Crime Prevention Effects of Closed Circuit Television

8/14/2019 Crime Prevention Effects of Closed Circuit Television

1/68

Home Office Research Study 252

Crime prevention effects ofclosed circuit television:a systematic review

Brandon C. Welsh and David P. Farrington

The views expressed in this report are those of the authors, not necessarilythose of the Home Off ice (nor do they reflect Government poli cy).

Home Office Research, D evelopment and Statistics DirectorateAugust 2002


2/68

Crime prevention effects of closed circuit television: a systematic review

Home Office Research Studies

T he H ome O ff ice R esearch S tudies are reports on research undertaken by or on behalf of theH ome O ff ice. T hey cover the range of subjects for whi ch the H ome Secretary hasresponsibi li ty. Other publications produced by the R esearch, Development and StatisticsDirectorate include Findings, Statistical Bulletins and Statistical Papers.

The Research, Development and Statistics Directorate

RDS is part of the Home Office. The Home Office's purpose is to build a safe, just and tolerantsociety in which the rights and responsibilities of individuals, families and communities areproperly balanced and the protection and security of the public are maintained.

RDS is also part of National Statistics (NS). One of the aims of NS is to inform Parliament and

the citizen about the state of the nation and provide a window on the work and performanceof government, allowing the impact of government policies and actions to be assessed.

Therefore -

Research Development and Statistics Directorate exists to improve policy making, decisiontaking and practice in support of the Home Office purpose and aims, to provide the public andParliament with information necessary for informed debate and to publish information forfuture use.

First published 2002Application for reproduction should be made to the Communication Development Unit,R oom 201, H ome Offi ce, 50 Queen A nnes G ate, London SW 1H 9A T . C rown copyri ght 2002 ISBN 1 84082 882 X

ISSN 0072 6435


3/68

Foreword

This review summarises the findings of previous studies from both the USA and Britain on theeffectiveness of C CT V in crime reduction. Forty six relevant studies were assessed according tostrict methodological criteria:

that CCTV was the main intervention studied;that there was an outcome measure of crime;that crime levels before and after the intervention were measured; thatthe studies included a comparable control area.

The authors considered only 22 of these surveys to be rigorous enough for inclusion in theirmeta-analysis. T he review draws conclusions on the effectiveness of CCT V generally and onits effectiveness in terms of specif ic settings (e.g. car parks, public transport or city centres).

O veral l , the best current evi dence suggests that CC T V reduces crime to a small degree.CCT V is most effective in reducing vehicle crime in car park s, but it had li ttle or no effect oncrime in public transport and city centre settings.

Importantly, the review draws attention to the shortcomings of many of the previousevaluations and highlights common methodological problems that either resulted in theirexclusion f rom the review or in their li mited value in the debate.

The review includes a useful summary of the knowledge gaps in relation to the impact of

CCT V on crime and sets out the key elements needed in future research and evaluation i f these questions are to be addressed.

Carole F Will isHead of Policing and R educing Crime Unit

i


4/68


Acknowledgements

We thank Hugh Arnold, London Borough of Sutton; Professor Trevor Bennett, University of G lamorgan; Professor Jason D itton, U niv ersity of Sheff ield and Scottish C entre forCriminology; Professor John E. E ck, University of C incinnati; Professor L orraine M azerolle,G rif fi th Uni versity; Professor Sara M cLafferty, H unter College; David Sk inns, DoncasterCollege; Dr Peter Squires, University of Brighton; and Professor Pierre Tremblay, Universityof M ontreal, for providing helpful assistance in obtaining copies of evaluation studies usedin this report. Appreciation is also extended to Professor Nick Tilley, Nottingham TrentUniversity, for comments on the proposal for this research; Deborah Friedman, University of M assachusetts L owell, for help with the collection of reports; Jennifer Wylie, for translationservices; and Professor M artin G il l, L eicester University, for helpful comments on the report.T hanks also go to Professor R oss Homel, Grif fi th University, A ustralia and Professor G rahamFarrell, University of Cincinnati, USA, for acting as independent assessors for this report.

Brandon C. W elshD avid P. Farrington

Brandon C . Welsh is an A ssistant Professor in the Department of Criminal Justice, University of M assachusetts Lowell. D avid P. Farrington is Professor of Psychological C riminology in theInstitute of Criminology, U niversity of C ambridge.

ii


5/68

Contents

Foreword iAcknowledgements iiList of tables ivList of figures ivSummary v

1. Background 1

2. M ethod 3Criteria for inclusion of evaluation studies 3Search strategies 5K ey features of evaluations 6Evaluations not meeting inclusion criteria 8

3. R esults 13City centre or public housing 13Publi c transport 27Car parks 34Pooled meta-analysi s results 39

4. Conclusions 41Summary of main findings 41

Pri orities for research 42Policy impl ications 44

A ppendix 1: Literature reviews consulted 47A ppendix 2: Evaluation reports that could not be obtained 49

R eferences 51

iii


6/68


List of Tables

2.1 CC TV evaluations not meeting inclusion criteria 93.1 CCTV evaluations in city centres or public housing 153.2 M eta-analysis of CCTV evaluations in city centres or public housing 263.3 CCTV evaluations in public transport 293.4 M eta-analysis of CCTV evaluations in public transport or car parks 343.5 CCTV evaluations in car parks 35

List of Figures

3.1 CCT V evaluations 40

iv


7/68

Summary

Closed circuit television serves many functions and is used in both public and private settings. Theprevention of crime (i.e., personal and property) is among its primary objectives in public space. This report aims to evaluate the evidence on the effectiveness of C CT V in preventing crime.

Determining what works to reduce crime requires examination of the results of prior evaluation

studies. This is better than drawing conclusions about what works from personal experience,from anecdotal evidence, from widespread beliefs, or from a single study which waswellfunded or highly publicised. This is the foundation of an evidence-based approach to preventingcrime, and the systematic review represents an innovative, scientific method for contributing toevidence-based prevention of crime.

T his report has two main objectives: (1) to report on the fi ndings of a systematic review -incorporating meta-analytic techniques - of the available research evidence on the effects of

CCTV on crime, and (2) to inform publi c poli cy and practice on preventing crime through theuseof C CT V interventions.

Systematic reviews use rigorous methods for locating, appraising, and synthesising evidencefrom prior evaluation studies, and they are reported with the same level of detail thatcharacterises high quality reports of original research.

Evaluations meeting the following criteria were included in this review:(1) CCT V was the focus of the intervention

(2) there was an outcome measure of crime(3) the evaluation design was of high methodological quality, with the minimum design

involving before-and-after measures of crime in experimental and control areas(4) there was at least one experimental area and one comparable control area(5) the total number of crimes in each area before the intervention was at least 20.

The following four search strategies were carried out to identify CCT V evaluations meeting thecriteria for inclusion in this review:

(1) searches of on-line databases(2) searches of reviews of the literature on the effectiveness of C CT V in preventing crime(3) searches of bibliographies of C CT V reports(4) contacts with leading researchers.

v


8/68


Both published and unpublished reports were considered in the searches, and the searcheswere international in scope and were not limi ted to the English language.

T he search strategies resulted in 22 CCTV evaluations meeting the criteria for inclusion. Theevaluations were carried out in three main settings: (1) city centre or public housing, (2) publictransport, and (3) car parks.

Of the 22 included evaluations, half (11) found a desirable effect on crime and five found anundesirable effect on crime. Five evaluations found a null effect on crime (i.e., clear evidence of noeffect), while the remaining one was classified as finding an uncertain effect on crime (i.e.,unclear evidence of an effect).

R esults from a meta-analysis provide a clearer picture of the crime prevention effectiveness of CCTV . From 18 evaluations - the other four did not provide the needed data to be included inthe meta-analysis - it was concluded that CCT V had a signifi cant desirable effect on crime,although the overall reduction in crime was a very small four per cent. H alf of the studies (nineout of 18) showed evidence of a desirable effect of CCT V on crime. A ll nine of these studies

were carri ed out in the UK . Conversely, the other nine studies showed no evidence of anydesirable effect of C CT V on crime. All f ive N orth A merican studies were in this group.

T he meta-analysis also examined the effect of C CTV on the most frequently measured crimetypes. I t was found that CC T V had no effect on violent crimes (from five studies), but had asignificant desirable effect on vehicle crimes (from eight studies).

Across the three settings, mixed results were found for the crime prevention effectiveness of CCT V . In the city centre and public housing setting, there was evidence that CCT V led to a

negligible reduction in crime of about two per cent in experimental areas compared with controlareas. CCTV had a very small but significant effect on crime in the five UK evaluations in thissetting (three desirable and two undesirable), but had no eff ect on crime in the four N orthA merican evaluations.

The four evaluations of CCT V in public transportation systems present confli cting evidence of effectiveness: two found a desirable effect, one found no effect, and one found an undesirableeffect on crime. For the two effective studies, the use of other interventions makes it difficult to saywi th certainty that CCT V produced the observed crime reductions. T he pooled effect size for all fourstudies was a non-signif icant six per cent decrease in crime.

vi


9/68

Summary

In car parks, there was evidence that CC T V led to a statistically signif icant reduction in crime of about41 per cent in experimental areas compared with control areas. For all of the studies in thissetting other measures were in operation at the same time as CCT V .

