR) know the reason certain kinds of crime were
counted was
motivated by a political goal: shielding top police officials from
the periodic
crime cmsades of the tabloid press." The sensationalism of
crime reporting,
and the consequent threat to the job stability of a police chief,
literally could be
count-ered by the "objectivity" of numbers.^" Of course, since
its beginning,
the UCR has been plagued by a second subjective decision point
prior to
arriving at a number: What counts as an instance of a
phenomenon?^'
64. The specifically political nature of this decision point is noted
by Deborah
Stone who points out that counting requires classification,
which in tum
requires judgments about inclusion and exclusion—who or what
is in or out.̂ ^
Again, the history of the UCR nicely illustrates this dilemma.
Crime categories
have been known to expand or contract depending on the
circumstances at
hand. Thus, a mtinicipality heavily dependent on tourism may
employ very
narrow definitions of what constitutes a criminal act, thereby
keeping crime
rates low. At budget time, a crime wave, based on expanding the
category of
what coimts as a crime, may be helpful in obtaining additional
resources.
The third decision point that illustrates the subjective nature of
a number is
the choice of a procedure for counting to one. '̂' This may seem
straightforward—one thing is one thing. But alas, one thing may
be made up of
multiple parts; thus, the question becomes whether to count the
parts separately
or as a unit. The UCR solves this dilemma by introducing a time
dimension in
17. Faith E. Lutze & Megan L. Symons, The Evolution of
Domestic Violence Policy
Through Masculine Institutions: From Discipline to Protection
to Collaborative Empowerment, 2
CRIMINOLOGY & PUB. POL'Y 319,322 (2003).
65. 18. Bennett Wolff, Expanding the Right of Sexual Privacy, 27
LOY. L. REV. 1279, 1281
(1981).
19. Michael D. Maltz, Crime Statistics: A Historical
Perspective, 23 CRIME &
DELINQUENCY 32, 33 (1977).
20. /¿a t33-34.
21. Id
22. DEBORAH STONE, POLICY PARADOX: THE ART OF
POLITICAL DECISION MAKING 146
(3d ed. 2002).
23. Crime Reporting in the Age of Technology, CRIMINAL
JUSTICE INFO. SERVS. (Fed.
Bureau of Investigation, U.S. Dep't of Justice, Washington,
D.C), 2000, at 12.
24. See supra note 14 and accompanying text.
96 SAINT LOUIS UNIVERSITY PUBLIC LAW REVIEW [Vol.
XXXI1:93
deciding how to count to one.̂ ^ For example, if a car is
vandalized as part of a
series of auto vandalisms that all occurred within the same,
designated time
period, then that incident is counted as one occurrence of
vandalism, even if
twenty cars were involved.̂ *
The above example introduces the fourth challenge to numeric
objectivity—the inability to understand what a number means
66. unless the
context in which the number was produced is also provided.^'
Put another way,
this is the siren song of numbers. Numbers appear to be
objective and not
tainted by context. Moreover, with too much contextual
description, the
elegant efficiency of numeric description is lost. This is a
particular problem in
attempting to use numbers either to provide transparency or to
improve
performance through accountability.^' There is a well-known
aphorism that
attests to the need for context:
The Govemment are very keen on amassing statistics—they
collect them, add
them, raise them to the nth power, take the cube root and
prepare wonderful
diagrams. But you must never forget that every one of those
figures comes in
the first instance from chowly dar (village watchman), who just
puts down
what he damn pleases.
This nineteenth century British waming continues to be of
relevance in
assessing modem, twenty-first century police intelligence units
charged with
analyzing data for the constmction of actionable law
enforcement
interventions.^ ' In a recent study of British police intelligence
analysis units,
one analyst is quoted as saying, "the quota of information that
we work on is as
good as the officer puts on there. If you look at the standards of
67. the [officers']
reports, they're absolutely appalling. You know, and they've got
the house
number wrong, they've got the beat wrong."''̂ The respondent
goes on to note
that these mistakes result in the identification of false hot
spots—^places of
peak crime occurrences. The important point of this, however,
is that the
analysts make the best of the data they are given and report a
de-contextualized
version of an event (a series of numbers). As the authors note,
those numbers
fail to convey either the fluid nature of a criminal occurrence,
and, thus, some
25. FED. BUREAU OF INVESTIGATION, U.S. DEP'T OF
JUSTICE, UNIFORM CRIME REPORTING
HANDBOOK 12 (2004).
