Content uploaded by Trevor Hugh Bennett
Author content
All content in this area was uploaded by Trevor Hugh Bennett on Sep 21, 2015
Content may be subject to copyright.
Systematic review
Does neighborhood watch reduce crime? A systematic review
and meta-analysis
TREVOR BENNETT*
Centre for Criminology, University of Glamorgan, Pontypridd, CF37 1DL, UK
*corresponding author: E-mail: thbennet@glam.ac.uk
KATY HOLLOWAY
Centre for Criminology, University of Glamorgan, Pontypridd, UK
DAVID P. FARRINGTON
Institute of Criminology, University of Cambridge, Cambridge, UK
Abstract. Neighborhood watch grew out of a movement in the USA during the late 1960s that
promoted greater involvement of citizens in the prevention of crime. Recent estimates suggest that over
a quarter of the UK population and over 40% of the US population live in areas covered by
neighborhood watch schemes. The current paper presents the results of a recent systematic review
of evaluations of neighborhood watch. The main findings of the narrative review were that about half of
the schemes evaluated showed that neighborhood watch was effective in reducing crime, with most
of the other evaluations having uncertain effects. The main findings of the meta-analysis were that 15 of
the 18 studies provided evidence that neighborhood watch reduced crime. While the results of the
review are encouraging, it was concluded that more high-quality research needs to be done to help
explain why study variations exist.
Key words: crime reduction, effectiveness, meta-analysis, neighborhood watch, systematic review
Introduction
Neighborhood watch (NW) grew out of a movement in the USA that promoted
greater involvement of citizens in the prevention of crime (Titus 1984). It is also
known as block watch, apartment watch, home watch, citizen alert and community
watch. One of the first evaluations of neighborhood watch programs in the USA
was of the Seattle Community Crime Prevention Project launched in 1973 (Cirel et
al. 1977). One of the first evaluations of neighborhood watch schemes in the UK
was of the Home Watch program implemented in 1982 in Cheshire (Anderton
1985).
Since the 1980s, the number of neighborhood watch schemes in the UK has
expanded considerably. The report of the 2000 British Crime Survey estimated that
over a quarter (27%) of all households (approximately 6 million households) in
England and Wales were members of a neighborhood watch scheme (Sims 2001).
This amounted to over 155,000 active schemes. A similar expansion has occurred
in the USA. The report of The 2000 National Crime Prevention Survey (National
Crime Prevention Council 2001) estimated that 41% of the American population
lived in communities covered by neighborhood watch. The report concluded, BThis
Journal of Experimental Criminology (2006) 2:437Y458 #Springer 2006
DOI: 10.1007/s11292-006-9018-5
makes Neighborhood Watch the largest single organized crime prevention activity
in the nation^(p. 39). Considering such large investments in terms of resources
and community involvement, it is important for researchers to ask whether
neighborhood watch is effective in reducing crime.
The theory of neighborhood watch
The most frequently suggested mechanism by which neighborhood watch is
supposed to reduce crime is by residents looking out for suspicious activities and
reporting these to the police (Bennett 1990). The link between reporting and crime
reduction is not usually elaborated in the literature. However, it has been argued
that visible surveillance might reduce crime as a result of its deterrent effect on the
perceptions and decision making of potential offenders (Rosenbaum 1987). Hence,
watching and reporting might deter offenders if they are aware of the likelihood of
local residents reporting suspicious behavior and if they perceive this as increasing
their risks of being caught.
Neighborhood watch might also lead to a reduction in crime through the
reduction of opportunities for crime. One method discussed in the literature is
through the creation of signs of occupancy. Some of the methods by which
members of neighborhood watch schemes might create signs of occupancy
were discussed in the report of the Seattle scheme (Cirel et al. 1977). These
include removing newspapers and milk from outside neighbors_homes when
they are away, mowing the lawn, and filling up trash cans. The way in which
signs of occupancy might reduce crime might be through the effect that this has
on the perceptions of potential offenders in terms of their likelihood of getting
caught.
Neighborhood watch might also lead to a reduction in crime through the various
mechanisms of social control. Informal social control is not one of the mechanisms
for reducing crime stated in the publicity material of these schemes. Nevertheless,
they might indirectly serve to enhance community cohesion and increase the
ability of communities to control crime (Greenberg et al. 1985). Informal social
control can affect community crime through the generation of acceptable norms of
behavior and by direct intervention by residents.
It is also possible that neighborhood watch schemes might reduce crime through
enhancing police detection. Neighborhood watch might serve to increase the flow
of useful information from the public to the police. An increase in information
concerning crimes in progress and suspicious persons and events might lead to a
greater number of arrests and convictions and result (when a custodial sentence is
passed) in a reduction in crime through the incapacitation of local offenders
(Bennett 1990).
It is also feasible that neighborhood watch might reduce crime through the other
components of the program package. It has been argued that property marking might
lead to a reduction in crime as a result of making the disposal of marked property more
difficult (Laycock 1985). This might reduce offending rates if potential offenders
viewed marked property as increasing the risk of detection. Home security surveys
TREVOR BENNETT ET AL.438
might lead to a reduction in crime as a result of making it physically more difficult
for an offender to enter the property (Bennett and Wright 1984).
Program elements
Neighborhood watch is often implemented as part of a comprehensive package.
The typical package is sometimes referred to as the Bbig three^and includes
neighborhood watch, property marking and home security surveys (Titus 1984).
Some programs include other elements, such as a recruitment drive for special
constables, increased regular foot patrols, citizen patrols, educational programs for
young people, auxiliary police units, and victim support services.
Neighborhood watch schemes vary in terms of the size of the area covered.
Some of the earlier schemes in the USA and the UK were based on areas covering
just a few households. More recent schemes sometimes cover many thousand
households (Knowles et al. 1983). One of the smallest schemes included in the
review was the Bcocoon^neighborhood watch program in Rochdale in England,
covering just one dwelling and its immediate neighbors (Forrester et al. 1990). One
of the largest was the Manhattan Beach neighborhood watch scheme in Los
Angeles, covering a population of over 30,000 residents (Knowles et al. 1983).
Neighborhood watch schemes can be both public and police initiated. Schemes
launched in the UK during the early period of a program tended to be police
initiated (e.g., the early neighborhood watch schemes in London). More recently,
neighborhood watch schemes have been launched mainly at the request of the
public. Some police departments continue initiating their own schemes, even when
the program is fully developed. A program implemented in Detroit, for example,
developed a section of police-initiated schemes in order to promote neighborhood
watch in areas that were unlikely to generate public-initiated requests (Turner and
Barker 1983).