A dvancing knowledge about the crime prevention benefits of CCT V schemes should begin withattention to the methodological rigour of the evaluation designs. T he use of a control condition isimportant in ruling out some of the major threats to internal validity, but efforts are also neededto make the experimental and control conditions comparable. Attention to methodologicalproblems or changes to programmes that take place during and after implementation is needed.

Statistical power analysis is needed in advance to determine if numbers are sufficient to detectthe strength of likely effects. There is also the need for longer follow-up periods to see how fareffects persist. R esearch is needed to help identify the active ingredients and causal mechanismsof successful C C T V programmes and future experiments are needed which attempt todisentangle elements of effective programmes. R esearch is also needed on the financial costsand benefits of C CTV programmes. Future evaluations need to include alternative methods of measuring crime (surveys as well as police records).

The studies included in the present review show that CCT V can be most effective in reducingcrime in car parks. Exactly what are the optimal circumstances for effective use of CCTVschemes is not entirely clear at present, and needs to be established by future evaluationresearch. Interestingly, the success of the C CT V schemes in car park s was limi ted to a reductionin vehi cle crimes (the only cri me type measured) and all f ive schemes included otherinterventions, such as improved lighting and notices about CCT V cameras. Conversely, theevaluations of C CT V schemes in city centres and publi c housing measured a much larger rangeof crime types and the schemes did not involve, with one exception, other interventions. TheseCCTV schemes, and those focused on public transport, had only a small effect on crime. Could

it be that a package of interventions focused on a specifi c crime type is what made the CCT V -led schemes in car parks effective?

Overall , it might be concluded that CC T V reduces crime to a small degree. Future CC T Vschemes should be carefully implemented in different settings and should employ high qualityevaluation designs with long follow-up periods. In the end, an evidence-based approach tocrime prevention which uses the highest level of science available offers the strongest formula forbuilding a safer society.

vii


10/68


viii


11/68

1. Background

Closed circuit television serves many functions and is used in both public and privatesettings. T he prevention of crime (i .e., personal and property) is among i ts pri maryobjectives in publ ic space, and this i s the focus of the present report.

A s an intervention targeted at crime, CC T V is a type of situational crime prevention (e.g.,Clarke, 1995). According to Clarke and Homels (1997) classification of situational crimeprevention, CC T V is vi ewed as a technique of formal surveil lance . I n this regard, CC T Vcameras are seen to enhance or take the place of securi ty personnel.

The mechanisms by w hich C C T V may prevent crim e are numerous. T hese have beenarticul ated by A rmi tage and her col leagues (1999, pp. 226-27), and are as fol lows:

- C aught in the act - perpetrators wi l l be detected, and possibly removed or

deterred.

- Youve been framed - C CT V deters potential offenders who perceive an elevatedrisk of apprehension.

- N osy parker - CCT V may lead more people to feel able to frequent the surveill edplaces. T his wi ll increase the extent of natural surveill ance by newcomers, whichmay deter potential offenders.

- Effective deployment - CCT V directs security personnel to ambiguous situations,which may head off their translation into crime.

- Publicity - C C T V could symboli se efforts to tak e crime seriously, and theperception of those efforts may both energise law-abiding citizens and/or detercrime.

- T ime for crim e - CC T V may be perceiv ed as reducing the time available tocommit crime, preventing those crimes that require extended time and effort.

- M emory jogging - the presence of CCT V may induce people to take elementarysecurity precautions, such as locking their car, by jogging their memory.

1


12/68


- A nticipated shaming - the presence of CC T V may induce people to tak eelementary securi ty precautions, for fear that they wil l be shamed by being shownon CCT V .

- A ppeal to the cautious - cautious people migrate to the areas with CCT V to shop,leave their cars, and so on. Their caution and security-mindedness reduce the risk.

- R eporting changes - people report (and/or police record) fewer of the crimes thatoccur, either because they wish to show the [desirable] effects of C CT V or out of abelief that the C ouncil i s doing its best and nothin g shoul d be done todiscourage it.

T he growth in the use of C CT V to prevent crime in recent years, especially in the UnitedK ingdom (N orris and A rmstrong, 1999) and, surprisi ngly to a much lesser extent, in theU ni ted S tates (N ieto, 1997), and the increased attention to research on evaluating i tseffectiveness against crime (Eck, 1997, 2002; Phillips, 1999), were important reasons forcarrying out the present research.

D etermin ing what work s to reduce crime requir es us to exami ne the resul ts of pr iorevaluation studies. This is better than drawing conclusions about what works from personalexperience, from anecdotal evidence, from widespread beliefs, or from a single study whichwas well-funded or highly publicised. This is the foundation of an evidence-based approachto preventing crime, and the systematic review (see below), which serves as the basis of thisreport, represents an innovative, scientific method for contributing to evidence-basedprevention of crime.

This report has two main objectives: (1) to report on the findings of a systematic review -incorporating meta-analytic techniques - of the available research evidence on the effects of CCT V on crime, and (2) to inform public poli cy and practice on preventing crime throughtheuse of CC T V interventions.

This report is divided into four chapters. The second chapter reports on the criteria forinclusion of C CT V evaluations in this review and the methods used to search for, code, andanalyse evaluation reports of CCT V programmes. The third chapter discusses the researchfindings organised by the setting in which C CT V evaluations were conducted, and the finalchapter summarises the main findings and identifies priorities for future research and policyimplications.

2


13/68

2. Method

The present report presents a systematic review of the effects of CCT V on crime and followsclosely the methodology of this review technique. Systematic reviews use rigorous methodsfor locating, appraising and synthesising evidence from prior evaluation studies, and theyare reported with the same level of detail that characterises high quality reports of originalresearch. According to Johnson et al. (2000, p. 35), systematic reviews essentially take anepidemiological look at the methodology and results sections of a specific population of studies to reach a research-based consensus on a given study topic. They have explicitobjectives, explicit criteria for including or excluding studies, extensive searches for eligibleevaluation studies from all over the world, careful extraction and coding of key features of studies, and a structured and detailed report of the methods and conclusions of the review.All of this contributes greatly to the ease of their interpretation and replication by otherresearchers. It is beyond the scope of this report to discuss all of the features of systematicreviews, but interested readers should consult key reports on the topic (see e.g., Farrington

and Petrosino, 2000; Johnson et al., 2000; Farrington and Welsh, 2001; Farrington et al.,2001).

Criteria for inclusion of evaluation studies

In selecting evaluations for inclusion in this review, the following criteria were used:

(1) CC T V was the focus of the intervention. For evaluations involving one or more

other interventions, only those evaluations in which CC T V was the mainintervention were included. The determination of the main intervention was basedon the author identifying it as such or, if the author did not do this, the importanceof C C T V relative to the other in terventions. For a small number of i ncludedevaluations with multiple interventions, the main intervention was not identified,but it was clear from the report that CCT V was the most important intervention. I tis desirable to include only evaluations where CCT V was the main intervention,because in other cases it is impossible to disentangle the effects of C CTV from theeffects of other interventions.

(2) There was an outcome measure of crime. The most relevant crime outcomes wereviolent and property crimes (especially vehicle crimes).

3


14/68


(3) T he evaluation design was of high methodological quality, w ith the minimum designinvolving before-and-after measures of crime in experimental and control areas. Theunit of interest is the area (including car parks and underground stations).

(4) T here was at least one experi mental area and one comparabl e control area.Studies involving residential, busi ness or commercial areas (e.g., city centres),and other publ ic and pri vate areas (e.g., underground stations, car park s) wereeligible for inclusion. Studies that compared an experimental area with theremainder of a city were excluded, because the control area was non-

comparable.

(5) T he total number of crimes in each area before the intervention was at least 20.T he main measure of effect size was based on changes in crime rates betweenthe before and after time periods. It was considered that a measure of changebased on an N below 20 was potentiall y mi sleading. A lso, any study wi th fewerthan 20 crimes before would have insufficient statistical power to detect changesin crime. The criterion of 20 is probably too low, but we were reluctant to exclude

studies unless their numbers were clearly inadequate.

I t is worth saying a few more words about criterion 3. I deally , the gold standard of therandomised experiment, which is the most convincing method of evaluating crime preventionprogrammes (Farri ngton, 1983), would have been used. T he k ey feature of randomisedcontrolled trials, which are widely used in medical evaluations, is that the experimental andcontrol groups are equated before the experimental intervention on all possible extraneousvari ables. H ence, any subsequent diff erences between them must be attributable to theintervention. Technically, randomised experiments have the highest possible internal validity

in unambiguously attributing an effect to a cause (Shadish et al., 2002).