26. Id
27. See supra note 14 and accompanying text.
28. 5eeUchida, Bridgeforth & Wellford, .yt/pra note 14, at 25.
29. See Martin Innes et al., 'The Appliance of Science?' The
Theory and Practice of Crime
Intelligence Analysis, 45 BRIT. J. CRIMINOLOGY 50-52
(2005).
30. JosiAH STAMP, SOME ECONOMIC FACTORS OF
MODERN LIFE 258-59 (1929).
31. Innes et al., supra note 29, at 50-52.
32. Id
33. Id
68. 2012] THE NUMBERS DILEMMA . 97
of its underlying causes, or the interpretive work of the analyzer
who fills in
gaps in the data so that it becomes "sensical."^''
The inability to convey context exacerbates a fifth characteristic
of
numbers, the objectification of evaluative criteria—i.e.,
disguising judgment as
measurement.̂ ^ Deborah Stone gives a wonderful example of
this process:
"Paul Samuelson's best-selling (economic) textbook declared in
its 1970
edition that ftill-employment was about 3.5 percent
tmemployment; by the
time of its 1985 edition, the natural rate of tmemployment had
grown to around
6 percent."^*
Of course, today, the Obama administration would celebrate a 7
percent
tmemployment rate. The tendency of numbers to disguise a
judgment as a
measurement makes it difficult to hold the entities reporting the
numbers
accountable and to ensure organizational transparency.'"
Finally, as noted previously, the act of counting something
confers a status
on, or suggests the importance of, the thing counted.''̂
Therefore, both the
counter and the coimted react to the process of numbering.^'
Organizationally,
69. this refers to producing good numbers whether the entity is a
police agency
employing COMPSTAT or a school system participating in high
stakes testing.
The number becomes a goal in and of itself—separate from what
the number
represents. This unfortunately became the experience of the
NYPD over the
course of implementing COMPSTAT."" In a study by John
Etemo and Eli
Silverman, field commanders quickly leamed that they needed
to look good
(i.e. have good numbers) when presenting at COMPSTAT
meetings if they
wanted to avoid public humiliation.'" This resulted first in
commanders
spending inordinate amounts of time constmcting data charts,
rather than
actually implementing the crime control sfrategies that the
analysis might have
suggested.''̂ Secondly, there was a strong temptation to
manipulate the
numbers.''̂ Thus, to tmly understand what a number means, it is
necessary to
know about the organization that produces the number. It is to
this issue that
we now tum.
34. Id
35. See supra note 14 and accompanying text.
36. STONE, supra note 22, at 169.
37. Id at 176-77.
38. Id at 178.
39. Wat 187.
40. See John A. Etemo & Eli B. Silverman, The New York City
70. Police Department's
Compstat: Dream or Nightmare?, 8 INT'L J. OF POLICE SCI. &
MGMT. 218, 219-26 (2006).
41. /¿at223.
42. Id at 228.
43. Id at 227.
98 SAINT LOUIS UNIVERSITY PUBLIC LAW REVIEW [Vol.
XXX11:93
II. ORGANIZATIONAL DYNAMICS AND THE PROBLEM OF
REFORM
Daniel Kahneman, in a popular sutnmary of current brain
research,
suggests that there are, at least metaphorically, two systems that
control
thitiking.'*'* System one is intuitive, quick thinking, and bases
conclusions on
mental frames developed through past experience.'" These
fVames or beliefs
are the templates used to explain the world as someone
encounters it.'*̂ System
two is slower and engages in deliberative assessments of
information.'*'
However, system two responses require effort, and according to
Kahneman,
system two often lacks the requisite effort."^ Therefore, if
system one presents
an explanation that seems to make sense and accords with past
experience,
system two will accept the conclusions of system one without
ftirther analysis
71. and cntique.
