Adding More Police Is Unlikely to Reduce Crime: A Meta-Analysis of Police Agency Size and Crime Research

Full text

Translational
Criminology
THE MAGAZINE OF THE CENTER FOR EVIDENCE-BASED CRIME POLICY, GEORGE MASON UNIVERSITY
SPRING 2017
Inside this issue…
From the Directors
Integrating Research and Researchers
into the U.S. Marshals Service
New Partnership Uses Research
and Data to Improve Understanding
of Ofcer-Involved Shootings
The Caruth Police Institute:
A Model of Embedded Criminology
Moving Toward a Research-Based
Framework for School Policing
Adding More Police Is Unlikely
to Reduce Crime: A Meta-Analysis of
Police Agency Size and Crime Research
The Role of Consultants in
the Applied Research Process
Problem-Oriented Policing as a
Form of Translational Criminology
NIJ’s LEADS Programs
Police Research on the Front Lines
Research Note: Gaining Access to
Security Environments in Norway
Hot Off the Press
Promoting knowledge exchange to shape criminal justice research,
practice, and policy

John E. Eck YongJei Lee Nicholas Corsaro
1 is was not a subjective assessment. “Condence” is measured statistically
by the statistical variation in a study’s ndings (standard error): lower
variation means higher condence.
14 www.cebcp.org
Adding More Police Is Unlikely
to Reduce Crime: A Meta-Analysis
of Police Agency Size and Crime Research
BY JOHN E. ECK, YONGJEI LEE, AND NICOLAS CORSARO
John E. Eck is professor of criminal justice at the University
of Cincinnati. He has studied police eectiveness since 1977.
YongJei Lee is an advanced graduate student at the University
of Cincinnati specializing in crime patterns and policing.
Nicholas Corsaro is associate professor of criminal justice
at the University of Cincinnati specializing in police eectiveness
and operations.
Perhaps the most common question police ask researchers is
“How many police do we need?” In times of city budget cuts,
elected and police ocials worry that cutting police numbers
will drive crime up. Researchers have attempted to be helpful.
Since the 1970s, social scientists have tried to provide empirical
evidence on how police agency size—number of ocers—inuences
crime. rough 2014, there were 62 such studies examining police
agencies in the United States. Reading through these studies one
would be whipsawed by their conclusions. One study would show
hiring more police reduces crime, and the next study would show the
opposite. Over time, the sophistication of the research methods has
increased, yet the ndings continue to oscillate. After the year 2000,
however, studies seemed to show that adding police has a modest
crime reduction impact, though there were several contradictory
studies. Had researchers nally found the answer?
Because the question of police agency size is so important and
because there are so many studies with varying conclusions, we felt it
was important to systematically and rigorously examine all studies
and their ndings to determine if there was some general conclusion
we could draw from more than 40 years of scientic research.
How We Conducted Our Study
We looked at all the studies published in English that examined
police agencies in the United States and that attempted to determine
the relationship between numbers of police and crime volume. ere
were 62 such studies published between 1972 and 2013. Most of
these looked at several crimes, so these studies contained multiple
separate ndings. ere were 229 such ndings in these 62 studies
that we analyzed using meta-analytic methods.
What We Discovered
ere is no consensus among the studies and ndings about the
usefulness of adding more police. About 32 studies had at least one
nding that showed that adding police can reduce crime, and about
30 studies had no ndings suggesting that adding police would
reduce crime (some had ndings that adding police was associated
with more crime). is disagreement among studies is constant over
time, though since 2000 it appears there are more studies indicating
hiring police is benecial than there are studies saying the opposite.
is sort of analysis is called a vote count, because it treats every
study as equally valid: One simply tallies the votes to draw a conclu-
sion. Unfortunately, some studies are better than others, some
ndings are more valid than others, and within many studies there
were contradictory ndings. So another approach is needed.
When we combined the ndings from all the studies and adjusted
for condence in their conclusions,1 we found the eect on crime
of adding or subtracting police is miniscule and not statistically
signicant. Practically, this means police agency size has no impact
on crime. Further analysis showed that this is true over time, it is true
regardless of the type of statistical analysis used by the researchers,
and it is true regardless of how police force size vis measured.
Changing police agency strategy to address crime is far more
eective than hiring more ocers. We compare our ndings about
police force size to ndings from studies that have systematically
reviewed policing strategies. Hot-spots policing, focused deterrence,
and problem-oriented policing are more eective than hiring more
police. Research indicates that even developing a competent