The randomised experiment, however, is only the most convincing method of evaluation if asufficiently large number of units is randomly assigned to ensure that the experimental groupis equivalent to the control group on all possible extraneous variables (wi thin the li mits of statistical f luctuation). A s a rule of thumb, at least 50 uni ts in each category are needed.This number is relatively easy to achieve with individuals but very difficult to achieve withlarger units such as areas, as in the evaluation of C CT V schemes. For larger units such asareas, the best and most feasible design usually involves before-and-after measures inexperimental and control conditions together with statistical control of extraneous variables(Farrington, 1997). T he use of a control condition that is comparable with the experimentalcondition is necessary in order to exclude threats to internal validity.

4


15/68

Method

Search strategies

The following four search strategies were carried out to identify CCTV evaluations meeting thecriteria for inclusion in this review:

(1) searches of on-line databases (see below)(2) searches of reviews of the li terature on the eff ectiveness of C CT V in preventing

crime (for a list of reviews consulted, see Appendix 1)(3) searches of bibliographies of C CTV reports(4) contacts with leading researchers (see Acknowledgements).

Both published and unpublished reports were included in the searches. Furthermore, thesearches were international in scope and were not limited to the English language (one non-English language evaluation report is included in the review). Searches (1) through (3) werecompleted in January 2001 and refl ect material publ ished or known up to 31 D ecember2000.

The following eight databases were searched:

(1) Criminal Justice Abstracts(2) National Criminal Justice Reference Service (NCJRS) Abstracts(3) Sociological Abstracts(4) Social Science Abstracts (SocialSciA bs)(5) Educational R esources Information Clearinghouse (ER IC)(6) Government Publications Office Monthly Catalog (GPO Monthly)(7) Psychology Information (PsychInfo)(8) Public A ffairs Information Service (PA IS) International

T hese databases were selected because they had the most comprehensi ve coverage of criminological, criminal justice, and social science literatures. They are also among the topdatabases recommended by the Crime and Justice G roup of the Campbell Collaboration, andother systematic reviews of interventions in the field of crime and justice have used them (e.g.,Petrosino, 2000; Petrosino et al., 2000).

The following terms were used to search the eight databases noted above: closed circuittelevi sion, CC T V , cameras, social control, surveill ance, and formal surveill ance. W henappli cable, crime w as then added to each of these terms (e.g., C C T V and crime) tonarrow the search parameters.

5


16/68


T hese search strategies resulted in the collection of 22 CC TV evaluations meeting the criteria forinclusion in this review. A few of the evaluations identified, which may or may not have metthe criteria for inclusion, could not be obtained. The reports of these evaluations are listed inA ppendix 2.

Key features of evaluations

T ables 3.1, 3.3, and 3.5 summari se key features of the 22 included CCT V evaluations.

A uthor, publication date, and location. T he authors and dates of the most relevantevaluation reports are listed here, along with the location of the programme. Theevaluations have been li sted in chronological order, according to the date of publication.

Context of intervention. This is defined as the physical setting in which the CCTVintervention took place.

Type and duration of intervention. The intervention is identified and any keyfeatures are listed. The length of time the programme was in operation is alsonoted here.

Sample size. The number and any special features of the experimental andcontrol areas are identif ied.

O ther interventions. I nterventions other than CC T V whi ch were employed at thetime of the programme are identif ied.

O utcome measure of interest and data source. A s noted above, crime was theoutcome measure of i nterest to this review. H ere the specif ic crime types as wellas the data source of the outcome measure are identified.

R esearch design and before-after time period. A s noted above, the minimumresearch design for an evaluation to be included in this review involves before-and-after measures of crime in comparable experimental and control areas. If matching or other statistical analysis techniques were used as part of theevaluation of programme effects, these too are noted here. The before and aftertime periods of the evaluation are also noted.

6


17/68

Method

R esults. In summarising results, the focus was on the most relevant crime outcomesfor this review (i.e., property and violent crime types) and comparisons betweenexperimental and control areas. The results of significance tests are listed, but theywere rarely provided by researchers. Si mil arly, few effect size measures wereprovided. The problem with signif icance tests is that they depend partly on samplesize and partly on strength of effect. A significant result in a large sample couldcorrespond to a rather small effect size, and conversely a large effect size in asmall sample may not be statistically significant. Consequently, this report relies onmeasures of effect size (and associated confidence intervals) where possible.

Each of the evaluations were rated on their effectiveness in reducing crime. Each evaluation isassigned to one of the following four categories:

(1) desirable effect: significant decrease in crime(2) undesirable effect: significant increase in crime(3) null effect: clear evidence of no effect on crime(4) uncertain effect: unclear evidence of an effect on crime.

Category 4 was assigned to those evaluations in which methodological problems (i.e., smallnumbers of crimes or contamination of control areas) confounded the reported results to thepoint that the evaluation could not be assigned to one of the other three categories. It wasdiff icult to rate those evaluations whi ch reported the percentage change in crime (frombefore to after the programme was implemented), but did not provide data on the numberof crimes in the before and after periods. Instead of giving these evaluations a rating of uncertain effect, they were rated subjectively on the basis of the reported percentagechange in crime.

Other dimensions. CCTV evaluations differ on many different dimensions, and it isimpossible to include more than a few in summary tables. T wo important issuesthat are addressed, not in the tables, but in the accompanying text, aredispl acement and dif fusion of benefi ts. D ispl acement is of ten defined as theunintended increase in targeted crimes in other locations following from thein troduction of a crime reduction scheme (for a discussion of benign ordesirable effects of displ acement, see Barr and Pease, 1990). Fi ve dif ferent formsof displacement have been identified by R eppetto (1976): temporal (change intime), tactical (change in method), target (change in victim), territorial (change inplace), and functional (change in type of crime). Diffusion of benefits is defined asthe unintended decrease in non-targeted crimes following from a crime reductionscheme, or the complete reverse of displacement (Clarke and Weisburd, 1994).

7


18/68


In order to investigate territorial displacement and diffusion of benefits, the minimum designinvolves one experimental area, one adjacent area, and one non-adjacent control area. If crime decreased in the experimental area, increased in the adjacent area, and stayedconstant in the control area, this might be evidence of displacement. If crime decreased inthe experimental and adjacent areas and stayed constant or increased in the control area,this might be evidence of diffusion of benefits. V ery few of the included evaluations had bothadjacent and non-adjacent but comparable control areas. M ore had an adjacent controlarea and the remainder of the city as another control area, for example.

Evaluations not meeting inclusion criteria

When coding CC T V evaluations, many did not meet the criteria for inclusion and thus have notbeen included in the present review. Altogether, 24 CCT V evaluations were excluded. Table2.1 lists these evaluations, summarises their key features, and identifies the reasons forexclusion. The reasons for discussing these evaluations here are two-fold: first, it conformswith the widely-held practice in systematic reviews of listing excluded studies and second, it

allows readers to judge for themselves the strength of observed effects in excludedevaluations compared with those included.

A s shown in T able 2.1, 17 of the 24 evaluations w ere excluded because no control areawas used in evaluating the impact of the intervention. Another four evaluations wereexcluded because no comparable control area was used. The remaining three evaluations(K ings L ynn, in B rown, 1995; Squires, 1998b, d) were excluded because they did notreport crime data. M issing information on the few k ey features li sted in T able 2.1 was notmuch of a problem with the 24 evaluations, although three failed to specify the length of the

fol low-up period. For the 21 evaluations that did provi de information on the fol low-upperiod, nin e inv olved follow-ups of less than one year. M any of the CC T V schemesappeared to be successful in reducing a range of crimes, including robbery, assault,burglary, motor vehicle theft and vandalism. H owever, a number of the evaluations of theseschemes were limited by small numbers of crimes. Because of methodological problems it isdiff icult to give much credence to the results of these evaluations.

8


19/68

Table 2.1: CCTV evaluations not meeting inclusion criteria

Author, publication R eason for not Other Sample Follow-up anddate, and location including programme interventions size results

Burrows (1991) No control area Changes in store 1 store (T esco -- n.a.;design and large retail er) unk nown losses:procedures approx. 12,000 to 5,000 per

week; cash losses (from tills):

approx. 500 to 20 per week N ational A ssociation of N o control area n.a. 189 2 years;Convenience Stores, convenience robbery: -15.2% (1.58 to 1.34mul tiple sites, (1991), USA stores per store per year, N S)Poyner (1992), N o control area M edia publ icity 5 buses 8 months;N orth Sh ields and school visits vandali sm: -52.9% (51 to 24)Carr and Spring (1993),V ictoria, Australia

N o control area M ultiple (e.g., T rain, tram, and 2 years;improved lighting, bus systems of crimes against persons: -42.2%police) Public T ransport (57.3 to 33.1 per month);

System vandalism: -83.6% (700 to 115broken windows, weekly average)

T il ley (1993a), Sal ford N o control area N one 3 businesses 12 months;total crimes: -14.3% (35 to 30)

1. T il ley (1993b), N o control area M edia publi city 1 station car park 4 months;Lewisham and notices vehicle crimes: -75.0% (24 to 6)of CCT V

2. T il ley (1993b), H ull N o comparable control N one E=1 car park , 8 months;area C=city centre as E vs C : theft of vehicles: -88.9% (27

a whole to 3) vs -5.6% (430 to 406); theftfrom vehicles: -76.3% (38 to 9) vs+2.8% (961 to 988)


20/68

Author, publi cation R eason for not Other Sample Follow-up anddate, and location including programme interventions size results

3. T il ley (1993b), N o comparable controlWolverhampton area

N otices of CC T V E=1 car park, 13 months;C=subdivision as E vs C : theft of vehicles: -18.2% (11a whole to 9) vs +3% (data n.a.); theft from

vehicles: -46.4% (28 to 15) vs -3%(data n.a.)