This observation is parallel to an observation made by Herbert
Simon
about organizational decision-making. Decision-making in
organizations
does not involve analyzing all of the information available, but,
instead, is
represented by a process of "satisficing."" The decision-maker
uses mies of
thumb, that is, how this current situation is like a previous
situation and what
the appropriate response was in that previous situation.'^ The
use of analogous
situations to make decisions mirror what has been discovered by
brain
research—neither as individuals nor as members of
organizations do we
optimize our decision-making, picking the best of all possible
altematives.
Instead, we satisfice, picking the altemative that best seems to
fit, without
expending a great deal of energy to review—in a systematic
way—all other
possible choices.'^
Of course, this approach, both individually and
organizationally, has some
advantages.''* It conserves energy and allows for efficient
decision-making."
In most cases, such decision-making processes or standard
operating
44. DANIEL KAHNEMAN, THINKING, FASTAND SLOW 20
(2011).
72. 45. Id at 105.
46. / ¿a t 21-22.
47. /¿a t49 .
48. / ¿ a t31 .
49. Id. at 24.
50. KEVIN B. SMITH & CHRISTOPHER W. LARIMER, THE
PUBLIC POLICY THEORY PRIMER
54 (2009).
51. Id
52. Id at 52.
53. Id at 52-53.
54. See id. at 53.
55. / ¿ a t 54.
2012] THE NUMBERS DILEMMA 99
procedures result in the individual or the organization moving
forward to
accomplish the task at hand.̂ ^
The disadvantage of this course of action is that often
individuals and
organizations do not engage in an analytical process that would
prevent major
problems.̂ ^ Higher order analytical routines are most often
employed after the
CO
fact—what went wrong and why? Post-mortems sometimes
result in changed
ways of behaving, but these soon become standard operating
procedures and
73. satisficing once again becomes the preferred mode of decision-
making.^'
This suggests that changing the core activity of an organization
and its
routines of decision-making is very, very difficult. This may be
particularly
tme in the public sector, where assessing performance is
hindered by the lack
of a concrete product or outcome.* '̂ Performance criteria are
instead infiuenced
by political considerations, budgets, and the public
In this kind of decision environment, it makes sense to ask
whether any
particular decision is based on technical or institutional criteria.
A decision
based on technical criteria responds to a change in the outside
environment. '̂'
A new product or service is required to keep the organization
competitive.^"
Therefore, stmctures are designed or redesigned to efficiently
and effectively
meet this demand.̂ ^ The organization that successfully
navigates its
environment's technical demands is rewarded with more
resources.*^ Failure to
do so jeopardizes resources.̂ ^ When an organization employs
technical
decision processes, it is engaging in a more sophisticated and
analytical routine
than simply satisficing.*^
A decision based on institutional criteria is much closer to the
satisficing
model or the system one brain model.*' The criteria for judging
74. the success of
56. KEVIN B. SMITH & CHRISTOPHER W. LARIMER, THE
PUBLIC POLICY THEORY PRIMER
54 (2009).
57. See id. at 53; K. A. Arehibald, Three Views of the Expert's
Role in Policymaking:
Systems Analysis, Incrementalism, and the Clinical Approach, 1
POL'Y SCI. 73, 76, 82 (1970).
58. See Archibald, supra note 57, at 76.
59. Id
60. / ¿ a t 8 2 .
61. SMITH & LARIMER, supra note 50, at 114-15.
62. See id.
63. See Sidney G. Winter, The Satisficing Principle in
Capability Learning, 21 STRATEGIC
MGMT. J., 981,982(2000).
64. See id.
65. See id.
66. See id. at 983; James Willis et al.. Making Sense of
COMPSTAT: A Theory-Based
Analysis of Organizational Change in Three Police
Departments, 41 LAW & SOC'Y REV. 147,
150(2007).
67. See Winter, supra note 63, at 982.
68. W. at 983.
69. See Willis et al.,5«pra note 66, at 151.
75. 100 SAINT LOUIS UNIVERSITY PUBLIC LAW REVIEW
[Vol. XXXI1:93
the organization involve neither efficiency nor effectiveness.'"
The criteria
instead involve judging the legitimacy of the organization based
on cultural
beliefs about how an organization should look and act.''
