PROTOCOL: The Effectiveness of Neighbourhood Watch

Abstract

Neighbourhood watch (also known as block watch, apartment watch, home watch and community watch) grew out of a movement in the US that promoted greater involvement of citizens in the prevention of crime (Titus, 1984). One of the first recorded neighbourhood watch programmes in the US was the Seattle Community Crime Prevention Project launched in 1973 (Cirel et al., 1977). One of the first recorded neighbourhood watch schemes in the UK was the Home Watch programme implemented in 1982 in Cheshire (Anderton, 1985). Since the 1980s, the number of neighbourhood watch schemes in the UK has expanded considerably. It was recently estimated from the results of the 2000 British Crime Survey that in that year over a quarter (27%) of all households (approximately six million households) in England and Wales were members of a neighbourhood watch scheme (Sims, 2001). The National Neighbourhood Watch Association estimated for the same time period that there were 155,000 schemes in operation. It is reasonable for researchers to ask, therefore, whether such large investments in time and resources are effective in reducing crime (the primary objective of neighbourhood watch). One of the first evaluations of neighbourhood watch in the UK was conducted in Bristol (Veater, 1984). The study included pretest and posttest victim and public attitude surveys conducted in the scheme area. Police-recorded crime data were collected for an adjacent area for comparison. The results of the victimisation surveys showed that the offence rate fell from a total of 247 reported offences in 979 households (a rate of 25 offences per 100 households) in the first round to 174 reported offences in 1,060 households (a rate of 16 offences per 100 households) in the second round. During the period of the research, the number of police-recorded crimes committed in the control area increased. The report concluded that the programme was effective. In many respects, the evaluation was advanced for its time and included victim surveys as well as police recorded crime. However, like many of the evaluations that followed, there were a number of methodological problems. The changes in victimisation rates were not subjected to statistical significance testing. The response rate for both surveys was fairly low (just over 50%) and potential differences in the demographic profile of the experimental and control areas were not taken into account. Another early study conducted by the police in Merseyside in England also experienced problems of evaluation design and research quality (Jenkins and Latimer, 1987). The study was based on a comparison of police-recorded crime rates for a period 12 months before and 12 months after the launch of the Home Watch scheme in four areas of Merseyside. The four schemes were implemented in areas covering between 43 and 97 households. A public attitudes survey was conducted among residents. In three of the four areas the number of residential burglaries was lower in the posttest period than in the pretest period and the schemes were claimed to have been a success. However, the numbers involved were very small. In the first area, the number of burglaries fell from 7 to 6, in the second area, from 19 to 5, and in the third area, from 4 to ‘nearly zero’. One of the most sophisticated evaluations of neighbourhood watch was conducted in nine areas in Chicago (Rosenbaum et al., 1985). The research was based on a quasi-experimental, untreated control group design with pretest and posttest surveys. Crime and public attitude surveys were conducted in programme and non-programme areas before the launch and again one year following the launch. The results showed that two programme areas experienced a significant increase in the number of victimisations per respondent over the study period. One area showed a decrease in victimisations, which was ‘marginally significant’. The other areas showed no change in victimisation. The results of existing reviews of the evaluation research on neighbourhood watch show that the results are mixed. However, there is one notable theme to the results. Studies conducted by the police tend to conclude that neighbourhood watch is effective in preventing crime, while studies conducted by independent researchers tend to conclude that neighbourhood watch is ineffective (Rosenbaum, 1987, Bennett, 1990). The body of research is variable in quality. In the absence of a systematic review, it is difficult to arrive at an overall conclusion about the effectiveness of neighbourhood watch in preventing crime. In order to determine programme effectiveness from evaluation research, it is necessary to conduct a rigorous review of the literature that controls for research quality and standardises results. The most frequently recorded mechanism by which neighbourhood watch is supposed to reduce crime is as a result of residents looking out for suspicious activities and reporting these to the police. The link between reporting and crime reduction is not usually elaborated in the literature. It has been argued that visible surveillance might reduce crime as a result of its effect on the perceptions and decision making of potential offenders. Hence, watching and reporting might deter offenders if they are aware of the propensity of the local residents to report suspicious behaviour and if they perceive this as increasing the risks of being caught. It is also possible that neighbourhood watch schemes might reduce crime as a result of an increase in 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. Neighbourhood watch might also reduce crime through various mechanisms of social control. Neighbourhood watch might enhance community cohesion, community activism, and collective efficacy, which might also increase the ability of communities to control crime in residential