Continued
Spring 2017 | TRANSLATIONAL CRIMINOLOGY 15
Adding More Police Is Unlikely
to Reduce Crime: A Meta-Analysis
of Police Agency Size and Crime Research
BY JOHN E. ECK, YONGJEI LEE, AND NICOLAS CORSARO
neighborhood watch program is more eective against crime
than hiring more police. is is shown in Figure 1. e height
of each bar represents the “eect size.”
Unfortunately, there is no way to translate eect size into some
easily interpretable description (e.g., we cannot say that an increase
of one point in eect sizes drops crime by some number). But we can
compare the relative eect sizes to get a judgement about how much
more eective some approaches to crime are, relative to others. And
we can give an indicator of how condent we can be about the eect
sizes. In Figure 1, we see that police force size has a tiny eect size
relative to all other eect sizes. It is less than one quarter of the next
largest eect size. e whiskers (dashed lines) on each bar show a 95
percent condence range. is means that we can be 95 percent sure
that the true eect size is between the top and bottom of the whisker.
Because the police force size whisker crosses the zero axis at the
bottom, we cannot be sure that the true eect size for adding and
subtracting police is not zero. is is what we mean by the eect size
is not signicant.
For the other four alternatives, the whiskers do not touch zero, so
there is less than a 5 percent chance the true eect size is zero. Because
the eect sizes of problem-oriented policing, neighborhood watch,and
Figure 2: The economics of adding more police and the effects
of changing strategy.
This curve shows the hypothetical relationship between police agency size
and crime. On the far left, if there were no police, crime would be high: adding
even a few police would have a large impact on reducing crime. As police
agencies get larger, adding the same number of police has less and less
impact on crime. So, at the right, adding more police has no detectable impact.
This downward sloping curve assumes that the strategy of policing does
not change (for example, the police always use random patrolling and
simply add more police). However, changing to a more effective strategy
(from A to B) makes police more effective, regardless of police force size.
Our ndings suggest that most police agencies in the United States
operate in this region. Modest uctuations in police agency size have
tiny, undetectable impacts on crime. Switching to a better policing
strategy has far more impact on crime.
Police force size
Strategy A
Strategy B
Crime
Figure 1: Changing policing strategy is far more effective than
adding police.
hot-spots policing all have overlapping whiskers, we must treat them as
equally eective. Focused deterrence is the most eective strategy,
though unlike its alternatives, it is extremely crime specic (addressing
gun killings by groups and some group-related drug dealing).
Why Our Results are Probably Right
ere are several reasons our results are probably correct. First, we
looked at all the research over four decades. We looked for systematic
changes in the ndings over time. Perhaps older ndings showed no
eect while recent ndings show more positive results. Contrary to
our rst impressions, there has never been a period of research where
the overall set of ndings were dierent from our general conclusion.
We did not cherry-pick the studies nor did we look at some small
fraction of the research. It is common for people, including police
and researchers, to select the studies they nd most agreeable and
highlight their results. We left no study out of our review, and we
treated all studies and all ndings the same way.
We also carefully looked at how dierent research methods
inuenced the ndings. Perhaps some research methods were better
at detecting the inuence of police force size on crime than others.
We found no evidence that dierent research methods or measures
changed the ndings.
Another reason we are probably correct is that economic theory
predicts that the usefulness of hiring an extra worker goes down as
more workers are hired. At some point, a business gains nothing
from hiring more workers. is well-established theory, illustrated
in Figure 2, is consistent with our ndings.
Mean Effect Size
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Problem-
Oriented
Neighborhood
Watch
Crime Hot
Spots
Focused-
Deterrence
Police Force
Size
0.030
0.126 0.160
0.604
0.184
Note: The types of crimes examined for the dierent strategies dier. Focused-deterrence evaluations looked at
violent crime and drug dealing. Crime hot spots studies include measures of disorder. Neighborhood watch studies
are influenced heavily by property crimes. Problem-oriented policing evaluations looked at a variety of dierent
types of crimes and disorder. Further, these strategies are not mutually exclusive—a police agency could use any
combination. In fact, examination of crime hot spots experiments show that when this strategy involves problem
solving, it is more eective at fighting crime than when it only involves patrolling or aggressive enforcement
(Braga et al., 2014).
Braga, A. A., Papachristos, A. V., and Hureau, D. M. (2014). The Eects of Hot Spots Policing on Crime:
An Updated Systematic Review and Meta-Analysis. Justice Quarterly 31(4): 633-63

