POP Center Responses Video Surveillance of Public Places, 2nd Ed. Page 8
Appendix B: Site-specific Evaluations of Video Surveillance in Public Places
The following table summarizes a number of CCTV systems and the results of their evaluations. It is not an exhaustive list, as some studies may have been inadvertently omitted during the literature search for this guide. Also, a number of studies have been excluded. The main reasons for exclusion were when the evaluation report did not include sufficient information to corroborate any reported crime reduction, or where the evaluation was conducted by a party perceived to be heavily invested in the system.† This commonly occurred when a system was reported as a success in a newspaper article based solely on the comments of a city manager or local police department. When some evaluations reported findings that did not appear to accurately reflect the changing pattern of crime, they were either excluded, or the language was changed to a more general tone. As a result of this last caveat, if you require further information you should refer to the original study reports. This is the best way to judge the reliability of the findings and conclusions, as the quality of studies varies considerably.
† This is not to suggest or imply an inappropriate behavior on the evaluator’s part. Simply, the evaluator’s impartiality cannot be guaranteed and, therefore, the evaluation was excluded.
The table below emphasizes studies that have a strong quantitative component. This is not intended to negate the value of qualitative analysis, but to reflect the likely audience for the report. Most CCTV systems are implemented to tackle, at least as one aim, levels of reported crime. These are usually apparent in police recorded-crime records and so the table reflects more positively on reports that demonstrate they have examined and evaluated recorded-crime statistics in a robust manner. Studies are ordered by implementation date, with the most recent first.
Location | Camera organization | Implement-ation | Effect on crime | Effect on fear of crime | Operation | Evaluation | Research design |
Gothenburg, Sweden | 28 cameras in 3 neighborhoods 5 cameras are at each spot, 4 are fixed and 1 is movable | January 2018 and April 2018 | CCTV was associated with a reduction in violence but no significant change in property crime or crime clearance. Effects vary by site. | No information available | Cameras can be actively monitored but the extent to which this occurred is unknown | Gerell (2020) | Changes in crime and crime clearance in the 3 neighborhoods are compared to 6 control neighborhoods. Weighted displacement difference models are used to understand changes in crime and chi-square test are conducted for changes in crime-clearance rates. |
Philadelphia, Pennsylvania | 86 CCTV cameras grouped into 13 clusters | January 2003 and December 2012 | Cameras had no effect on the level of violent street felonies. No significant impact was found for disorder crimes. | No information available | Cameras have a patrol function where they patrolled a specific area but can also be operated by the camera monitor. Cameras also recorded all footage 24 hours a day 7 days a week and footage was stored for 12 days. | Ratcliffe & Groff, 2019 | A quasi-experimental repeated measure design which considered counts of crime events for both violent street felonies and disorder crimes in 13 spatial units of 120 temporal periods using a multilevel random effects model |
Newark, New Jersey | 64 CCTV cameras grouped into 38 schemes | July 2011 and September 2011 | The experimental strategy was associated with significant reduction in violent crime and social disorder in treatment areas compared to control areas | No information available | Cameras were monitored normally by 2 camera operators. During the experiment an additional operator was added to monitor treatment cameras. | Piza, Caplan, Kennedy & Gilchrist, 2015 | A randomized block design control trial was used to assign each of the 38 CCTV schemes to either treatment or control group. Schemes were grouped into pairs based on their calls for service for violent crime, social disorder, and narcotics activity. |
Surrey, British Columbia | 12 cameras were installed at the car park: 11 fixed and one adjustable cameras | August 2009 | Police data did not show much of an impact in of CCTV. Insurance data also do not show much of an impact. | No information available | Camera recordings were stored for 7 days and were available upon request | Reid & Andresen, 2014 | Structural break tests employed via linear regression were used to assess three trends in a variety of spatial units controlling for seasonal effects |
Newark, New Jersey | 73 dome cameras in plain view of pedestrians | March 2008 and July 2008 | No significant difference between strategically and randomly place cameras. Significant decrease in auto thefts. No significant displacement and small diffusion benefits. | No information available | No information available | Caplan, Kennedy, & Petrossian, 2011 | Quasi-experimental design. Geographic information system mapping was used to create the 73 camera and control location boundaries. Crime data was compared for 13 months before and after camera installation. |