In the USA, block watches are usually run by a block captain who is responsible
to a block co-ordinator or block organizer. The block co-ordinator acts as the
liaison person to the local police department. Neighborhood watch schemes in the
UK often include street co-ordinators (equivalent to block captains) and area co-
ordinators (equivalent to the block organizer). There is little information in the
literature on the numbers and types of neighborhood watch meetings. The evidence
that does exist suggests that some schemes have public meetings that involve all
the residents participating in the scheme, while others have meetings that involve
only the organizers of the scheme (Bennett 1990).
The funding of neighborhood watch schemes is nearly always a joint venture
between the local police department and the scheme members through their fund-
raising activities. The relative contribution of the two sources varies considerably.
Some schemes in the USA are provided with no more than an information package
from the local police. Others are provided with police facilities for the production
of newsletters and the use of police premises for meetings (Turner and Barker
1983). Apart from police funding, the majority of schemes are encouraged to raise
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 439
some funds from other sources, such as voluntary contributions, local businesses,
and the proceeds of fe
ˆtes, and raffles.
Previous reviews
There are several previous reviews that include evaluations of neighborhood watch
programs. One of the earliest conducted in the USA was by Titus (1984), who
summarized the results of nearly 40 community crime prevention programs. Most
of these included elements of neighborhood watch. The majority of studies was
conducted by police departments or included data from police departments. Nearly
all found that neighborhood watch areas tended to have relatively low levels of
crime. However, most of the evaluations were described as Bweak^in terms of
their ability to guard against threats to validity.
Another review of the literature looked mainly at community watch programs in
the UK (Husain 1990). This study reviewed the results of nine existing evaluations
and concluded that there was little evidence that NW prevented crime.
One of the most recent reviews of the literature on the effectiveness of
community watch programs selected only evaluations with the strongest research
designs (Sherman 1997). The author included only studies that used random
assignment or studies that monitored both watch areas and similar comparison
areas without community watch. The review found just four evaluations that
matched these criteria. The results of these evaluations were largely negative. The
author concluded, BThe oldest and best-known community policing program,
Neighborhood Watch, is ineffective at preventing crime^(pp. 8Y25). Similar
conclusions were drawn in the later update of this report (Sherman and Eck 2002).
Methods
Criteria for inclusion of studies
The criteria for inclusion of studies in the current review were based on three broad
categories: the type of intervention, the type of outcome and the type of evaluation
design.
The main aim of the type of intervention criterion was to include studies that
evaluated neighborhood watch schemes. In practice, this is more difficult to
determine than it might seem, as neighborhood watch schemes are often implemented
alongside other program elements. As mentioned, the most common other elements
are property marking and security surveys. Neighborhood watch is also sometimes
implemented as part of broader area improvements and may exist alongside other
unrelated crime reduction initiatives. Hence, the selection criteria relating to the type
of intervention included only the following program types and combinations:
1. stand-alone neighborhood watch schemes (comprising solely a watch component)
2. neighborhood watch schemes that include Bthe big three^(neighborhood watch,
property marking and security surveys), so long as there was a watch component
TREVOR BENNETT ET AL.440
3. neighborhood watch schemes that include two components of Bthe big three^,
so long as there was a watch component
4. comprehensive programs that include neighborhood watch (any version of the
above) and other unrelated schemes (such as environmental improvements), so
long as the independent effects of the neighborhood watch component were
identified in the evaluation or neighborhood watch was the major component of the
program
The main aim of the type of outcome criterion was to focus the evaluation on
crime outcomes. We were not interested, in this review, in determining the impact
of neighborhood watch on fear of crime, residents_satisfaction with their area, or
policeYcommunity relations. Instead, we sought to determine whether neighbor-
hood watch succeeded in meeting its primary objective of reducing residential
burglary and related neighborhood crimes. The types of crimes included in the
review were:
1. crimes against residents
2. crimes against dwellings
3. other (street) crimes occurring in residential areas
The aim of the type of evaluation design criterion was to include studies of the
highest quality in terms of the research methods used. The main method for
selecting rigorous evaluations was based on the Maryland Scientific Methods Scale
(SMS) (Sherman 1997; Sherman and Eck 2002). This is a 5-point scale ranging
from level 1 (the weakest design) to level 5 (the strongest design) in terms of
overall internal validity. Sherman and Eck (2002) argue that evaluations should be at
least level 3 in order to make it possible to conclude, with a reasonable level of
certainty, that the program worked. This review of evaluations uses level 3 as the
minimum acceptable for inclusion in the review. This level requires that the evaluation
must comprise at least a comparison of one or more experimental units and one or
more comparable control units over time. Hence, the minimum requirement for
inclusion of evaluations in this review of neighborhood watch is that they are based on
before and after measures of crime in experimental and comparison areas.
Search strategy
The main goal of the search strategy was to be as exhaustive as possible in
obtaining relevant evaluations. This meant that we were willing to include
published and unpublished literature studies, with no restriction on country of
origin or source sector (e.g., academic, government, policy, or voluntary). We
could only include studies written in English, as we had no research funds for
translation. We used the following search strategies for locating studies:
1. Searches of on-line databases (especially for reports and articles).We conducted
searches of the following electronic databases and websites: International
Bibliography of the Social Sciences (IBSS), Web of Science, Criminal Justice
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 441
Abstracts, National Criminal Justice Reference Service Abstracts, Sociological
Abstracts, Psychological Abstracts (PsycINFO), Social Science Abstracts, UK
Government Publications (Home Office), Dissertation Abstracts (ASSIA),
ProQuest, and C2-SPECTR.
2. Searches of on-line library catalogs (especially for books). These included the
Radzinowicz Library, University of Cambridge and the Rutgers University
Library.
3. Searches of reviews of the literature on the effectiveness of neighborhood watch
in preventing crime. These included reviews by Titus (1984), Husain (1990) and
Sherman (1997).
4. Searches of bibliographies of publications on neighborhood watch. These
included the references in all publications selected as eligible for the review.
5. Contacting leading researchers. These included Dennis Rosenbaum and Wesley
Skogan, who worked on one of the largest evaluations of neighborhood watch.
We used the following search terms when searching online databases:
Bneighbourhood watch^,Bneighborhood watch^,Bstreet watch^,Bblock watch^,
Bapartment watch^,Bhome watch^,Bcommunity watch^,Bhome alert^,Bblock
association^,Bcrime alert^,Bblock clubs^,Bcrime watch^,Bbig three^.
Eligible publications
The above process resulted in the identification of a total of 1,595 publications.