Chatterton and Frenz N o control area N otices of CCT V 15 housing 5-10 months;(1994), M erseyside schemes (sheltered burglary (completions and attempts):

accommodation ) -78.8% (4.25 to 0.9 per month)aD avidson and Farr N o control area M ultiple (e.g., 5 housing blocks 15 months;(1994), M itchelhil l Estate, target hardening, total crimeb: -63.1% (28.7 to 10.6G lasgow local management) average per quarter year)Brown (1995), N o crime data for N one E=car park s and 32 months;K ings Lynn experimental or control adjacent streets, E vs C: theft of vehicles: decline

areas C=rest of poli ce (data n.a.) vs ? (data n.a.); theftdivi sion from vehicles: decline (data n.a.) vs

decline (data n.a.); burglary (datan.a.) vs ? (data n.a.)

Squ ires and M easor N o comparable control N one E=pol ice beats 1-4, 12 months;(1996), Br ighton area C=rest of Br ighton E vs C : total crimes: under -10%

(data n.a.) vs -1% (data n.a.)Bromley and T homas N o control area M ultipl e (e.g., D if ferent types of n.a. (no before measures);(1997), Cardi ff and staff at exits, car parks vehicle crimes: Cardiff (8.3/100Sw ansea pain ting) spaces) vs. Sw ansea (13.7/100

spaces)G il l and T urbin (1998, N o control area N one 2 retail stores n.a.;1999), Leeds and stock losses from theft (before-duringSheffi eld phases and Leeds store only): 600

to 200 per week


21/68


22/68

A uthor, publication R eason for not O ther Sample Follow-up anddate, and location including programme interventions size results

3. Si varajasingam and N o control area N one 1 city centre or 2 years;Shepherd (1999), R hyl town area A & E recorded assault: +46.0%

( 1 ,249 to 1 ,823 ) ; pol ice -recordedassault: -24.0% (526 to 400)

1. T aylor (1999), N o control area M ultipl e (e.g., 154 businesses 11 months;Leicester (W est End) silent alarm) commercial burglary: decli ne

(data n.a.)

2. T aylor (1999), N o control area M ultipl e (e.g., n.a. 24 months;Leicester (Belgrave) silent alarm) commercial burglary: decline

(data n.a.)a T he total number of off ences were 51 in the before period and 9 in the after period. I n 13 of the 15 schemes, no offenses of burglary w ere recorded for the

period after CC T V was installed. O ne scheme had no burglaries in either period, and in another, there was a slight increase after camera installation(Chatterton and Frenz, 1994, p. 136).

b T he indivi dual crimes and their before-after comparisons (average per quarter year) were as follows: bu rglary (19.0 to 5.4), theft of and from vehicles (4.7 to1.4), theft other (2.0 to 2.2), vandalism (2.3 to 0.8), and crimes against the person (0.67 to 0.8). The before and after periods consisted of six quarters or 18months and 5 quarters or 15 months, respectively.

c T he figures in parentheses reflect the v alue of goods lost expressed as a percentage of all goods sold (B eck and Wil li s, 1999, p. 257).N otes: Locations were in the UK unless otherwi se specifi ed; E = experimental area; C = control area; n.a. = not available; A & E = accident and emergency

department; NS = non-significant.


23/68

3. Results

T his chapter discusses the results of the 22 included CCT V evaluations. It also summariseskey features of the evaluations which are important in the assessment of programme effects(e.g., other interventions, sample size, follow-up periods). The evaluations have beenorganised according to the setting in which the intervention took place. Three main settingswere delineated: (1) city centre or public housing, (2) publi c transport, and (3) car park s.

City centre or public housing

Thirteen evaluations were identified that met the methodological criteria for inclusion in this

review and assessed the impact of CC T V on crime in the setting of a city centre (N =11) orpubli c housing (N =2). T hree of the evaluations are reported in M azeroll e et al. (2000). O f the three settings, this contains the largest number of evaluations. Selected evaluations arediscussed below and see Table 3.1 for summary information on each of the 13 evaluations.

S even of the 13 evaluations were carri ed out in England, five in the U .S ., and one inScotland. On average, the duration of the follow-up evaluations was 10.9 months, rangingfrom a low of three months in the evaluation by M usheno et al. (1978) to a high of 24months in the evaluations by Short and Ditton (1995) and Skinns (1998b). Only one of the

evaluations (Skinns, 1998a) included other interventions in addition to the main intervention of C C T V . M any of the evaluations used mul tiple experimental areas (e.g., police beats,apartment bui ldi ngs), meaning that the coverage of the CC T V intervention w as quiteextensive in the city or town centre. M ul tipl e control areas (e.g., adjacent poli ce beats,remainder of city) were also used by some of the evaluations.

A s shown in T able 3.1, the city centre or public housing CC T V evaluations showed mixedresults in their effectiveness in reducing crime. Five of the 13 evaluations were considered tohave a desirable effect on crime, while three were considered to have an undesirable effect(increased crime). The remaini ng fi ve evaluations were considered to have a null (clearevidence of no effect; N =4) or uncertain (unclear evi dence of an eff ect; N=1) effect oncrime.

13


24/68


T wo evaluations of city centre CC T V schemes were conducted by Brown (1995). The firstevaluation took place in N ewcastle-upon-Tyne and involved the installation of 14 CCT Vcameras in f our poli ce beats in the ci ty centre (the experimental area). T he control areacomprised the seven remaining police beats of the city centre, which surrounded theexperimental area. It is important to note that two cameras were installed in police beatswhich were part of the control area.

Fifteen months after the start of the programme, the monthly average of total crimes wasreduced by 21.6 per cent (from 343 to 269) in the experimental area and 29.7 per cent

(from 676 to 475) in the control area, which overall was an undesirable effect of C CT V . T hemeasure of total crimes includes burglary, criminal damage, theft of vehicles, theft fromvehicles, theft other, and juvenile disorder. Table 3.1 presents the results of the interventionfor a number of these crimes. Reductions were observed in burglary, theft of vehicles, andtheft from vehicles in both the experimental and control areas, with the reductions in theexperimental area outpacing those in the control area. However, the number of these crimesin the experimental area was small. For example, burglary was reduced by 57.5 per cent inthe experimental area (from 40 to 17) and 38.7 per cent in the control area (from 75 to

46). Brown (1995) found little evidence of territorial or functional (change in type of crime)displacement, but did find some evidence of diffusion of benefits, particularly for the crimesof burglary and criminal damage.

The second evaluation by Brown (1995) was carried out in Birmingham. In this programme, 14CCT V cameras were installed in the centre of the city, wi th the cameras covering for themostpart shopping streets and partial ly open mark et areas , as wel l as some of the financialdistri ct. Three control areas were establi shed, wi th streets in control area 1 (C 1) receivingpartial coverage by the CCT V system (see Table 3.1). T herefore, the experimental area was

compared with control areas 2 and 3 combined.

A fter 12 months, total crimes, according to vi ctim survey reports, were reduced in theexperi mental area, whi le total cri mes increased in each of the three control areas. T heactual number of crimes was much greater in the experimental area than in any of thecontrol areas. Some evidence of what appears to be functional displacement (change intype of crime) was found, with offenders switching from robbery and theft from the person totheft from vehicles.