Malaga, Spain | 17 cameras in the two square miles of the main shopping center | March 2007 | CCTV system did not significantly reduce crime. There is a possible displacement effect occurring and is more evident for crimes against property. | No significant difference in the fear of being a victim of crime | Cameras are pan, tilt, zoom high resolution models | Cerezo, 2013 | Quasi-experimental design. Included experimental area, control area, and buffer areas. Considered both crime rates and victimization rates. |
Philadelphia, Pennsylvania | 2 different camera types were used in the pilot program: 8 pan, tilt, zoom and 10 regular cameras. Cameras were located at 10 different sites. | January 2005 and August 2007 | The introduction of CCTV cameras was associated with a 13% reduction in all crime in the target areas. However, not all sites showed a benefit from the camera placement. | No information available | The 8 pan, tilt, zoom cameras were actively monitored. 10 cameras did not allow for live monitoring although officers nearby with the correct equipment could theoretically view feed from cameras. The system recorded up to 5 days of activity. | Ratcliffe, Taniguchi, & Taylor, 2009 | Two different evaluation techniques are used: hierarchical linear modeling and weighted displacement quotients |
Philadelphia, Pennsylvania | 10 cameras monitored by police department installed in 4 areas and 8 cameras that record continuously installed in 8 locations | July 2006 | 13% reduction in crime through end of August 2017 overall. 4 camera sites had no reduction and 4 sites reduced crime and diffused benefits to surrounding streets. | No information available | 10 cameras monitored by officers. 8 cameras officers can monitor from in patrol cars while nearby and cameras feed set continuously to hard drive. | Ratcliffe & Taniguchi, 2008 | Hierarchical linear modeling was used to control for seasonal effects and preexisting temporal trends at each camera location. In order to also assess the effects of different cameras at different locations weighted displacement quotient analysis was utilized. |
Los Angeles, California | 5 cameras in Hollywood Boulevard & 6 cameras in Jordan Downs | February 2005 for Hollywood Boulevard, and October 2006 for Jordan Downs | Not statistically significant differences for violent crime, property crime, or displacement | No information available | All cameras were active monitoring systems | Cameron, Kolodinski, May & Williams, 2008 | Quasi-experimental design to examine monthly crime data before and after the introduction of CCTV. Areas were categorized into target, buffer, or control areas. Relative effect size statistical tests were conducted. |
Schenectady, New York | 11 cameras with locations determined based on spatial concentration of crime | October 2003 – January 2007 | Total crime did decrease in the 150-foot area around the camera. Cameras were associated with declines in person rather than property crime. Cameras were very successful at reducing disorder. Cameras that were more visible were better at reducing crime. Mixed results were found for displacement and diffusion benefits. | No information available | Cameras operated on a patrol sequence with an average of 7 present viewing locations. Little viewing of live feed and footage was stored for 2 weeks. | McLean, Worden, & Kim, 2013 | Interrupted time series analysis was conducted. Each of the 4 time periods where cameras were introduced were treated as interventions in the model. |
Kabukicho, Tokyo, Japan | No information available | March 2002 | Reduction in vehicle crime, slight reduction in violence, substantial reduction in larceny, within 50 meters of cameras | No information available | No information available | Harada et al., 2004 | Geocoding crime events improved accuracy and better determined which crimes were within the CCTV area |
Cincinnati, Ohio | Cameras sited in 3 city locations | Early 1999 | Some reduction in calls for service and anti-social behavior in 2 sites (one with some diffusion), but an increase in anti-social behavior in a third location, as well as some displacement on implementation | No information available | No information available | Mazerolle et al., 2002 | An ARIMA time series analysis of data derived from interpretation of video footage was combined with police incident data |
Oslo, Norway | 6 cameras | January 1999 | Decrease in robbery/theft from person and bicycle theft | None | Civilians working at a police station | Winge & Knutsson, 2003 | The data have some limitations, and the surveys are not large; however, the incident data were examined for experiment, control, and displacement areas |
East Brighton, UK | 10 cameras in a housing project | Summer 1998 | Crime continued a long-term increase | Feelings of lack of safety continued after CCTV’s introduction | No information available | Squires, 2003 | Some factors were out of the researchers’ control. There were potentially significant differences between pre- and post-survey groups, and the crime analysis does not break down the data into more meaningful offense categories. |
Greater Easterhouse, Glasgow, Scotland | Not reported | May 1998 | No overall crime reduction. Drug offenses and violent crime increased, but at a lower rate than in other areas. Other crime types not reported in the paper. | No information available | Civilian operators working at a police station | Hood, 2003 | Adequate, but not all quantitative results reported |
Camberwell, London, UK | 17 cameras in a town center | January 1998 | Street, vehicle, and violent crime decreased at a faster rate than before CCTV’s introduction, while the buffer and comparison areas saw an increase in crime | Of public surveyed, who knew about the cameras, 69% felt safer | Civilian, based at a public car park and linked to a police station | Sarno et al., 1999 | 4 years of crime data examined and supported with numerous qualitative approaches |