Overall, 335 of these were selected by the researchers as potentially relevant
evaluations. The titles and abstracts of these studies were scrutinized to determine
whether they were evaluations and whether they were duplicates of studies already
obtained. In total, 225 of the 335 were identified as unique publications. An
attempt was then made to obtain a copy of each of the selected publications. This
resulted in 137 (61%) studies being obtained. The main reasons for not obtaining
publications were that they could not be located following various attempts to
obtain them by inter-library loan, through the internet, or by contacting the authors.
Another reason for the losses was that many of the evaluations were included in
reports published by police departments and had not been deposited in copyright
libraries that hold copies of all national publications.
Of those publications obtained, 107 were defined as ineligible. The main reason
for ineligibility was that the publication did not include an outcome evaluation of
neighborhood watch (n= 60). For example, they were descriptions of neighborhood
watch programs or process evaluations. A further 27 studies were excluded on the
grounds of methodological quality, for example there was no control area. Overall,
the search resulted in 30 eligible publications covering 19 unique studies that
evaluated 43 separate neighborhood watch schemes. Two studies covering seven
evaluations were excluded on the grounds that the results were presented in
graphical form only (Mukherjee and Wilson 1988 and Husain 1990). This left 17
studies covering 36 evaluations that were included in the narrative review. A sub-
sample of 12 studies covering 18 evaluations watch was included in the meta-
TREVOR BENNETT ET AL.442
analysis. Studies were included in the meta-analysis if the data were presented in a
form that could be used for calculating effect sizes.
Coding study characteristics
Studies determined as eligible for inclusion in the systematic review were coded,
and the data were entered into a database. One researcher entered the data, and this
was then checked for accuracy by a second researcher. Any discrepancies in
coding were discussed, and an agreement was reached on the correct values to be
used. The database included basic information about the study (e.g., author(s), year
of publication, country of study), details of the program (e.g., type of program,
program elements, size of area, type of area), research design (e.g., type of design,
sample size, length of follow-up period, type of comparison areas), and outcomes
(type of offence, pre- and post-test measures for experimental and control areas). In
some cases evaluations produced multiple outcome measures. This occurred when
there were multiple methods of measuring the same outcome and when the same
outcome was measured at multiple points in time. In these cases a method of
selecting outcomes was established. When multiple outcome measures were
provided (e.g., multiple outcome measures of crime) we listed the results for each
measure. However, any single analysis was based on only one of these measures.
The measure chosen was based on a system for prioritizing the results (i.e.,
burglary first, followed by all property crimes and then all crimes). When the same
outcome was measured at multiple points in time, we selected the year before and
the year after the implementation of the scheme as the basis for our analyses.
Failing this, we chose other periods in accordance with the above priority system
(i.e., periods nearest to the point of implementation were chosen first).
Statistical procedures
In order to carry out a meta-analysis of the effects of neighborhood watch, one
needs a comparable effect size measure for each evaluation, together with its
variance (see Lipsey and Wilson 2001). All evaluations included in the review
employed the same research design (pre-test and post-test measures for
experimental and control areas). The majority of evaluations used police-recorded
data to provide an outcome measure of crime, while the remainder used self-report
victimization data. The outcome measure in each study was the number of crimes
(i.e., burglaries, property crimes, or all crimes, in that order) recorded by the police
or the proportion of residents in an area who had been victimized.
The most appropriate effect size for data based on proportions is the Bodds
ratio^(OR)
1
. This is calculated as shown in the following table:
Before intervention After intervention
Experimental a b
Control c d
where a, b, c, d are numbers of crimes and OR=ad/bc
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 443
The chance value of the OR is 1.0. To the extent that the OR exceeds 1.0, it
might be concluded that the intervention (i.e., neighborhood watch) was beneficial.
To the extent that the OR falls below 1.0, it might be concluded that the
intervention was damaging. It is possible that some schemes might artefactually
serve to increase the number of recorded crimes, for example, if increased
surveillance led to an increase in the number of crimes reported to the police. The
method of determining the mean effect size for groups of studies varied, depending
on whether the results were based on police-recorded crimes or on survey results
on the prevalence of victimization.
The analysis based on police-recorded crimes was adjusted slightly to deal with
the problem of possible Bover-dispersion^(i.e., greater than expected variance).
The variance of the OR is usually calculated from its natural logarithm (LOR). In
order for a summary effect size in a meta-analysis to be produced, each effect size
is usually weighted by the inverse of its variance (1/V). This estimate of the
variance is based on the assumption that total numbers of crimes (a, b, c, d) have a
Poisson distribution. If the number of crimes has a Poisson distribution, its
variance should be the same as its mean. However, the large number of changing
extraneous factors may cause over-dispersion; that is, where the variance of the
number of crimes, VAR, may exceed the number of crimes, N. Hence, the standard
formula for V(LOR) was multiplied by an over-dispersion factor, D, where
D¼VAR=N
Farrington et al. (2005; The effects of CCTV on crime: Meta-analysis of an
English national quasi-experimental multi-site evaluation. University of Cambridge,
unpublished paper) estimated VAR from monthly numbers of crimes and found the
following equation:
D¼:0008*Nþ1:2
D increased linearly with N and was correlated (0.77) with N. The median number of
crimes in their study was 760, suggesting that the median value of D was about 2.
However, Farrington et al. (2005; The effects of CCTV on crime: Meta-analysis of
an English national quasi-experimental multi-site evaluation. University of Cam-
bridge, unpublished paper) argued that this is an overestimate because the monthly
variance is inflated by seasonal variations, which do not apply to N and VAR.
Nevertheless, in order to obtain a conservative estimate, V(LOR), calculated from the
usual formula above, was doubled in all cases involving police-recorded crime data.
2
This adjustment corrects for over-dispersion within studies, not for heterogeneity
between studies.
Results
Description of evaluations
Table 1gives an overview of the characteristics of the 43 evaluations included in
the narrative review. All were conducted during the period 1977 to 1994. No
TREVOR BENNETT ET AL.444
eligible evaluations were found after the mid-1990s. About half were conducted in
North America, and about half in the UK. There was one study from Canada and
one study from Australia. Most evaluations (31) concerned a neighborhood watch
scheme with no other program elements, but 12 assessed a comprehensive
package, including property marking and/or security surveys. Only a minority of
evaluations (11) was based on any kind of matching of comparison areas with
experimental areas. The remainder used either similar nearby areas or larger areas
with no matching at all (e.g., the remainder of the police force area). In some cases
the comparison area was the whole police force, including the neighborhood watch
area. The majority of evaluations (25) used police-recorded crime data as the main
outcome measure, while 18 used victimization survey data. The final section of the
table shows that 26 of the evaluations were categorized as published and 17 as not
published. Evaluations were defined as published if they were reported in a book,
journal or official government report, as those were likely to have been externally
reviewed before distribution. Unpublished evaluations included police reports and
reports from survey research companies, which were unlikely to have been
externally reviewed before distribution.