I n the programme evaluated by S arno (1995), 11 CCT V cameras were install ed in the towncentre of the London Borough of Sutton as part of the Safer Sutton Initiative launched in theearly 1990s. The remaining part of the police sector in the town centre, which did not

14


25/68

Table 3.1: CCTV Evaluations in City Centres or Public Housing

Author, Publi cation Context of Type and Sample Other Outcome Research Design ResultsDate, and Location Intervention Duration of Size Interventions Measure of and Before- After

Intervention Interest and T ime PeriodData Source

M usheno, Levine, Public CC T V E=3 N one Crime Before-after, E vs C: totaland Palumbo housing monitoring buildi ngs, (multiple experimental- crimes: -9.4% (32(1978), Bronxdale system C =3 offences); control to 29) vs -19.2%

H ouses, N ew York (cameras in buildi ngs victim survey (26 to 21)City, US A lobby and N ote: project Before=3 months; (uncertain effect)

elevators; had 26 A fter=3 monthsmonitors in high-rises;apartments); 533 months apartments

in each1. Brown (1995), C ity or town CCT V ; 15 E=4 beats of N one Cri me Before-after, E vs C (monthlyN ewcastle-upon-Tyne centre months central area, (multiple experimental average): total

C=7 N ote: 14 of offences); control crimes: -21.6%remaining 16 cameras poli ce (343 to 269) vsbeats of city are in E; records Before=26 months; -29.7% (676 tocentre remaini ng 2 A fter=15 months 475); burglary :N ote: T here are in C -57.5% (40 to 17,

are 2 other p


26/68

(18 to 9, p


27/68

Short and D itton T own centre CCTV ; E=6 police None Crime (total Before-after, E vs C3: total(1995, 1996) and 24 months beats, C1= and mul tipl e experimental crimes: -35% (dataD itton and Short rest of 6 categories); control n.a.) vs -12%(1998, 1999), poli ce beats poli ce (data n.a.)A irdr ie (not in records Before=24 months; (desirable effect)

camera After=24 monthsvision), C2= Note: D ata notrest of police provided to allowsub-divi sion, for comparisons of C3= rest of E with C1 or C2policedivision

Sk inns (1998a, b), Town centre CC TV ; E=all or parts Help points Crime (total Before-after, E vs C: totalDoncaster 12 months of streets in for publi c to and selected experimental police-recorded

visi on of contact CCT V offences); control crimes: -21.3%cameras in control poli ce (5,832 to 4,591)commercial rooms records Before=24 months; vs +11.9% (1,789areas, After=24 months to 2,002)C=comm- (desirable effect)ercial areas N ote: T here wereof 4 adjacent 2 Es and 6 Cs used.townships The C used here is

because the authorsays it was the mostcomparable to EN ote: This E hasbeen used becauseit includes theother E


28/68

Squi res (1998a), T own centre CC T V ; 7 E=town N one Crime (total, Before-after, E vs C: total crimes:I lford months centre, violent, and experimental- -17% (data n.a.)

C=areas selected control vs +9% (data n.a.)adjacent to offences); (desirable effect)town centre police Before=6 months;

records After=7 monthsN ote: 2 other Csused, but lesslik ely to becomparable to E

Armitage, Smyth, T own centre CCTV ; 20 E=police None Crime (total Before-after, E vs C1: totaland Pease (1999), months beats with and multipl e experimental- crimes: -28%Burnley CCTV , offences); control (1,805 to 1,410)

C1=beats police vs -1% (6,242 tohaving a records Before=12 months; 6,180); violence:common After=12 monthsa -35% (117 to 87)boundary vs -20% (267 towith CC TV 223); vehiclebeats, crimes: -48% (375C2=other to 253) vs -8%beats in (1,842 to 1,706);police burglary: -41%div ision (143 to 101) vs

+9% (2,208 to2,426)E vs C2: totalcrimes: -28% vs+9% (1,069 to1,175); violence:-35% vs 0% (32 to32); vehicle crimes:


29/68

-48% vs -8% (309to 285); burglary:-41% vs +34%(366 to 555)(desirable effect)

1. M azerolle, Ci ty centre CC TV ; 3 E=1 site with None Calls for Before-after, E vs C (weeklyH urley, and months CCTV , C= service experimental- average): +1.8%Chamlin (2000), 1,000 foot (weekly control (901 to 917) vs

Cincinnati radius BZ average); 0% (36 to 36)(Northside), US A police Before=23 months; (null effect)records A fter=6 months

Note: 2 other Csof 200 and 500foot radii wereused and areincluded in the1,000 foot radius C

2. M azerolle, Ci ty CC TV ; 3 E=1 site with None Calls for Before-after, E vs C (weeklyH urley, and centre/park months CCTV , C= service experimental- average): +9.8%Chamlin (2000), 1,000 foot (week ly control (1,062 to 1,166)Cincinnati (H opkins radius BZ average); (vs 0% (22 to 22)Park ), US A police Before=23 months; (null effect)

records A fter=4 monthsNote: 2 other Csof 200 and 500foot radii wereused and areincluded in the1,000 foot radius C


30/68

3. M azerolle, Ci ty centre CCT V ; 2 E=1 site with None Call s for Before-after, E vs C (weeklyH urley, and months CC TV , C= service experimental- average): +16.9%Chamlin (2000), 1,000 foot (week ly control (1,005 to 1,175)Cincinnati (Findlay radius BZ average); vs +17.1% (111M arket), USA police Before=24.5 months; to 130)

records After=3.5 months (null effect)N ote: 2 other C s of 200 and 500 footradii were usedand are includedin the 1,000 footradius C

Will iamson and Publi c CCT V ; 18 E=9 buildings N one Cri me (total Before-after, E vs C: change inM cLafferty (2000), housing months (1,220 apart- and mul tiple experimental- total crimes insi deBrooklyn, N ew ments; A lbany categories) control wi th projects: 0% vsYork , USA project), insi de housing matching -5.3%; change in

C=no. of projects and total crimes insidebuildings n.a. inside zones Before=18 months; 0.1 mile BZ : 0%(R oosevelt of 0.1 to 0.5 After=18 months vs -4.0%; changeproject) miles radii in major felonies

around inside projects:projects; -22.8% vs -14.5%;police records change in

major felonies

inside 0.1 mile BZ :-6.4% vs -8.6%(data n.a.)(null effect)


31/68

Farrington, Bennett, City centre CCT V ; 11 E=city centre, N one Cr ime (total Before-after, E vs C: total crimes:and Welsh (2002), months C= secondary and multiple experimental- -13.8% (2,600 toCambridge centre categories); control 2,242) vs -26.9%

police records (1,324 to 968);A lso victim Before=11 months; violent crimes:survey data A fter=11 months -6.0% (151 to 142)on crime and vs -33.8% (77 todisorder 51); vehicle crimes:

-53.1% (224 to105) vs -54.0%(250 to 115);percentagevictimized: +8.0%(26.4% to 28.5%)vs +19.3% (11.4%to 13.6%)(undesirable effect)

a There was an additional eight months of follow-up, but the authors reported crime data as percentage changes relative to the 12-month before period, so it wasnot possible accurately to calculate the number of incidents for the additional eight months.

Notes: Locations were in the UK unless otherwise specif ied; BZ = buff er zone (area surroundi ng experimental area); E = experimental area; C = control area; n.a. = notavailable.


32/68


receive any C C T V coverage, served as the control area. (One other control area was used, butit was not comparable to the experimental area.) Twelve months after the programme began,total police-recorded crime (not including vehicle crime) had decreased by 12.8 percent inthe experimental area but by 18 per cent in the control area. Sarn o did not investigatethe possibility of displacement or diffusion of benefits.

Short and D itton (1995) evaluated a CCT V scheme in A irdrie town centre, whi ch involved12 cameras spread over six police beats; this comprised the experimental area, and thecomparable control area was the remainder of the six police beats not in camera vision.(T wo other control areas were used, but the only data supplied was for the rest of the poli cedivision.) After 24 months, total police-recorded crime had decreased by 35 per cent in theexperim ental area compared wi th a 12 per cent decli ne in the control area. Short andDitton found some evidence of diffusion of crime prevention benefits from the experimentalarea to the control area.

T he programme evaluated by Skinns was a multi-agency, police-led, town centre system,consisting of 63 cameras located in the commercial centre, multi-storey car parks and main

town centre arterial roads (1998a, p. 176). The programme has been included here, asopposed to in the setting of car parks, because the main focus of the intervention was thetown centre. As noted above, another intervention was used: help points were establishedwi thin the experimental area to aid the public in contacting the main C CT V control room.T he experimental area included all or parts of streets in vi sion of the cameras. (A notherexperimental area was used but it is included in this experimental area.) The control areaincludes commercial areas of four adjacent townships. Five other control areas were used,but Skinns noted that these control areas were less comparable with the experimental areathan the one used in this present report for experimental-control comparisons.

Twenty-four months after the start of the programme, total police-recorded crime hadreduced in the experimental area by 21.3 per cent, but it had increased in the control area by11.9 per cent. The author found no evidence that total crimes were displaced from theexperimental area to the control area. The increase in crime in the control area was judged bythe author to be due to pre-existing trends.

I n the programme evaluated by Squi res (1998a), an unknown number of C CT V cameraswere installed in Ilford town centre to address a range of crime problems; areas adjacent tothe town centre served as the control condition. (T wo other control areas were used, buttheir comparability with the experimental area is less likely.) Seven months after theprogramme began, total police-recorded crime had fallen by 17 per cent in the

22


33/68

Results

experimental area, but had increased by 9 per cent in the control area. Squires found someevidence that crimes, particularly robbery and residential burglary, had been displaced fromthe town centre to adjacent areas (the control area).

In the programme evaluated by Armitage and her colleagues (1999), an unknown number of cameras were installed in the town centre of Burnley. T he experimental area consisted of policebeats in the town centre wi th CC T V coverage. Two control areas were used. The first comprisedthose police beats which shared a common boundary with the beats covered by C C T V . T hesecond control area consisted of other police beats in the police division. The first control areawas more comparable to the experimental area.