e Caruth Police Institute, continued from page 11
basic academy curriculum to include more scenario-based learning
opportunities, consulting on a rewrite of eld ocer training
operations, and helping foster public/private partnerships to raise
funds for a study to assess DPD’s current personnel allocations and
to assist them with creating a more exible stang protocol. Still
other collaborations include a grant-funded initiative wherein CPI
and DPD are working together to rethink how mental health calls are
managed including evaluating police response, creating a multidisci-
plinary team with Dallas Fire and Rescue, and training ocers to
better recognize mental health problems in the community.
Moving forward, CPI’s mission will continue to be one that
promotes providing leadership professional development opportunities
for police and conducting meaningful research on policing issues that
matter. Yet, to remain relevant, the institute needs to continue to evolve
and change with the times and with environmental demands. As CPI
expands to provide its services to other law enforcement and criminal
justice personnel regionally and nationwide, it will focus on doing so
using a community-centric approach that encourages open and honest
discussion by police and the community about law enforcement
education and research. In this way, CPI will continue to be an
innovator in education and research and to play the unique role
in the police/academic paradigm it was originally intended to ll.
References
Braga, A. A. (2013). Embedded criminologists in police departments.
Ideas in American Policing, 17, 120.
Braga, A. A., & Davis, E. F. (2014). Implementing science in police
agencies: e embedded research model. Policing, 8(4), 294-306.
Buerger, M. E. (2010). Policing and research: Two cultures separated by
an almost-common language. Police Practice and Research, 135-143.
16 www.cebcp.org
ere is historical evidence that massive reductions in policing can
dramatically increase crime. In the United States, the best example is
the 1919 Boston police strike. So it is likely that going from zero to
many police will have an impact. is is shown in curve A on the left
in Figure 2. at is not the typical situation cities face, however.
Rather, police stang changes are likely to occur on the far right
of curve A. Our ndings are consistent with this interpretation.
To substantially impact crime, a police agency must change how it
does business. is is equivalent to shifting from curve A to curve B.
e comparison of eect sizes (Figure 1) support this argument.
ird, historically, jurisdictions do not change their police force
size relative to their populations very much. Despite political
statements about surging police, this almost never happens, based on
evidence we report. Most hiring is to replace ocers who leave the
agency, and most reductions are due to attrition. Most increases and
decreases are in dribs and drabs compared to the number of police
already employed. erefore, the typical magnitudes of changes to
police agency sizes are too small to make much dierence in crime
numbers or rates.
In short, we are condent in our ndings because they are
supported by economic theory, by the empirical data we reviewed,
and by common-sense interpretation of the reality of police hiring
practices. Any other interpretation of the impact of police on crime
must contradict theory, evidence, and common sense.
e principle limitation is that the outcome examined was the
impact on crime. Police agency size might have impacts on ocer
health and safety, on police uses of force, or on the quality of the
contacts with the public. e research we reviewed did not address
these or other outcomes.
What Policy Makers Can Do
• If crime is a problem, then change the policing strategy. Once
that is established, hire police necessary to carry out the strategy.
• Do not worry if your police agency shrinks a bit due to budget
shortfalls. Crime will not skyrocket. When tax revenues increase,
replace those ocers if they are needed.
• Base hiring decisions on how many police are needed to carry
out the functions of a police agency. If there are insucient
police to competently investigate serious crime, then hiring more
might be sensible. If there is evidence that ocer safety is imper-
iled because there are too few police, then consider hiring more.
If you want to have more police out of their cars in the community
engaged with the public, then consider hiring more.
• Consider hiring civilians as force multipliers for critical
services. Crime analysts, for example, are essential for most
advanced policing strategies. A few highly trained non-sworn
employees conducting crime and intelligence analysis may be
more useful than hiring more police, for example.
• Adopt evidence-based policing strategies shown to impact
crime and other public demands on the police. is is not
simple or easy, but it is eective. Supercial adoption, or tempo-
rary adoption, will not help. is needs to be undertaken with
the long-term objective of fundamentally changing the way
policing is carried out if it is to have a sustained impact on crime.
To read more about this study, see Lee, Y., Eck, J. E., & Corsaro, N.
(2016). Conclusions from the History of Research into the Eects of
Police Force Size on Crime—1968 through 2013: An Historical
Systematic Review. Journal of Experimental Criminology 12(3):
431-451.

Center for Evidence-Based Crime Policy
George Mason University
4400 University Drive, MS 6D12
Fairfax, VA 22030
www.cebcp.org