Narrative review
One reason for presenting the results of the narrative review is that it includes more
studies (n= 36) than does the meta-analysis (n= 18). Only a minority of the
evaluations provided the data needed to calculate an effect size. It is important to
know whether there are any differences in results between studies included in or
excluded from the meta-analysis.
Table 1. Description of evaluations.
Parameter Item n
Year of study 1977Y1988 20
1989Y1994 23
Country USA 21
UK 20
Other 2
Type of scheme NW only 31
NW+ 12
Type of comparison Matched 11
Non-matched (similar size) 18
Non-matched (remainder of force) 14
Data source Police data 25
Survey data 18
Published Published 17
Not published 26
n= 43
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 445
Table 2. Outcome effectiveness of neighborhood watch based on studies included in the narrative review (Exp experimental, Con control, PD police-recorded crime
data, SD survey data).
Author/publication date of main report
Used in
meta-analysis
Data
source Published
Result percentage
crime difference
Relative percentage change
(jfavorable, + unfavorable Outcome
Anderton (1985) * PD No Exp j10%; Con +3% j13% Positive
Bennett (1990) (1) * SD Yes Exp j22%; Con j28% +6% Uncertain
Bennett (1990) (2) * SD Yes Exp +37%; Con j28% +65% Negative
Bennett and Lavrakas (1989) (1) SD Yes Percentage not available Sig. negative Negative
Bennett and Lavrakas (1989) (2) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (3) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (4) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (5) SD Yes Percentage not available Sig. positive Positive
Bennett and Lavrakas (1989) (6) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (7) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (8) SD Yes Percentage not available Sig. negative Negative
Bennett and Lavrakas (1989) (9) SD Yes Percentage not available n.s. Uncertain
Bennett and Lavrakas (1989) (10) SD Yes Percentage not available n.s. Uncertain
Cirel et al. (1977) * SD Yes Exp j61.3%; Con j4.0% j57% Positive
Forrester et al. (1988) * PD Yes Exp j38%; Con +1% j39% Positive
Henig (1984) * PD No Exp j100%; Con j35% j65% Positive
TREVOR BENNETT ET AL.446
Hulin (1979) PD Yes Exp j26%; Con +10% j36% Positive
Jenkins and Latimer (1986) (1) * PD No Exp j25%; Con +2% j27% Positive
Jenkins and Latimer (1986) (2) * PD No Exp +1100%; Con +20% +1,080% Negative
Jenkins and Latimer (1986) (3) * PD No Exp j75%; Con j29% j46% Positive
Jenkins and Latimer (1986) (4) * PD No Exp j71%; Con j25% j46% Positive
Knowles et al. (1983) PD Yes Exp j28%; Con +12.9% j41% Positive
Latessa and Travis (1987) PD Yes Exp j11%; Con j2% j9% Positive
Lewis et al. (1988) (1) SD Yes Exp j21%; Con j11% Sig. positive Positive
Lewis et al. (1988) (2) SD Yes Exp +23%; Con j27% Sig. negative Negative
Lewis et al. (1988) (3) SD Yes Exp +10%; Con j18% Sig. negative Negative
Lewis et al. (1988) (4) SD Yes Percentage not available n.s. Uncertain
Lewis et al. (1988) (5) SD Yes Percentage not available n.s. Uncertain
Lowman (1983) * PD Yes Exp j33%; Con 0% j33% Positive
Matthews and Trickey (1994a) (1) * PD No Exp j20%; Con j17% j3% Uncertain
Matthews and Trickey (1994b) (2) * PD No Exp +24%; Con +45% j21% Positive
Research & Forecasts Inc. (1983) * PD No Exp j48%; Con j4% j44% Positive
Tilley and Webb (1994) (1) * PD Yes Exp j41%; Con j11% j30% Positive
Tilley and Webb (1994) (2) * PD Yes Exp 0%; Con +12% j12% Positive
Tilley and Webb (1994) (3) * PD Yes Exp j13%; Con +12% j25% Positive
Veater (1984) * PD No Exp j25%; Con +31% j56% Positive
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 447
Summaries of the results of the 36 evaluations included in the narrative review are
shown in Tables 2and 3. Table 2provides information about each study, including
the name of the authors, whether it was used in the meta-analysis, the data source,
and the outcome results. One of the aims of the summary table is to show whether
the study found that neighborhood watch had a positive effect (a greater reduction
or smaller increase in crime than the comparison area), an uncertain effect, or a
negative effect (a smaller reduction or greater increase in crime than the
comparison area). This was calculated from the published results of the study in
one of two ways, depending on whether the results were numbers or coefficients.
When the results were presented as raw numbers of crimes or as percentages, a
relative change score was calculated showing the difference in the change in the
experimental area compared with the change in the comparison area. In order to
assess whether the difference was noteworthy, a coding rule was devised, which
defined a relative reduction in the neighborhood watch area of 9% or more as a
positive effect and a relative increase in the neighborhood watch area of 10% or
more as a negative effect. The difference in the two percentages reflects an increase
in proportion of 10% (to 1.10) and a decrease in proportion of 10% (to 0.91). When
the results were presented as adjusted means or as regression coefficients, the
significance and direction of the effect were presented (e.g., a significant positive
effect, no significant effect or a significant negative effect). It was possible, using
these two methods, to produce outcome estimates for 36 of the 43 studies; these 36
are shown in Table 2. The remainder presented results in the form of graphical
outputs only, which could not be used to produce reliable estimates of effect size.
A summary of the results is shown in Table 3. The narrative analysis concluded
that 53% of evaluations (19 studies) showed that neighborhood watch had a
significant desirable effect on crime. The remainder showed an uncertain effect (11
studies) or an undesirable effect (six studies). Overall, it can be concluded that the
results are mixed, with almost as many evaluations providing evidence that
neighborhood watch was effective as those providing uncertain evidence or
evidence of an unfavorable effect. However, this pattern of results is concordant
with the hypothesis that neighborhood watch has a small but desirable effect.
The second reason for conducting this analysis was to determine whether there
were any important differences between those studies included in the meta-analysis
and those excluded. Table 3shows that there were significant differences (chi
squared = 9.38; 2 df, PG0.01) in the results of the two groups of studies. Over
Table 3. Summary of outcome effectiveness of neighborhood watch based on studies included in the
narrative review.