areas. Evaluations of neighbourhood watch need to take into account the problems of causal order when investigating mechanisms. It is possible that neighbourhood watch schemes are implemented in areas that already have low levels of crime and high levels of social control. Hence, the evaluation would need to take into account the temporal order of change in both the independent and dependent variables. This should be possible in studies that include pre-test and post-test measures. Neighbourhood watch is often implemented as part of a comprehensive package. The typical package is sometimes referred to as the ‘big three’ and includes neighbourhood watch, property-marking and home security surveys. Some programmes include a third or fourth element such as a recruitment drive for special constables, increased regular foot patrols, citizen patrols, educational programmes for young people, auxiliary police units, and victim support services. Neighbourhood watch schemes in the United States are often based on small blocks of dwellings and include no more than 20–30 households. The size of the Block Watch units in the Seattle programme (one of the first programmes in the United States) varied between just 10 and 15 households. Schemes in Britain tend to be larger and some cover residential areas of 3,000 or more households. Neighbourhood watch schemes can be both public and police initiated. Schemes launched in the UK during the early period of a programme tended to be police initiated (e.g. the early neighbourhood watch schemes in the Metropolitan Police District). More recently, neighbourhood watch schemes have been launched mainly at the request of the public. Some police departments continue initiating their own schemes, even when the programme is fully developed. A programme implemented in Detroit, for example, maintained a section of police-initiated schemes in order to promote neighbourhood watch in areas that were unlikely to generate public-initiated requests. Block Watches are usually run by a block captain who is responsible to a block coordinator or block organiser. The block co-ordinator acts as the liaison person to the local police department. Neighbourhood watch schemes in the UK often include street co-ordinators (equivalent to block captains) and area co-ordinators (equivalent to the block organiser). There is little information in the literature on the number and type of neighbourhood watch meetings. The evidence that does exist suggests that some schemes have public meetings that involve all of the residents participating in the scheme, while others have meetings that involve only the organisers of the scheme. The funding of neighbourhood 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 United States 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. Apart from police funding, the majority of schemes are encouraged to raise some funds from other sources such as voluntary contributions, local businesses, and the proceeds of fetes, and raffles. There is strong consensus in the literature that the main aim of neighbourhood watch is crime prevention. There are small variations among the programmes in terms of which crimes are targeted. The vast majority of programmes identify residential burglary as the sole or most important target crime of neighbourhood watch. Some programmes focus solely on residential burglary, but most list other offences which it is hoped neighbourhood watch will reduce. The list of other offences is sometimes explicit (e.g. ‘street robberies, auto thefts and vandalism’) or general (e.g. ‘street crime’ and ‘property crime’). The primary aim of the review is to assess the effects of neighbourhood watch on crime. The main types of intervention studied are neighbourhood watch schemes. In practice, these are called many different things (e.g. home watch, block watch, home alert, and street watch) and it is not possible to identify the type of intervention solely from the name used. Hence, the type of intervention will be defined by its characteristics and mechanisms rather than by its name. Watch programmes can be based on a diversity of populations, including boat owners, farmers, and business employees, and a diversity of locations, including car parks, yacht marinas, and the countryside. We aim to evaluate schemes based on residents living in neighbourhoods. The mechanisms described above rule out neighbourhood wardens and similar citizen patrols. Citizen patrols are based: (a) on the appointment of residents to a particular role, and (b) on agreement to conduct particular duties such as patrolling the streets. Watch schemes are based solely on residents operating in their capacity as residents. When crime measures are based on police recorded crimes, the main outcome measure will be the total number of crimes recorded in the areas studied. When crime measures are based on victimisation surveys, the main outcome measures will be the total number of victimisations, the prevalence of victimisation, and the incidence of victimisation per household. Other outcome measures used in the research will be listed. These will not be used in the meta-analyses. However, they will be made available and discussed in a separate narrative review. Our aim is to include the ‘best’ evaluations. However, the decision about what to include needs to be pragmatic and should not be so stringent as to rule out all available evaluations. The criteria for selecting rigorous evaluations would be based on the Maryland Scientific Methods Scale (SMS) (Sherman et al., 1997). This is a five point scale ranging from level 1 (the weakest design) to level 5 (the strongest design) in terms of overall internal validity. Sherman et al. argue that evaluations should be at least level 3 in order to conclude, with a reasonable level of certainty, that the programme worked. The current review of evaluations