East Street, London, UK | 12 cameras covering a street market | January 1998 | Vehicle crime and criminal damage decreased, though street crime increased (mainly in theft from the person; robberies decreased) | Of public surveyed who knew about the cameras, 53% felt safer | Civilian, based at a public car park and linked to a police station | Sarno et al., 1999 | 4 years of crime data examined and supported with numerous qualitative approaches |
Five UK towns | Varied | March to July 1997 | Assault-related emergency room visits decreased, recorded violence increased, suggesting that police intervention due to CCTV surveillance increased arrests and reduced the escalation of violence | No information available | No information available | Sivarajasingam, Shepherd, & Matthews, 2003 | 2 years of pre-and post-intervention data were explored for 5 experiment and 5 control towns and cities |
Ilford, Essex, UK | Town center. Number of cameras not available. | May/June 1997 | Reduction over 5 months for every crime type examined. Lesser reductions outside implementation area for a number of crime types. Crime in the CCTV area also declined compared to the same months in the previous year. | Modest improvement after CCTV implementation | No information available | Squires, 1998 | A longer data period would have been able to correct the apparent seasonality |
Elephant and Castle, London, UK | 34 cameras around a shopping center | January 1997 | Recorded crime fell 17% in both target and buffer areas. Steep decline in street robberies attributed to CCTV. | Of public surveyed who knew about the cameras, about 60% felt safer. | Civilian, based at a shopping center and linked to a police station | Sarno et al., 1999 | 4 years of crime data examined and supported with numerous qualitative approaches |
Amsterdam, The Netherlands | 29 cameras, in 3 areas, with variable viewing hours | Early 1997 to mid-2001 | General reduction in crime levels. Some displacement to other areas, though still a net reduction. Some immediate diffusion of benefits. | Slight improvement in only one area | Variable hours, with two systems operational only during peak hours | Flight, Heerwaarden, & Soomeren, 2003 | The systems were evaluated by means of an analysis of police records for one year before, and one year after CCTV implementation at each site, though the quantitative data were not fully explored |
Gillingham, UK | 7 town center cameras | 1997 | Reduction in vehicle crime and robberies | No information available | Civilian | Griffith, n.d. | The evaluation compared crime rates in the target area with a comparison site in a similar town with 5 years of aggregated data |
Peckham, London, UK | 14 cameras in a public retail area | October 1995 | Inconclusive, due to limitations in access to recorded crime data | Of public surveyed who knew about the cameras, about 60% felt safer | Civilian, based at a public car park and linked to a police station | Sarno, Hough, & Bulos, 1999 | Crime analysis was complicated by limited access to crime data due to the introduction date of a crime recording system. Researchers did manually gather data for a pre- and post-implementation period. Limitations in crime data outside the researchers’ control. |
Burnley, UK | No information available | 1995 | Substantial decline in most crime types. Some diffusion effect for most crime types. | No information available | No information available | Armitage, Smyth, & Pease, 1999 | The study used a long-time series of data and also explored hourly temporal patterns |
Glasgow, Scotland | 32 city center cameras | November 1994 | Marginal, though the system has helped with some major crime investigations | Marginal | Civilian | Ditton et al., 1999 | 3 years of crime data had seasonal variation removed before trend analysis, and pre- and post-surveys were conducted in control areas |
Newcastle Upon Tyne, UK | 16 city center cameras | December 1992 | Reduction in burglary (57%), theft from vehicle (50%), vehicle theft (47%), and criminal damage (34%). Reductions occurred in areas outside the CCTV area, but not to the same level. | No information available | Police and civilians in a police station | Brown, 1995 | Crime data examined for 26 months before, and 15 months after, implementation |
Airdrie, Scotland | 12 town center cameras | November 1992 | Overall, 21% reduction, especially crimes of dishonesty and vandalism. Some crime types increased, but this may be due to increased detections. | No information available | Civilian operators working at a police station | Short & Ditton, 1996 | Researchers controlled for seasonality and used a long-time series before and after CCTV implementation |
Birmingham, UK | 9 city center cameras initially | 1991-1992 | Apparent crime control benefits (in robbery, burglary, and theft from person). Possible displacement of robbery and theft from person out of the area, as well as displacement of offending from vehicle theft to theft from vehicles. Some evidence of reduced personal victimization in CCTV area. | A positive change only in people who were aware the cameras had been installed | Civilian staff employed by the police | Brown, 1995 | Nearly 4 years of data were used for the study, but the data were aggregated only to monthly beat counts |