Review
Positive
effect Uncertain
Negative
effect Total
Significance
of difference
Studies in the meta-analysis 14 (78%) 2 (11%) 2 (11%) 18 (100%)
Studies not in the meta-analysis 5 (28%) 9 (50%) 4 (22%) 18 (100%)
Total studies 19 (53%) 11 (31%) 6 (17%) 36 (100%) PG0.01**
**Chi-square test.
TREVOR BENNETT ET AL.
448
three-quarters of studies included in the meta-analysis (14 of 18) showed a positive
effect of neighborhood watch on crime, compared with approximately one-quarter
(5 out 18) of those not included in the meta-analysis. Hence, this difference in the
results is important and needs to be taken into account when arriving at an overall
conclusion about the effectiveness of neighborhood watch.
Meta-analysis
Meta-analyses have a number of advantages and disadvantages over narrative
reviews. The main advantages are that they can standardize the results across studies
to produce a uniform effect size for each individual study and a weighted mean effect
size for groups of studies. The main disadvantage of meta-analyses is that they are
often based on a sub-set of all eligible studies which provide the necessary data for
the calculations. By comparison, the main advantages of narrative reviews are that
they can include a larger number of eligible studies and the results can be presented
in a more easily interpretable way. The main disadvantage is that such reviews may
be biased, and it is often not possible to summarize accurately the strength of effect
across studies. In practice, it is useful to include the results of both.
A summary of the results of the 18 evaluations included in the meta-analysis is
given in a forest plot in Figure 1. The figure shows that 15 evaluations had an OR
greater than 1 (showing a favorable effect on crime) and three had an OR less than
1 (showing an unfavorable effect). Hence, in the majority of evaluations,
neighborhood watch was associated with a desirable change in crime (a greater
reduction or a smaller increase).
One of the aims of the meta-analysis is to calculate a weighted mean effect size
(here, the OR) to determine, overall, how well neighborhood watch works. There
are two commonly used methods of calculating the weighted mean effect size. The
fixed effects (FE) method assumes that all measured effect sizes randomly vary
about the mean. In estimating this mean, each effect size is weighted by the inverse
of its variance (1/VAR), so that studies based on larger samples are given greater
weight than those based on smaller samples. However, the studies may not all be
drawn from the same sampling distribution of effect sizes. One method of
addressing the problem of heterogeneity in effect sizes is to use the random effects
(RE) model. This assumes that the variance of the effect size is the sum of two
components, one reflecting random variation about the mean and the other
reflecting the variation of effect size over studies. The RE method minimizes
heterogeneity by adding a constant to the variance of each effect size (for the
formula, see Lipsey and Wilson 2001, p. 119). When this is done, studies with
larger sample sizes no longer have such a disproportionate influence on the mean
effect size. Each study has a more similar weighting, which seems undesirable,
since larger studies have narrower confidence intervals about the mean, which
estimates the mean of their sampling distribution more accurately. As there are
advantages and disadvantages to using FE and RE models, it is usually considered
good practice to report findings for both. In the following, we adopt this approach
and present the results using both FE and RE models.
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 449
Figure 1shows that the weighted mean OR for the 18 evaluations combined was
1.19 using the FE model. This was statistically significant (z = 7.25, PG0.0001).
An OR of 1.19 can be interpreted to mean that crime increased by 19% in the
control area compared with the experimental area or that it decreased by 16%
(using 1/OR) in the experimental area compared with the control area. However,
the 18 studies were significantly heterogeneous according to the Q statistic
(Q = 35.72, 17 d.f., PG0.005). As the FE model is based on the assumption that the
evaluations are homogeneous and drawn from the same sampling distribution of
effect sizes, it is recommended, when this assumption is not met, that another
method be used for conducting the analysis (Lipsey and Wilson 2001).
There are two main ways of dealing with the problem of heterogeneity. One is
to split the evaluations into distinct homogeneous groups and conduct meta-
analyses separately for each group. The other is to use the RE method. In the
former case, this method would lose the overall findings for all studies combined,
and the analysis would be limited to specific sub-groups. In the latter case small-
sample studies would be given a similar importance as large sample studies. As we
were interested in presenting the results for the entire sample, the RE model was
Jenkins and Latimer (1986) [3]
Henig (1984)
Jenkins and Latimer (1986) [4]
Cirel et al. (1977)
Research and Forecasts Inc. (1983)
Veater (1984)
Forrester et al. (1988)
Tilley and Webb (1994) [1]
Lowman (1983)
Jenkins and Latimer (1986) [1]
Tilley and Webb (1994) [3]
Matthews and Trickey (1994b)
Anderton (1985)
Tilley and Webb (1994) [2]
Matthews and Trickey (1994a)
Bennett (1990) [1]
Bennett (1990) [2]
Jenkins and Latimer (1986) [2]
Mean FE
Mean RE
Mean MVA
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
0.00 0.01 0.10 1.00 10.00 100.0
0
Weighted mean FEOR=1.19; p=.0000; Q=35.72; p<.005
Weighted mean REOR=1.36; p=.0003; Q=13.35; p=ns
Weighted mean MVAOR=1.19; p=.0006; Q=17.0; p=ns
Notes: Odds ratios and confidence intervals shown on a lo
g
arithmic scale.
Figure 1. The effectiveness of neighborhood watch (FEOR fixed effects odds ratio, REOR random
effects odds ratio, MVAOR multiplicative variance adjustment odds ratio).
TREVOR BENNETT ET AL.
450
chosen as the better method of addressing this problem. The weighted mean OR for
the 18 evaluations combined was 1.36 using the RE model (statistically significant
at PG0.0004). An odds ratio of 1.36 means that crime increased by 36% in the
control area compared with the experimental area or decreased by 26% in the
experimental area compared with the control area.
The different results obtained by the FE and RE methods is troubling. When faced
with this problem in a meta-analysis of the effect of CCTV on crime, Jones (2005;
Measuring effect size in area-based crime prevention research. Unpublished thesis,
University of Cambridge) compared four other statistical methods of estimating a
weighted mean effect size. Three produced the same results, which seemed the most
defensible. The simplest of these methods was a random effects model with a
multiplicative variance adjustment (MVA) rather than the additive variance adjustment
(AVA) used in the usual RE model. Unlike the FE model, the MVA model exactly
adjusts for both over-dispersion and heterogeneity and exactly fits the data.
The MVA model generally yields the same weighted mean effect size and
confidence interval whether the true V(LOR) is two or three times (in this case, up
Table 4. Mean effect size based on the fixed effect method and two versions of the random effect
method.