Perceived legitimacy
rather than efficacy is what counts.'^ Organizations operating in
this kind of
decision climate gain recognition and resoiu-ces by: conforming
to cultural
beliefs and expectations about what it is they are supposed to
do; and by
becoming isomorphic with other institutions in their
environment that have
been rewarded for particular behaviors."
Research describing the adoption of COMPSTAT suggests that
the
adoption has been spurred primarily by institutional
considerations.''' The rapid
deployment of COMPSTAT mirrors an almost fadlike
acceptance of the
process rather than a careñil and analytical investigation of how
the process
might adopt to a particular local situation or what pros and cons
such adoption
might entail.'^
Change spurred by the dynamic of institutional isomorphism is
likely to be
superficial. Core technologies and procedures are unlikely to be
impacted."
76. This certainly seems to be the case with COMPSTAT.'̂
In a study of three police departments that had adopted
COMPSTAT,
James Willis and his colleagues concluded that COMPSTAT
simply raised
reactive policing strategies to new levels. Police would respond
quickly to a
spike in crime, producing what the authors called a "whack-a-
mole" effect.""
Contrary to the assumptions of careful plarming and the
development of long
term strategies, COMPSTAT simply encouraged business as
usual, only even
more so. Further, while commanders felt responsible for crime
in their
districts, there was very little communication to beat officers
conceming
COMPSTAT trends—with the result that officers' daily routines
were not
impacted by the analysis derived from the gathered data. '̂
70. Id
71. Id.
72. Matthew J. Giblin, Structural Elaboration and Institutional
Isomorphism: The Case of
Crime Analysis Units, 29 POLICING: INT'L J. POLICE
STRATEGIES & MGMT. 643, 645 (2006).
73. Id
74. Willis et al., supra note 66, at 161.
75. Eterno & Silverman, supra note 40, at 219.
76. Matthew J. Giblin & George W. Burruss, Developing a
Measurement Model of
77. Institutional Processes in Policing, 32 POLICING: INT'L J.
POLICE STRATEGIES & MGMT., 351,
355 (2009).
77. See id
78. See Willis et al. supra note 66, at 152, 174-75.
79. / ¿a t 174.
80. / ¿a t 175.
81. / ¿a t 164-65.
2012] THE NUMBERS DILEMMA 101
Other studies have suggested that COMPSTAT also increased
the gulf
between the command culture and the street cop culture.̂ ^ For
example, in
New York, commanders took credit for dips in the crime rate,
while passing
blame to subordinates if the crime rate spiked in an area.̂ ^
The concept of work culture is an important element in trying to
understand why organizational change in general, and change
toward more
accountability and transparency, in particular, is difficult to
achieve. '̂' In public
sector agencies, the organizational dynamics just described,
together with the
cultural dynamics of such organizations, make the likelihood of
fundamental
change even more remote.̂ ^ We ttim to the role of cultural
dynamics next.
78. III. THE CULTURES OF MODERN DAY POLICING
Our understanding of the cultural complexity of the modem
police agency
has come a long way from the studies of the 1970s, which
concentrated
primarily on the culture of the street officer.̂ * By the end of the
decade, it was
clear that police organizations had at least two distinct culttires:
a street cop
culture and a management culttire.̂ ^ Today, it is recognized
that there are
multiple cultures within police agencies consisting not only of
beat officers and
higher ranked managers—but civilian cultures which span a
fairly large
professional hierarchy from clerical support staff, through 911
call-takers and
dispatchers, to highly trained and credentialed personnel in such
vmits as
research and development, human resource management, and
forensics.̂ ^
Although not studied quite as intensively as the police
subcultures, these
other subcultures are beginning to receive more attention.^' In a
study this
Author conducted on police call-takers, it was evident their
work culture was
influenced by the stress and strains of being sfreet level
bureaucrats with too
82. Eterno & Silverman, supra note 40, at 222-23.
83. W. at 224.
84. Linda Smircich, Concepts of Culture and Organizational
Analysis, 28 ADMIN. SCI. Q.
79. 339,346(1983).