would also use this level 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 the review of neighbourhood watch is that they are based on both before and after surveys and experimental and comparable control areas. This would include randomised experiments (including pre-test/post-test and controls) and post-test only designs (including controls) when the treatment (neighbourhood watch) was implemented by random allocation. However, as far as we know, there are no evaluations of these kinds. Hence, in practice the review is likely to be based on pretest-posttest designs with controls. The review would also be based on studies that used sample sizes large enough to generate sufficient power to detect a significant programme effect. Welsh and Farrington (2002) propose (in relation to measurements of crime) that there should be at least 20 crimes before the intervention. We will adopt the same measure and only exclude evaluations with clearly inadequate sample sizes (e.g. less than 20 crimes in the before period in the experimental or control groups). The results need to be presented in a way that can be used to calculate effect sizes. The study should include effect sizes or provide numbers from which effect sizes could be calculated. In some cases, when the information is not provided, it might be possible to contact the author to obtain the required information. If it proves to be impossible to calculate effect sizes, the study will be rejected from the main systematic review. However, the results of excluded studies will be included in a separate narrative review. The review includes published and unpublished literature. It will be based on documented evaluations. There is no restriction on country of origin. The evaluations must be available in English. There is no restriction on source sector (e.g. academic, government, policy, voluntary, etc.). There is no restriction in terms of year (e.g. implemention, study, or publication). There will be no restriction of the time period covered by the evaluation (e.g. short-term or long-term effects). We will start with the following search terms: neighbourhood watch, neighborhood watch, street watch, block watch, apartment watch, home watch, community watch, home alert, block association, crime alert, block clubs, crime watch, ‘big three’. This would involve assessing each study on each of the five criteria of validity. The main types of research design used for evaluating neighbourhood watch schemes have been discussed in previous sections. The most common is some kind of quasi-experimental design. The strongest quasi-experimental designs are based on before and after measures in experimental and comparable control areas. It is expected that this will be the most common design used among studies included in the review. The studies excluded from the review will be those that have adopted a more restricted version of the above. Excluded studies will be listed, but not included in the meta-analysis. Examples of included and excluded quasi-experimental designs can be found in the first section of the protocol. The guidelines ask whether multiple outcomes from the same or related evaluations are independent data points. When multiple outcome measures are provided (e.g. multiple outcome measures of crime) we would list the results for each measure. However, the meta-analysis would be based on only one measure. The measure chosen would be guided by a predetermined method of prioritising the results (e.g. victimisation data used in preference to official record data). The main measure of effect size will be the odds ratio, calculated from numbers of crimes before and after the intervention in experimental and control areas (as in Welsh and Farrington, 2002). One of the reviewers (KH) initially will extract information from the study using a specially designed data extraction spreadsheet. Another reviewer (TB) will check all of the entries against the original study reports. A copy of the entire data set, with a list of any disagreements, will be sent to the third reviewer (DPF) for checking and arbitration over disagreements. Where possible, the author(s) of the report will be contacted to help resolve any ambiguities and to provide missing information. The information extracted will include: author, publication date, study date, location, physical context of the intervention, type of intervention, duration of the intervention, duration of the evaluation, sample size, other interventions employed at the time, outcome measures, data source, research design, results, author(s) conclusion. Several meta-analyses will be carried out covering different types of programme and different types of crimes. The odds ratios calculated for each evaluation will be combined as summarised in Lipsey and Wilson (2001). Inverse variance weights will be used to combine effect sizes. When sufficient information exists, the effects of mediating and moderating factors will be used to refine the meta-analysis. Mediating variables are part of the causal sequence (e.g. surveillance and reporting to the police) and moderating variables are contextual variables that might modify the causal sequence (e.g. community and individual characteristics). The systematic review will be based on quantitative evaluations only. However, we will collect information on qualitative research and present the results of these studies separately as a narrative review. Qualitative findings may derive from the qualitative sections of quantitative research (both those included and excluded from the systematic review) and from wholly qualitative studies. The principal reviewer in collaboration with the other reviewers would update the review once every two years. No additional acknowledgements. One of the authors (TB) has conducted an evaluation of neighbourhood watch that might be included in the review. None at present.