Method
No. of
studies OR CI z P of z Q P of Q
Fixed effects method 18 1.19 1.13Y1.24 7.25 0.0000 35.72 0.0050
Random effects (additive method) 18 1.36 1.15Y1.61 3.63 0.0003 13.35 0.7125
Random effects (multiplicative method) 18 1.19 1.08Y1.31 3.45 0.0006 17.0 0.4544
Table 5. Variations in mean effect size by features of the methods and the program using the Badditive^
RE method.
Parameter
No. of
studies OR CI z P of z Q P of Q
Significance
of difference
in OR
Type of data Police data 15 1.38 1.16Y1.64 3.67 0.0002 9.58 0.7927 0.3173
Survey data 3 1.09 0.43Y2.73 0.17 0.865 1.91 0.3848
Type of
comparison
Matched 8 1.40 1.09Y1.79 2.66 0.0078 6.98 0.431 0.7188
Not matched 10 1.32 1.06Y1.65 2.49 0.0128 6.27 0.7126
Type of
scheme
NW only 8 1.43 1.27Y1.60 5.96 0.0000 j1.66 1 0.7263
NW plus 10 1.37 1.12Y1.68 3.10 0.0019 7.93 j0.5412
Size of
scheme area
Small 11 1.26 0.97Y1.64 1.73 0.836 9.49 0.4863 0.5222
Large 7 1.40 1.13Y1.73 3.10 0.0019 4.29 0.6375
Year 1977Y1988 11 1.56 1.21Y2.0 3.43 0.0006 6.57 0.7653 0.9385
1989Y1994 7 1.12 0.90Y1.39 1.01 0.3125 2.64 0.8525
Published Published 8 1.62 1.54Y1.71 18.03 0.0000 7.45 0.3836 0.1285
Not published 10 1.35 1.08Y1.69 2.64 0.0083 7.96 0.5382
All studies 18 1.36 1.15Y1.61 3.63 0.0003 13.35 0.7125
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 451
to four times) the value obtained in the usual formula. However, in our analyses the
results are slightly different because of the combination of police and survey data.
Whereas multi-site studies could cause great problems for the additive random
effects model (because the weighted mean effect size would vary according to the
level of aggregation chosen), the weighted mean effect size would be almost the
same if one used aggregated or disaggregated multi-site data with the MVA model.
We used the MVA model as a third way of estimating a weighted mean effect size. It
produces the same effect size as the FE model but has a larger confidence interval.
Table 4summarizes the results obtained with the three methods. All three show
that NW had a desirable effect in reducing crime. The RE model produces the
largest effect size because it gives similar weighting to large and small studies.
Comparison of the weighted mean effect sizes of the half of evaluations with the
larger sample sizes with the half with the smaller sample sizes showed that smaller
sample studies tended to have larger effect sizes (MVAOR = 1.31 compared with
MVAOR = 1.17)
3
. Giving all studies more equal weight (as is done in the RE
model) therefore tends to increase the overall effect size of the sample as a whole.
It is possible that the mean effect size of the results is affected by features of the
research design or the program being evaluated. In order to test for this, the results
were broken down into sub-groups. The results are presented for both the
Badditive^(Table 5) and Bmultiplicative^(Table 6) RE methods. However, they
are discussed below in relation to the Bmultiplicative^RE method only, as this is
considered to be the more defensible of the two methods in estimating the
weighted mean effect size and its confidence intervals.
The effect of variations in methods used was determined by comparing the
results by type of data and type of comparison group. The results presented in
Table 6. Variations in mean effect size by features of the methods and the program using the
Bmultiplicative^RE method.
Parameter
No. of
studies OR CI z P of z Q P of Q
Significance
of difference
in OR
Type of data Police data 15 1.19 1.07Y1.32 3.20 0.0014 14 0.4497 0.9045
Survey data 3 1.14 0.31Y4.15 0.20 0.8415 2 0.3679
Type of
comparison
Matched 8 1.48 1.13Y1.93 2.83 0.0047 7 0.4289 0.0819
Not matched 10 1.16 1.07Y1.27 3.48 0.0005 9 0.4373
Type of
scheme
NW only 8 1.30 1.02Y1.66 2.13 0.0332 7 0.4289 0.4965
NW plus 10 1.19 1.01Y1.39 2.08 0.0375 9 0.4373
Size of
scheme area
Small 11 1.27 0.97Y1.65 1.76 0.784 10 0.4405 0.6818
Large 7 1.19 0.96Y1.46 1.58 0.1141 6 0.4232
Year 1977Y1988 11 1.19 1.02Y1.39 2.20 0.0278 10 0.4405 0.4593
1989Y1994 7 1.12 1.02Y1.23 2.29 0.022 6 0.4232
Published Published 8 1.51 1.29Y1.77 5.02 0.0000 7 0.4289 0.0078
Not published 10 1.17 1.04Y1.31 2.59 0.0096 9 0.4373
All studies 18 1.19 1.08Y1.31 3.45 0.0006 17 0.4544
TREVOR BENNETT ET AL.
452
Table 6(using the Bmultiplicative^RE method) show that the mean OR was
positive and statistically significant (MVAOR = 1.19) for studies based on police
data but slightly lower and not statistically significant (MVAOR = 1.14) for studies
based on survey data. There was no significant difference between these two ORs.
Methods effects were also investigated by comparing type of comparison group.
The results showed that there was no significant difference in effect size by type of
comparison group used. Studies based on matched and non-matched comparison
areas were both positive and statistically significant. However, studies based on
matched comparison areas had higher mean ORs than non-matched comparison
areas (MVAOR = 1.48 compared with MVAOR = 1.16). This difference was nearly
statistically significant (P= 0.08).
We examined the effects of the features of the scheme by looking at the results
for different types of schemes and different types of scheme areas. The mean OR
was positive and significant for evaluations based on NW alone and those based on
NW with property marking and/or security surveys. The difference between these
conditions was not statistically significant. We investigated scheme effects also, by
looking at the size of area in which the NW schemes were implemented. The
results showed that larger areas tended to have larger ORs than the smaller areas.
However, these differences, again, were not statistically significant.
The next comparison was based on the year of the study. The results indicated
that schemes evaluated in the early and later periods of neighborhood watch both
showed a favorable significant effect on crime. There was no significant difference
in the results between the two periods. The final comparison considered the extent
to which the results varied by publication status. The results showed that there was
a significant difference between the mean ORs of published and unpublished
studies, with published studies associated with higher ORs. However, both
published and unpublished studies showed an association between neighborhood
watch and a reduction in crime.