85. Peter J. Robertson & Sonal J. Senevir, Outcomes of Planned
Organizational Change in
the Public Sector: A Meta-Analytic Comparison to the Private
Sector, 55 PUB. ADMIN. REV. 547,
548(1995).
86. See JEROME SKOLNICK, JUSTICE WITHOUT TRIAL 57
(4th ed. 2011).
87. ELIZABETH REUSS-IANNI, TWO CULTURES OF
POLICING: STREET COPS AND
MANAGEMENT COPS 1 (1983).
88. David A. Sklansky, Not Your Father's Police Department:
Making Sense of the New
Demographics of Law Enforcement, 96 J. CRIM. L. &
CRIMINOLOGY 1209, 1209-10, 1229-30
(2006).
89. See, e.g., WESLEY G. SKOGAN & MEGAN A.
ALDERDEN, NAT'L. INST. OF JUSTICE, JOB
SATISFACTION AMONG CIVILIANS IN POLICING l (2011),
available at http://www.nationalpolice
research.org/job-satisfaction-among-civilia/.
102 SAINT LOUIS UNIVERSITY PUBLIC LAW REVIEW
[Vol. XXX1I:93
few resources to meet increasing demands.'" Many of the
80. sfrategies they
adopted were attempts to gain greater confrol over their work
situation while
still trying to meet the demands for service emanating from 911
calls." Chief
among the sfrategies they employed was attempting to fit a
caller's demand
into a pre-existing definition of a situation that permitted the
dispatch of a
car.'^ Call-takers would often prompt callers as to how to
appropriately frame
their request.'^ A key conclusion of this study was that police
agencies do not
respond directly to a situation, but instead respond to an
organizationally
projected frame that takes ambiguous information and forms it
into an
understandable pattem to which the agency can then respond in
a routine
fashion.'"
A similar conclusion was reached in a study conducted by this
Author and
a colleague focused on a police research and development
(R&D) unit.'^ The
source of the ambiguous information for this unit was demands
for data that
would help other organizational units within the department
make a decision
about a problematic situation.'* Interestingly, the problem-
solving sequence for
the R&D tmit did not start with an analysis of the problem they
were supposed
to research, but with an analysis of the political situation and
the agenda of the
person making the request." Once there was consensus on the
81. real agenda and
on a solution that would meet both political and practical
realities, "research"
OR
was conducted to support this solution. Often, the research
consisted of
calling other police R&D units to see what they suggested in
similar
circumstances." This closed system of information processing
allowed the
organization to do what it would normally do anyway.
"Research" simply
reinforced standard operating procedures.
More recent research on police research units reinforces these
initial
observations.'"^ Martin Itmes and his colleagues adopt the term
"bricolage" to
describe the tasks performed by crime analysts in two British
police
90. See James F. Gilsinan, They is Clowning Tough: 911 and
the Social Construction of
Reality, 27 CRIMINOLOGY 329, 332 (1989).
91. See id
92. Id at 332-33.
93. See id. at 340.
94. Id at 341.
95. See generally James F. Gilsinan & J.R. Valentine, Bending
Granite: Attempts to Change
the Management Perspective of American Criminologists and
Police Reformers, 15 J. POLICE
82. SCI. & ADMIN. 196(1987).
96. SeeW. at 197-98.
97. Wat 198-99.
98. See/'rf. atl99.
99. W. at 197.
100. / ¿ a t 202, 203.
101. James F. Gilsinan & J.R. Valentine, iwpra note'95, at 203.
(1987).
102. Innes et al., supra note 29, at 54-55.
2012] THE NUMBERS DILEMMA 103
agencies.'"^ Crime analysts work with what they are given and
take messy,
contingent, and incomplete data to constmct an objective,
scientific product.'°'*
They put together the bric-a-brac of what they receive from
field reports and
constmct a coherent pattem of events."" The gaps in data are
filled in by what
everybody knows to be "tme" about how criminals operate.'"^
The resulting
product is a reproduction of the world that allows the
department to enact its
environment, i.e. project an image of a situation that allows for
the carrying out
of standard operating procedures.""
While units that can contribute to transparency and
accountability do not
operate in ways that can easily achieve either characteristic,