A fter 12 months, the experimental area, compared wi th the two control areas, showedsubstantial reductions in violent crime, burglary, vehicle crime, and total crime (see T able3.1). For example, total incidents of crime fell by 28 per cent (from 1,805 to 1,410) in theexperimental area compared with a slight decline of one per cent (from 6,242 to 6,180) incontrol area 1 and an increase of nine per cent (from 1,069 to 1,175) in control area 2. T heauthors found evidence of diffusion of benefits for the categories of total crime, violent crime,

and vehicle crime, and evidence of territorial displacement for burglary.

In the three Cincinnati programmes by M azeroll e et al. (2000) the outcome measure used toevaluate the impact on crim e was (w eek ly average) calls for poli ce servi ce, and theevaluation included one experimental and three control areas, the latter being buf ferzones of varyi ng distances around the experi mental area. T he outcome measure waslimited to total calls for police service. The authors also reported on police calls for disorder(disorderly persons, curfew violation, neighbour trouble, noise complaints, and suspiciouspersons or vehicles) and drugs for the three buffer zones, but not for the experimental site;

therefore, comparisons could not be made between experimental and control sites fordisorder and drug offences.

The impact of C CTV on calls for police service was fairly consistent across the three locations:calls for service increased in the experimental site and increased or remained the same in thethree control si tes or buffer zones. For the Findlay M arket programme, crime also increased in thetwo farthest buffer zones (500 and 1,000 feet away). Overall , CCT V did not have a desirableeffect on calls for service in the experimental sites of the three locations. All of these schemes hada null eff ect on crime. The authors investigated the possibi li ty of displacement in the Northsideand Findl ay M arket programmes. I n N orthside, the authors found li ttle or no evidence of displacement, while in Findlay M arket, the authors concluded that the results tend to suggestsome displacement of activi ty as reflected in calls for service (M azerolle et al., 2000, p. 24).

23


34/68


I n the programme evaluated by Farri ngton et al. (2002), 30 cameras were in stalled inC ambridge C ity centre. The control area was a secondary city centre shopping area (theG rafton centre) where there were no cameras on the streets. C omparing 11 months after thecameras were install ed wi th the comparable 11 month period before, poli ce-recordedcrimes had decreased by 13.9 per cent in the experimental area (from 2,600 to 2,242) butby 26.9 per cent in the control area (from 1,324 to 968). H ence, there was an undesirableeffect of C C T V on police-recorded crimes. V iolent crimes (assault and robbery) alsodecreased more in the control area, whi le vehicle crimes (theft of and f rom vehicles)decreased equally in the experimental and control areas. Interviews were also carried outwi th quota samples of persons in the areas before and after the CCT V installation, ask ingthem about their victimization (insul ted or bothered, threatened, assaulted, or mugged) inthe previous 12 months. T he percentage victimized i ncreased f rom 26.4 per cent to 28.5per cent in the experimental area and from 11.4 per cent to 13.6 per cent in the controlarea, suggesting that the installation of C CT V had no effect on vi ctimization. These resultssuggested that C C T V may have had no effect on crime but may have caused increasedreporting to and/or recording by the poli ce.

O nly two evaluations (M usheno et al., 1978; W il li amson and M cLafferty, 2000) wereidentified that met the methodological criteria for inclusion in this review and assessed theimpact of C CT V on crime in the setting of public housing. Both of the schemes took place inNew York City, but were implemented many years apart: the former in 1976 and the latter in1998. T he research design of the evaluation by W il li amson and M cLaf ferty (2000) wasparticularly rigorous, employing matching techniques to control for pre-existing differences(i.e., size of the housing communities, demographics, and neighbourhood location) betweenthe experimental and control areas. Concerning the research design of the other programme,M usheno et al. (1978) took efforts to make the respondents of the victim survey comparable in

the experimental and control areas; for example, half of the residents of the three experimental(all apartments received the intervention) and three control buildings were randomly selected toparticipate in the survey, which was administered before and after the CC T V intervention.

Both of the programmes did not involve interventions other than C C T V , although theapplication of CCT V dif fered somewhat between the two evaluations. In the programme byWil li amson and M cLafferty, cameras were installed at various locations in the experimentalproject (e.g., all elevators, lobbies, and roofs of buildings, and common areas and buildingwater tanks) and were monitored - from a remote location - 24 hours a day, seven days aweek, by uniformed officers of the New York City Police Department. In the other programme,cameras were installed in all of the lobbies and elevators of the experimental buildings, butwere monitored by the residents themselves: the cameras transmit pictures continuously to

24


35/68

Results

every residents television receiver ... The top half of the screen telecasts the lobby and thebottom half shows the inside of the elevator viewed from above. Sounds emitted in theselocations are also communi cated to tenants sets (M usheno et al., 1978, p. 648).

Another difference between the two evaluations is the scale of the intervention, for both thenumber of C C T V cameras install ed and the number of experimental si tes used. I n theevaluation by Wi ll iamson and M cLaf ferty, a total of 105 cameras were install ed at ninebuildings (the experimental project), comprising a total of more than 1,200 apartments; in theevaluation by M usheno et al., three bui ldings, compri sing a total of ju st over 150

apartments, were used as the experimental site (see Table 3.1). The authors did not report thenumber of cameras used, but considering that cameras were only installed in the lobbies andelevators, it is l ik ely that the numbers were quite low.

The evaluation by M usheno et al. showed that, three months after the cameras were installed,total incidents of crime were reduced in both the experimental and control sites: -9.4 per centand -19.2 per cent, respectively. H owever, as il lustrated in T able 3.1, the number of crimesrecorded was very low. This has the effect of inflating the before-after percentage changes

and limiting the examination of programme results to total crimes (the numbers for individualcrime types are even smaller). Because of small numbers, it was concluded that thisprogramme had an uncertain effect on crime. The authors did not investigate the possibility of displacement or diffusion of benefi ts, but it is l ik ely that neither occurred.

Wil li amson and McLafferty evaluated the impact of the CCT V intervention 18 months afterthe start of the programme and focused on crime inside the public housing projects andinside buffer zones of 0.1 to 0.5 miles radii around the projects. (For the buffer zones,only results inside 0.1 mil e are reported here, as the intervention i s less li k ely to aff ect

behaviour beyond this point.) The housing project that received the intervention did notshow any change in the total number of police-recorded crimes, either inside the project orinside the 0.1 mile buffer zone, while total crime in the control project dropped by 5.3 percent inside the project and 4.0 per cent inside the 0.1 mile buffer zone. When total crime isdisaggregated, a desirable programme effect is observed for major felonies in bothexperimental and control projects (see T able 3.1). H owever, the authors noted that thesubstantial decrease in major felonies around both public housing projects seems to be partof a larger downward trend that was occurring not only in Brooklyn but across New York City in the late 1990s (W ill iamson and M cLaf ferty, 2000, p. 7). T he authors investigatedthe possibility of displacement and diffusion of benefits and concluded that there is no clearevidence of either, as the change in crime around the two housing projects does not varypredictably with distance (ibid., p. 7).

25


36/68


T able 3.2 presents the results of a meta-analysi s of the CCT V evaluations in city centres orpublic housing. In order to carry out a meta-analysis, a comparable measure of effect size isneeded in each project. This has to be based on the number of crimes in the experimentaland control areas before and after the CCT V intervention, because this is the only informationthat is regularly provided in these evaluations. H ere, the odds ratio is used as the measure of effect size. For example, in Doncaster, the odds of a crime after given a crime before in thecontrol area were 2,002/1,780 or 1.12. T he odds of a crime after given a crime before inthe experi mental area were 4,591/5,832 or 0.79. T he odds ratio therefore was 1.12/0.79or 1.42. This was statistically highly si gnif icant (z = 9.24, p


37/68

Results

In order to produce a summary effect size in a meta-analysis, each effect size has to have astandard error. T hi s was one reason for choosing the odds ratio, wh ich has a k nownstandard error. T he average effect size (weighted according to the standard error of eachstudy) was an odds ratio of 1.02, whi ch was not statistically signif icant (z = 1.40, n.s.).T hus, pooling the data from the nine studies, there was no evi dence that C C T V led to areduction in crime.