Overall, the results of the meta-analysis show that neighborhood watch had a
desirable effect in reducing crime. The main difference in results concerned the
type of research methods used. Evaluations based on police data were associated
with significant reductions in crime, whereas those based on survey data were not.
However, there were only three studies based on survey data.
Summary of the results
The main findings of the narrative review were that just over half of the schemes
evaluated showed that neighborhood watch was effective in reducing crime, while
the other half mainly produced uncertain results. The main finding of the meta-
analysis was that neighborhood watch was associated with a relative reduction in
crime of about 16% (using both the FE and MVA methods). The generally positive
finding of the narrative review is consistent with the small favorable effect found
in the meta-analysis. Hence, the dominant finding of the review, using both
methods, is that neighborhood watch is effective in reducing crime.
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 453
Conclusions
The results of previous systematic reviews of neighborhood watch presented in the
introduction were divided in terms of the conclusions drawn. Titus (1984)
concluded that neighborhood watch was effective but noted that the research
methods used to investigate this were weak. Husain (1990) concluded that there
was little evidence that neighborhood watch worked. Based on the four studies
meeting the selection criteria, Sherman and Eck (2002) concluded that neighbor-
hood watch was ineffective in reducing crime.
The strongest finding of this review relates to the mean effect size estimate
produced by the meta-analysis. This indicated that, across all studies combined,
neighborhood watch was associated with a reduction in crime. It is not immediately
clear why neighborhood watch is associated with a reduction in crime. The analysis
of the moderator variables failed to show any clear differences between studies in
terms of methods used or program design and outcome. However, it is possible that
the reductions in crime were associated with some of the essential features of
neighborhood watch schemes, as discussed earlier. Neighborhood watch might serve
to increase surveillance, reduce opportunities and enhance informal social control.
Unfortunately, this kind of information is not provided in the majority of
evaluations, and the precise reasons for the reduction cannot be determined.
Research implications
There are a number of implications that can be drawn from the review for future
research on the effectiveness of neighborhood watch.
First, the review has drawn attention to the common problem of a relatively
small number of good-quality studies in terms of research design. Among the 27
studies that were excluded on grounds of methodological quality, 19 had no
comparison group and eight presented only post-test data on crime.
Second, coupled with this, it is unclear why evaluations of neighborhood watch
stopped abruptly in the mid-1990s. It is possible that researchers felt that the
effectiveness or ineffectiveness of neighborhood watch was already established
and that there was no need for further investigation. In fact, the current review has
shown that the results were mixed, with some studies showing that it works
(especially those based on police data) and others showing that it does not work
(especially those based on survey data). It would helpful if more evaluations, with
better designs, were conducted to help resolve this problem.
Third, none of the studies was based on random allocation of areas to treatment or
control conditions. Instead, all studies were based on some version of a quasi-
experimental design. This is almost certainly a result of the difficulties involved in
implementing community-based programs in areas where communities have not
requested them. It is very difficult to conduct a randomized experiment with areas as
the unit of assignment. However, quasi-experimental designs are not ideal, and some
writers have argued that they can over-estimate the positive effects of schemes as a
TREVOR BENNETT ET AL.454
result of selection effects whereby the subjects or schemes most likely to change are
included in the experimental group (for a discussion see Wilson et al. 2006).
Fourth, there was some variation among studies in the method of selecting
comparison areas. The strongest evaluations used matched comparison areas.
However, only one-quarter (26%) of the studies included in the review were based
on comparisons of this type. Other evaluations selected similar (but not matched)
areas nearby (41% of studies). The weakest evaluations were those based on non-
matched comparison areas (33% of studies). These typically were the police division
or police force area as the control and sometimes included the experimental area.
Fifth, a particularly important problem for the current review was that fewer than
half of the eligible studies reported data that were suitable for a meta-analysis. This
was either because studies presented the results using an unusual statistical notation
or left out the data entirely (e.g., when the results were presented in graphical form
only). It would be helpful if published evaluations included, at a minimum, raw data,
cell sizes and other relevant information in order to facilitate future meta-analyses.
Finally, very few evaluations disaggregated the findings in a way that would
show differential effects for sub-groups and provide detailed information on the
features of the program. As there might be variations in outcome according to the
type of program implemented or the type of area in which it is implemented, it is
important that this information should be included in a research report.
Implications for policy
Neighborhood watch has often been described as one of the most widespread
methods of reducing crime. It is supported by UK and US governments and is
popular among the public and the police (Sims 2001). The current review provides
some support for this level of implementation. However, little is known about the
factors that influence whether or not it is effective. The results of this review have
shown that there is some variation across schemes in terms of the outcomes
achieved. Governments and those responsible for crime prevention policy should
investigate differences between more effective and less effective schemes in order
to guide good practice.
Notes
1This OR is different from the usual OR, which is based on numbers of
individuals (e.g., re-offending or not re-offending in experimental or control
conditions). The use of the term BOR^might be justified on the grounds that a/b,
for example, indicates the odds of a crime occurring before or after the
intervention.
2In order to test the sensitivity of the adjustment factor, the value of V(LOR) was
trebled (rather than doubled) and the results were recalculated. The results
showed that there was no change in OR for either the fixed effects (FE) or
multiplicative variance adjustment (MVA) method and only a very small change
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 455
in OR for the random effects (RE) method. There was also either no change or a
very small change in the confidence intervals across all three methods.
3The significance of the difference between the two odds ratios was calculated
from each LOR (logarithm or the odds ratio), VLOR (variance of LOR), and N
(number of studies on which each odds ratio was based).
Z¼LOR1LOR2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pooled variance
p
Pooled variance ¼N11ðÞ
VLOR1þN21ðÞ
VLOR2
N1þN22
References
Bennett, T. H. & Wright, R. (1984). Burglars on burglary: Prevention and the offender.
Aldershot, U.K.: Gower.
Greenberg, S. W., Rohe, W. M. & Williams, J. R. (1985). Informal citizen action and crime
prevention at the neighborhood level. Washington, D.C.: Government Printing Office.
Laycock, G. (1985). Property marking: A deterrent to domestic burglary? Crime Prevention
Unit Paper 3. London: Home Office.
Lipsey, M. W. & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage.
National Crime Prevention Council (2001). The 2000 National Crime Prevention Survey.
Washington, D.C.: National Crime Prevention Council.
Rosenbaum, D. P. (1987). The theory and research behind neighborhood watch: Is it a sound
fear and crime reduction strategy? Crime and Delinquency 33, 103Y134.
Sherman, L. W. (1997). Policing for crime prevention. In L. W. Sherman, D. C. Gottfredson, D. L.