London, UK | 4 different drug markets. Camera organization changed by site. | 1990s | Effective in dispersing drug markets in 2 areas; in a third, users appear to have adapted to the cameras’ presence | No information available | No information available | Edmunds et al., 1996 | Not able to assess from the information provided |
King's Lynn, UK | 60 cameras around the town | 1987-1994 | Vehicle crime continued ongoing reduction and reduced at a more significant rate compared to the surrounding police division. Burglary reduced in the evaluated CCTV area. Within 2 years, vehicle crime in the camera areas declined to nearly zero. | No information available | Civilian | Brown, 1995 | The evaluation was limited to cameras overlooking car parks only. The number of crime events is low, limiting the application of any statistical measures. |
In 2005 a large UK Home Office study was published (Gill & Spriggs, 2005). This study evaluated 13 CCTV projects comprising 14 separate systems. The systems were implemented in a variety of ways, including at public car parks, in town centers, in residential areas and housing estates, and in hospital areas. Furthermore, the systems varied in type. Some were fixed, others redeployable. Some were digital, others analogue. Some were monitored full time, others for less than 24 hours a day. The variations in the system therefore had an impact on the success of the system. The table below aims to concisely summarize the ten systems relevant to this report.
Research design: Strong. Police recorded crime statistics were examined in both the target area and the comparison areas. Some projects were also evaluated for displacement effects. Where possible (as was the case in nearly all studies) at least one to two years of pre-and post-intervention crime data were gathered. Time-series techniques were used to control for seasonal fluctuations. In 12 of the areas, public attitude surveys explored the public’s perceptions of the CCTV systems and fear of crime. Researchers also identified other crime prevention measures taking place in the evaluation areas so the individual contribution of CCTV could be explored. Note that in the original report the names of the locations were changed to preserve anonymity.
Location | Camera organization | Effect on crime | Effect on fear of crime |
City outskirts | 47 cameras installed in a deprived area of residential, park, hospital, and light industrial land use | Significant reduction in crime | 14% fewer respondents reported being worried about crime after CCTV installation. Other measures less clear. |
South City | 51 cameras added to an existing system in a mixed affluent/deprived city center area in southern England | 10% reduction in crime, though there was a 12% reduction in the control area with no CCTV. Increased public order. | About 7% fewer respondents reported being worried about crime after CCTV installation |
Shire Town | 12 cameras installed in the town center of a Midlands former mining town | Crime reduced 4% in the town, while it increased 3% in the control site | 12% fewer respondents at night and 4% during the day reported being worried about crime after CCTV installation. Greater reduction at night in control area. |
Market Town | 9 evaluated cameras. 2 new cameras, with further cameras added to an existing system, in the center of an affluent market town. | Crime increased 18% in the town, while only increasing 3% in the comparison site | No information available |
Borough Town | 40 new cameras installed in a small-town center aiming to reduce retail crime, alcohol problems, and criminal damage | No change in crime in the town center, while crime increased 14% in the comparison area | Fear of crime reduced |
Northern Estate | 11 new cameras introduced to a deprived public housing project in northern England | Crime decreased by 10% in the target area (especially burglary). Crime in the comparison area increased by 21%. | 3% fewer respondents reported being worried about crime after CCTV installation. Similar reductions in control area. |
Eastcap Estate | 12 new cameras (10 evaluated) implemented into a deprived public housing project in southeast England | Crime increased in the target area, but only by 2% compared to a 5% increase in the control site. Some displacement within the target area. | 3% increase in feelings of safety, matched with a similar level in control areas |
Dual Estate | 14 cameras (10 evaluated) installed to 3 areas of a deprived public housing project in southeast England | Crime increased 4% in the target area, and decreased 19% in the control area, suggesting a statistically significant difference | About 9-10% fewer respondents reported being worried about crime after CCTV installation. Significantly better findings than in control area. |
Borough | 8 new cameras used in a redeployable system which could be attached to any lamp post across a mixed/affluent residential area of southeast England | Crime increased by 73% in the target area, a statistically significant difference from the more modest 12% increase in the control area | No information available |
Deploy Estate | 11 new redeployable cameras implemented to different areas of a deprived public housing project | A 21% increase in crime recorded in the housing estate, compared to only a 3% increase in the control area | A slight improvement in those worried about crime in one area of the project compared to the comparison area. No change in the other area. |