T he nine effect sizes were signif icantly vari able (Q = 164.9, 8 df , p


38/68


39/68

Table 3.3: CCTV Evaluations in Public Transport

Author, Publi cation Context of Type and Sample Other Outcome R esearch Design R esultsDate, and Location Intervention Duration of Size Interventions M easure of and Before- After

Intervention Interest and Time PeriodData Source

Burrows (1979, Publ ic CCTV ; 12 E=4 stations N otices of Personal theft Before-after, E vs C1: robbery:1980), transport months on southern CCTV (also and robbery; experimental- -22.2% (9 to 7) vs

Underground, (subway) sector, special police BT P records control +23.1% (13 toLondon C1=15 other patrols 16); theft: -72.8%stations on preceded Before=12 months;(243 to 66) vssouthern CC TV ) A fter=12 months -26.5% (535 tosector, 393)C2=228 other E vs C2: robbery:Underground -22.2% vs +116.3%stations (43 to 93); theft:

-72.8% vs -39.4%(4,884 to 2,962)(desirable effect)

1. Webb and Public CC TV E=6 stations Passenger Robbery; BT P Before-after, E vs C1 (monthlyLaycock (1992), transport (expansion of); on south end alarms, records experimental-control average): -62.3%Underground, (subway) 26 months of Northern visible kiosk to (5.3 to 2.0) vs

London line, C1=6 monitor CCTV , Before=46 months; -50.0% (7.8 to 3.9)stations on mirrors, and A fter=26 months E vs C2: -62.3% vsnorth end of improved -12.2% (69.6 toline, C2=236 lighting N ote: special 61.1)other Under- policing used in E (desirable effect)ground stations during fir ststations 3 years (1985-87) N ote: for C2,

of before period G uardian A ngels


40/68

(i.e., fi rst 36 of patrols began in46 months of M ay 1989 (7 monthsbefore period); in into 26 months of 1988 (remaining after period)10 months of before period),policing activityreduced in E stations

2. Webb and Public CC TV E=1 station, Passenger Personal theft, Before-after, E vs C (monthlyLaycock (1992), transport (expansion of); C=1 station alarms, visible robbery, and experimental- average): robbery:Oxford Circus (subway) 32 months kiosk to assault; BT P control +47.1% (1.7 tostation, monitor CCTV , records 2.5) vs +21.4%Underground, and BT P Before= 28 months; (1.4 to 1.7); theft:London patrols After=32 months +11.0% (31.0 to

34.4) vs -1.9%(20.8 to 20.4);assault: +29.4%(1.7 to 2.2) vs+36.4% (1.1 to 1.5)(undesirable effect)

G randmaison and Public CC TV ; E=13 stations, None Crime (total Before-after, E vs C: total crimes:T remblay (1997), transport 18 months C=52 stations and multiple experimental- -20.0% (905 toM etro, M ontreal, (subway) offences); control with 724) vs -18.3%Canada police records statistical analyses (1,376 to 1,124);

robbery: -27.0%Before=18 months; (141 to 103) vsAfter=18 months -30.8% (312 to

216); assault:-27.5% (178 to129) vs +5.6%(233 to 246); total


41/68

theft and fraud:-15.5% (388 to328) vs -16.0%(507 to 426)(null effect)

Notes: Locations were in the UK unless otherwise specifi ed; BT P = Bri tish Tr ansport Police; E = experimental area; C = control area.


42/68


reductions of 50.0 per cent (f rom 7.8 to 3.9) and 12.2 per cent (f rom 69.6 to 61.1) incontrol areas 1 and 2, respectively . The authors found no evidence of robberies beingdisplaced to the two groupings of control stations or a third grouping (nine other stations atthe south end of the N orthern and V ictoria l ines) that did not receive the intervention.A lthough not stated by the authors in such terms, evidence of di ffusion of benefits is apparent.

T he multiple interventions that were implemented in the experimental and control stationsboth before and after the start of the programme, including special poli ce and G uardianA ngels patrols (see T able 3.3), mak e it diff icult to isolate the effect of CC T V , if any, onrobbery. On this matter, the words of the authors are instructive:

it seems likely that robbery has been kept down by improved management andstaffing of the system, including more revenue protection as well as station staff. Thepolicing changes may also have been helpful. It is also possible that the substantialphysical work involved in station modernisation and the introduction of automaticticket barriers in central area stations contributed by creating the impression of amore controlled and safer environment. (W ebb and L aycock , 1992, p. 11)

T he second Underground CCT V scheme evaluated by Webb and Laycock (1992) took pl ace inOxford Circus station located in central London. As noted above, this scheme did not justinvolve the expansion of C CT V , but also included other interventions: passenger alarms,visible kiosks to monitor CC T V operations, and patrols by the BT P. One station (TottenhamCourt Road) that did not receive CCT V cameras was used as the control station. T he schemewas evaluated after i t had been in operation for 32 months.

Disappointing results were reported for the programmes effects on passenger robbery, theft

(from the person), and assault. The authors noted that the robbery data were more reliablethan the data on theft; no mention was made of the reliability of the assault data. Table 3.3presents the results for before-after comparisons between the experimental and controlstations for all three offences. After 32 months, the monthly incidence of robberies increasedby almost half (47.1 per cent; from 1.7 to 2.5) in the experimental station, compared withan increase of more than one-fifth (21.4 per cent; from 1.4 to 1.7) in the control station. Theprogrammes impact on theft was also undesirable. The authors did not investigate thepossibility of displacement.

I n the M ontreal subway programme (G randmaison and T remblay, 1997), CCT V cameraswere installed in 13 stations (approximately ten cameras per station) over the course of 18months in the early 1990s. Fifty-two stations served as the control group. The programme

32


43/68

Results

was evaluated after 18 months of operation, and statis tical analy ses were conducted tocontrol for past crime trends in the experimental and control stations.

G randmaison and T remblay found an equal reduction in (police-recorded) crime in both theexperimental and control subway (M ontreal M etro) stations: -20.0 per cent and -18.3 per cent,respectively. H ence, there was li ttle evidence of any effect of the CCT V intervention. Themeasure of total crime included robbery, assault, purse snatching, other theft and fraud,vandal ism, and other offences. From 18 months before the start of the intervention to 18months afterwards, all categories of crimes were down in the experimental stations, whileal lcategori es except assaul t decreased in the control stations. T he authors di d notinvestigate the possibility of displacement or diffusion of benefits.

Overall, CCT V programmes in public transportation systems present confli cting evidence of effectiveness: two had a desirable effect, one had no effect, and one had an undesirableeffect on crime. H owever, for the two effective programmes in the London U nderground, theuse of other interventions makes it diffi cult to say wi th certainty that CCT V produced theobserved crime reductions, although in the programme by Burrows (1979) C CT V was more

than likely the cause.

T able 3.4 shows the results of a meta-analysis of the CCT V evaluations in publ ic transportsettings. I n all cases, the most comparable control area is used. T he odds ratio wassigni fi cant only in one case: the evaluation by B urrows (O R = 2.58, z = 6.39, p


44/68


45/68

Table 3.5: CCTV Evaluations in Car Parks

Author, Publi cation Context of Type and Sample Other Outcome R esearch Design R esultsDate, and Location Intervention Duration of Size Interventions M easure of and Before- After

Intervention Interest and Time PeriodData Source

Poyner (1991), Parking lot CC TV ; E=1 parking I mproved Theft from Before-after, E vs C (monthlyUniversity of Surrey, 10 months lot (no. 4), lighting and vehicles; experimental- average): theft from

G uildford C=1 parking foliage cut private control vehicles: -73.3%lot (no. 1) back (for both security (3.0 to 0.8) vs -93.8%E and C; only records Before=24 months; (1.6 to 0.1)E received After=10 months (undesirable effect)

CCTV)

1. T ill ey (1993b), Car park CC T V ; E=CC T V Security T heft of and Before-after, E vs C: theft of H artlepool 24 months covered car officers, from vehicles; experimental- vehicles: -59.0%

parks, C= notices of police control (21.2 to 8.7 pernon-CC T V CC T V , and records quarter year) vscovered car payment Before=15 months; -16.3% (16.0 toparks scheme A fter=30 months 13.4 per quarterN ote: no. of E year); theft fromand C car vehicles: -9.4%park s or (6.4 to 5.8 per

spaces n.a. quarter year) vs+3.1% (16.0 to16.5 per quarteryear)(desirable effect)

2. T ill ey (1993b), Car park CCT V ; E=1 car park, N otices of T heft of and Before-after, E vs C1: theft of Bradford 12 months C1=2 CC T V , from vehicles; experimental vehicles: -43.5%

adjacent car improved police records control (23 to 13) vs


46/68

parks, C2= lighting, and +5.9% (17 to 18);adjacent painting Before=12 months; theft from vehicles:street park ing A fter=12 months -68.8% (32 to 10)

N ote: C1 vs +4.5% (22 toreceived some N ote: a third C 23)CC T V is used, but is less E vs C2: theft of coverage for comparable than vehicles: -43.5% vslast 4 months C1 or C2 +31.8% (22 to

29); theft fromvehicles: -68.8% vs+6.1% (33 to 35)(desirable effect)

3. T ill ey (1993b), Car park CCT V ; E=3 car parks Li ghting, T heft of and Before-after, E vs C: theft of Coventry various (BA R , BO N , painting, and from vehicles; experimental- vehicles: -50.5%

WH I), C=2 fencing police control (91 to 45) vscar park s records -53.6% (56 to 26);(FAI, G R E) Before and after = theft from vehicles:

8 months (E) and -64.4% (276 to16 months (C) 101) vs -10.7%

(150 to 134)(desirable eff ect)

Sarno (1995, Car park CC TV ; E=3 car parks M ultiple (e.g., Vehicle crime; Before-after, E vs C1: -57.3%1996), London 12 months in part of locking police experimental- (349 to 149) vsBorough of Sutton Sutton police overnight, records control -36.5% (2,367 to

sector, C1=rest lighting) 1,504)of Su tton sector, Before=12 months; E vs C2: -57.3% vsC2=all of A fter=12 months -40.2% (6,346 toBorough of 3,798)Sutton (desirable effect)

N otes: All locations were in the UK ; E = experimental area; C = control area; n.a. = not availabl e.