MacKenzie, J. Eck, P. Reuter & S. Bushway (Eds.), Preventing crime: What works, what
doesn`t, what`spromising(pp. 8-1Y8-58). Washington, D.C.: US Office of Justice Programs.
Sims, L. (2001). Neighborhood watch: Findings from the 2000 British Crime Survey.
(Research Findings 150.) London: Home Office.
Titus, R. (1984). Residential burglary and the community response. In R. V. G. Clarke & T.
Hope (Eds.), Coping with burglary (pp. 97Y130). Boston: Kluwer-Nijhoff.
Turner, B. W. M. & Barker, P. J. (1983). Study tour of the United States of America: 7th
March 1983 to 21st March 1983. London: Metropolitan Police.
Wilson, D. B., Mitchell, O. & MacKenzie, D. L. (2006) A systematic review of drug court
effects on recidivism. Journal of Experimental Criminology (in press).
Studies included in the review
Anderton, K. J. (1985). The effectiveness of Home Watch schemes in Cheshire. Chester, UK:
Cheshire Constabulary.
Bennett, T. H. (1990). Evaluating neighbourhood watch. Cambridge Studies in Criminology
LXI. Aldershot, UK: Gower.
Bennett, S. F. & Lavrakas, P. J. (1989). Community-based crime prevention: An assessment of
the Eisenhower Foundation`s neighborhood program. Crime and Delinquency 35, 345Y364.
Cirel, P., Evans, P., McGillis, D. & Whitcomb, D. (1977). Community Crime Prevention
Program, Seattle: an exemplary project. Washington D.C.: Government Printing Office.
TREVOR BENNETT ET AL.456
Forrester, D., Chatterton, M., & Pease, K. (1988) The Kirkholt Prevention Project,
Rochdale. Crime Prevention Unit, Paper 13. London: Home Office.
Forrester, D., Frenz, S., O_Connell, M. & Pease, K. (1990). The Kirkholt Burglary
Prevention Project: Phase II. Crime Prevention Unit, Paper 23. London: Home Office.
Henig, J. R. (1984). Citizens against crime: An assessment of the neighborhood watch
program in Washington, D.C. (Occasional Paper, Center for Washington Area Studies.)
Washington: George Washington University.
Hulin, J. O. (1979). Community-based crime prevention project. Crime Prevention Review
6, 26Y34.
Husain, S. (1990). Neighborhood watch and crime: An assessment of impact. London: Police
Foundation.
Jenkins, A. D. & Latimer, I. (1986). Evaluation of Merseyside homewatch scheme.
Management development and force planning unit. Liverpool, UK: Merseyside Police.
Knowles, L., Lesser, C. & McKewen, F. (1983). Burglary prevention: A citizen initiated and
operated neighborhood watch program. The Police Chief 50, 36Y38.
Latessa, E. J. & Travis, L. F. (1987) Citizen crime prevention: problems and prospectives in
reducing crime. Journal of Security Administration, 10(1): 38Y51.
Lewis, D. A., Grant, J. A. & Rosenbaum, D. P. (1988). Social construction of reform: Crime
prevention and community organizations. New Brunswick, N.J.: Transaction Books.
Lowman, J. (1983). Target hardening burglary prevention and the problem of displacement
phenomena. In T. Fleming (Ed.), Deviant designations: Crime, law and deviance in
Canada (pp. 277Y304). Toronto: Butterworths.
Matthews, R. & Trickey, J. (1994a). Eyres Monsell crime reduction project. Leicester, U.K.:
University of Leicester, Centre for the Study of Public Order.
Matthews, R. & Trickey, J. (1994b). The New Parks crime reduction project. Leicester,
U.K.: Centre for the Study of Public Order, University of Leicester.
Mukherjee, S. & Wilson, P. (1988). Neighborhood watch: Issues and policy implications.
(Trends and Issues No. 8.) Canberra, Australia: Australian Institute of Criminology.
Research & Forecasts Inc. (1983). The Figgie Report, Part IV: reducing crime in America -
successful community efforts. Willoughby, Ohio: Figgie International.
Sherman, L. W. & Eck, J. (2002). Policing for crime prevention. In L. W. Sherman, D. P.
Farrington, B. C. Welsh & D. L. MacKenzie (Eds.), Evidence-based crime prevention (pp.
295Y329). London: Routledge.
Tilley, N. & Webb, J. (1994). Burglary reduction: Findings from the safer cities scheme.
(Crime Prevention Unit Paper 51.) London: Home Office.
Veater, P. (1984). Evaluation of Kingsdown neighborhood watch project, Bristol. Bristol,
U.K.: Avon and Somerset Constabulary.
About the authors
Professor Trevor Bennett is Director of the Centre for Criminology at the University of Glamorgan.
Prior to this he was Acting Director of the Institute of Criminology at the University of Cambridge and
Fellow of Wolfson College. He has published widely in the area of drug misuse and crime as well as in
the areas of policing, crime prevention and offender decision making. He was Director of the New
English and Welsh Arrestee Monitoring (NEW-ADAM) programme and is a member of the Scientific
Advisory Board for the Drug Use Monitoring in Australia program (DUMA).
DOES NEIGHBORHOOD WATCH REDUCE CRIME? 457
Dr Katy Holloway is a Research Fellow and a Lecturer in Criminology and Criminal Justice at the
University of Glamorgan. Prior to this she was a Research Associate at the Institute of Criminology at
the University of Cambridge, where she worked as a data analyst on the NEW-ADAM programme. She
completed an MPhil and PHD at the University of Cambridge. Her doctoral theses examined the
decision making of Mental Health Review Tribunals and the release of mentally disordered offenders
into the community.
David P. Farrington O.B.E., is professor of Psychological Criminology at the Institute of Criminology,
Cambridge University. He is a Fellow of the British Academy, of the Academy of Medical Sciences, of
the British Psychological society and of the American Society of Criminlogy, and an Honorary Life
Member of the British Society of Criminology and of the Division of Forensic Psychology of the British
Psychological society. He is Co-Chair of the Campbell Collaboration Crime and Justice Group, a
member of the Board of Directors of the International Society of Criminology, joint editor of Cambridge
Studies in Criminology and of the journal Criminal Behaviour and Mental Health, a member of the
editotial boards of 13 other journals. He received B.A., M.A. and Ph.D. degrees in psychology from
Cambridge University, the Sellin-Glueck Award of the American Society of Criminology for
international contributions to criminology, and the Sutherland Award of the American Society of
Criminology for outstanding contributions to criminology.
TREVOR BENNETT ET AL.
458
Reproducedwithpermissionof thecopyrightowner. Furtherreproductionprohibitedwithout permission.