47/68

Results

Four of the programmes had a desirable effect and one had an undesirable effect onvehicle crimes, which was the exclusive focus of each of the impact evaluations. Poyner(1991) evaluated a mul ti-component scheme at the Universi ty of Surrey in G ui ldf ord inwhich both the experimental and control parking lots (one in each condition) received up-graded lighting and foliage was cut back, but only the experimental parking lot receivedCCTV . T en months after the programme started, Poyner found that thefts from vehicles weresubstantially reduced in both the experimental and control parking lots. In the experimentalsite, the monthly average of incidents declined by almost three-quarters (73.3 per cent; from3.0 to 0.8), while in the control site, they were almost eliminated (a drop of 93.8 per cent; from1.6 to 0.1). Although the numbers are small , these resul ts suggest that C C T V hadundesirable eff ects on crime. H owever, the author concluded that there was evidence of diffusion of benefits.

Tilley (1993b) evaluated three C C T V programmes in car parks in the foll owing cities:Hartlepool, Bradford, and Coventry. Each scheme was part of the Safer Cities Programme.In H artlepool, C CT V cameras were installed in an unknown number of covered car park sand the control area included an unk nown number of non-CC T V covered car parks. Securi ty

personnel, notices of C C T V , and payment schemes were also part of the pack age of measures employed to reduce vehicle crimes. Twenty-four months after the programmebegan, thefts of and from vehicles had been substantially reduced in the experimentalcompared with the control car parks (see Table 3.5). Tilley (1993b, p. 9) concluded that, T he mark ed relative advantage of CCT V covered parks in relation to theft of cars clearlydeclines over time and there are signs that the underlying local trends [an increase in carthefts] begin to be resumed. The author suggests that the displacement of vehicle thefts fromcovered to non-covered car parks may be partly responsible for this.

I n B radford, CC T V cameras were install ed in one multi-story car park in the city centre.N otices of CCTV , improved lighting, and general improvements in the form of painting werealso implemented in the car park . T wo adjacent car park s and adjacent street park ingserved as the control areas. A third control area - a city centre sub-division - was also usedby T i ll ey, but it i s considered here to be less comparable than the other two w ith theexperimental area, and thus has not been used in experimental-control comparisons. It isimportant to note that the first control area - two adjacent car parks - also received someCCT V coverage, for the last four months of the 12-month follow-up period. Twelve monthsinto the programme, thefts of and from vehicles showed substantial reductions in theexperimental area, whi le both crimes showed increases in the two control areas (see Table3.5). A gain, displacement was not measured, and numbers of crimes were small.

37


48/68


I n the third car park CCT V scheme evaluated by T il ley (1993b), in C oventry, cameras wereinstalled at different times in five car parks, and not in a sixth. Yearly data for thefts of andfrom vehicles (for January to August) were presented for six years (1987-1992). Three carparks (Barrack s, Bond Street, W hitefriars) were designated as experimental car parks,because crime data were available for at least one year before and one year after theinstallation of cameras. Two car parks were designated as control car parks, either becausecameras were not install ed in them (Fai rf ax Street) or because the cameras were onlyinstalled in the last year (G reyfriars). For the control car park s, crime data in the two yearsbefore the average year of C CT V installation (1989) were compared with crime data in thetwo years afterwards. T he sixth car park (C ox Street) was not included in the analysesbecause cameras were installed in i t in the first year. There were other (lighting, paintingand fencing) improvements in these car parks during this time period. It was found that theftfrom vehicles decreased more in the experimental car parks, but theft of vehicles did not.

The most recent evaluation of the impact of C CT V on vehicle crime was carried out in the LondonBorough of Sutton (Sarno, 1995). CCTV cameras were installed in three car parks (experimentalarea) in one part of the Sutton police sector at high risk of vehicle crimes, and two control areas

were established: (1) the remainder of the Sutton police sector and (2) all of the Borough of Sutton. The first control area was considered to be comparable to the experimental area.

T he programme was evaluated after its first 12 months of operation. T otal vehicle crimes (theft of, theftfrom, criminal damage to, unauthorised taking of vehicles and vehicle interference; Sarno, 1995, p.22) were reduced by 57.3 per cent (from 349 to 149) in the experimental area, but there werealso lesser reductions in control areas 1 (36.5 per cent; from 2,367 to 1,504) and 2 (40.2 per cent;6,346 to 3,798). The author did not measure diffusion of benefits.

T ill ey (1993b) attempted to investigate mechanisms that may or may not have played a role inthe success of CCT V in preventing vehicle crimes in car park s. However, his conclusionsabout mechanisms were almost all negative. For example, the true probability of beingcaught did not increase, offenders were not removed by being caught, CCT V images wereinsufficiently clear to identify offenders, there was little increase in car park usage following theinstall ation of CCT V (so no convincing evidence of increased natural surveil lance or of cautious drivers being attracted to these car parks), and rarely any effective deployment of security staff. So why did CCT V allegedly have any effect? Ti lleys main suggestion was thatCCT V had an effect when it was combined wi th other crime prevention measures, but thisfai ls to address the problem of determining whether the eff ect was caused by C CT V or bythese other measures. Tilley made little attempt to address threats to internal validity (Cook andCampbell, 1979; Shadish et al., 2002).

38


49/68

Results

T able 3.4 shows the results of a meta-analysis of the five CC T V evaluations in car parks. In fourcases, the odds ratios showed a signif icant and desirable effect of C CT V . In the other case(Poyner, 1991), the effect was undesirable, but the small numbers meant that the odds ratiowas not significant (z = 1.35). When all five odds ratios were combined, the overalloddsratio was 1.70 (z = 7.45, p


50/68


Figure 3.1: CCTV evaluations1.0

Bradford

Underground S

Coventry

Birmingham

H artlepool

Sutton

Doncaster

Underground N

Burnley

M ontreal

Cincinnati F

Cincinnati N

Cincinnati H

Newcastle

Underground C

New York

Cambridge

Mean

N ote: Odds ratios and confidence intervals on logarithmic scale

40


51/68

4. Conclusions

Summary of main findings

A number of targeted and comprehensive searches of the published and unpublishedli terature and contacts with leading researchers produced 22 CC T V evaluations which met ourcriteria for inclusion in this review; 24 evaluations did not meet the inclusion criteria (mainlybecause they had no comparable control condition) and were excluded. The criteria forinclusion called for C CT V programmes whi ch employed rigorous evaluation designs to assesseffects on crime, with the minimum design involving before-and-after measures of crime inexperimental and comparable control areas.

Setting the threshold any higher - for example, requiring randomised experimental designs- was impractical, because no CC T V programme has been evaluated with this degree of scientif ic ri gour. T herefore, the methodological cri teri a used here sought to achieve a

balance between weak (e.g., simple one group, no control group, before-after designs) andstrong science. Faced with a similar dilemma, Sherman and his colleagues adopted thesame approach: T he report [Preventing Crime] tak es the middle road between reachingvery few conclusions with great certainty and reaching very many conclusions with very littlecertainty (1998, p. 6).

T he 22 included evaluations were carried out in three main settings: (1) city centres andpubl i c housing, (2) publ ic transport, and (3) car park s. E valu ations were not evenlydistributed across the three settings. The largest number of evaluations was in the city

centre/public housing setting (N =13).

Of the 22 included evaluations, half (11) found a desirable effect on crime and five found anundesirable eff ect on crime. Five evaluations found a nul l eff ect on crime (i.e., clearevidence of no effect), while the remaining one was classified as finding an uncertain effect oncrime (i .e., unclear evidence of an effect).

R esults from a meta-analysis provide a clearer picture of the crime prevention effectivenessof C C T V . From 18 evaluations - the other four did not provide the needed data to beincluded in the meta-analysis - it was concluded that C C T V had a signif icant desirableeffect on crime, although the overall reduction in crime was a rather small four per cent.H alf of the studies (nine out of 18) showed evidence of a desirable effect of C CT V on crime.

41


52/68


A ll nine of these studies were carri ed out in the UK . C onversely , the other nine studiesshowed no evidence of any desirable eff ect of C C T V on crime. All fi ve North A mericanstudies were in this group.

T he meta-analysis also examined the effect of C CTV on the most frequently measured crimetypes. I t was found that CC T V had no effect on violent crimes (from five studies), but had asignificant desirable effect on vehicle crimes (from eight studies).

Across the three settings, mixed results were found for the crime prevention effectiveness of CCT V . In the city centre and public housing setting, there was evidence that CC T V led to anegligible reduction in crime of about two per cent in experimental areas compared withcontrol areas. CC T V had a very small but signifi cant effect on crime in the five U K evaluations in this setting (three desirabl

Documents

Crime Prevention Effects of Closed Circuit Television