Original Article
CCTV in jail housing: An evaluation of technology-enhanced supervision Sara A. Debus-Sherrilla,*, Nancy G. La Vigneb and P. Mitchell Downeyc a ICF International, 9300 Lee Highway, Fairfax, VA 22031, USA. E-mail: sara.debus-sherrill@icfi.com b Urban Institute, 2100 M St. NW, Washington DC 20037, USA. E-mail:
[email protected] c Department of Economics, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA.
*Corresponding author. This project was supported by Grant No. 2006-RP-BX-0040, awarded by the National Institute of Justice, Office of Justice Programs, US Department of Justice. The opinions, findings and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Department of Justice.
Abstract Surveillance cameras are widely used in correctional settings, but there has been little research on their effectiveness, especially in jail settings. Employing a mixed methods approach, this article examines how implementing closed-circuit television in jail housing units in a large Eastern city influences inmate perceptions of safety and incidents of violence and misconduct. Data collected through surveys with 101 inmates and 68 months (56 months from before and 12 from after camera implementation) of administrative records of inmate infractions, incidents of self-harm and officer use of force were analyzed through χ2 tests, independent sample t-tests, and structural break (time series) analyses. Semi-structured interviews with 14 correctional staff were analyzed qualitatively to provide contextual information. Findings indicate that while inmate perceptions of safety changed after implementing cameras, analyses of reported incidents did not yield any effect. These mixed results may be because of a combination of deterrence and detection effects or to cameras not being paired with more effective monitoring. Security Journal advance online publication, 28 July 2014; doi:10.1057/sj.2014.31
Introduction Public surveillance cameras, often referred to as closed-circuit television (CCTV), have become an increasingly popular public safety tool. Although cameras have been used for some time by private businesses, their use has quickly expanded in recent years within public spaces (Norris et al, 2004; Gill, 2006), schools (Taylor, 2010) and correctional settings (Travis et al, 1989). Advances in technology over the past few decades have made cameras even more attractive, with features such as recording, panning, tilting and zooming capabilities. In a society that increasingly turns to technology for innovation and solutions, CCTV is an enticing option for those wishing to reduce crimes.
© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal www.palgrave-journals.com/sj/
1–18
Debus-Sherrill et al
The use of CCTV is supported by the Situational Crime Prevention (SCP) framework. SCP is an approach for conceptualizing and addressing crime based on rational choice and opportunity theories. It posits that potential offenders who are motivated to commit crimes may be deterred after weighing the potential costs and benefits if they perceive the act (i) has too high a risk of being caught, (ii) involves too much effort, (iii) results in too little of a reward or (iv) will induce feelings of guilt or shame; in addition, reducing facilitators and provocations for offending may also prevent crime (Clarke, 1992/1997; Clarke and Homel, 1997; Wortley, 2002). The SCP framework can be applied to a diverse array of settings, including corrections (Wortley, 2002). According to the tenets of SCP, surveillance cameras would increase the risk of apprehension if potential offenders were aware of them (assisted by being highly visible or accompanied by signage) and could reduce the reward if the offender was caught and stopped midway through the act. Surveillance systems may also have a secondary impact, serving to increase others’ perceptions of safety and thus their presence in monitored spaces (Gill, 2006; Ratcliffe, 2006), which in turn may increase guardianship and reduce crime (Welsh and Farrington, 2002; Welsh and Farrington, 2004). Although CCTV is supported by theory, its rising use is not accompanied by strong evidence of its effectiveness, especially in the United States. Most public surveillance evaluations have been conducted in Europe, and particularly in the United Kingdom (UK), where use of the technology is more widespread. Two meta-analyses found mixed results for CCTV’s impacts on crime in public areas, with varying effects by location (studies of US cities revealed particularly disappointing results) or type of area (for example, residential versus downtown), resulting in little or no overall reduction in crime (Welsh and Farrington, 2002, 2004, 2008; Cameron et al, 2007). More recent studies in the United States (King et al, 2008; Ratcliffe and Taniguchi, 2008; La Vigne et al, 2011a) have also found mixed findings on CCTV in public places. Moreover, research on offender decision making has confirmed that offenders are not overly concerned with the presence of surveillance cameras when planning a criminal act (Gill and Loveday, 2003). However, recent research suggests that cameras can be effective, provided that cameras are actively monitored, rather than solely recording video footage of activities that can later be retrieved for investigative purposes (La Vigne et al, 2011a).
CCTV in corrections While public surveillance cameras have only marginal empirical support for their effectiveness, there is even less known about CCTV in correctional settings. Few studies have investigated this topic, and only two studies to date have examined the effectiveness of camera systems in correctional settings. A 1989 survey found that about half of prisons built during the previous decade used surveillance technology, with CCTV most commonly used in the entrance and visitation areas (Travis et al, 1989). A content analysis of legislation, prison audits and manager interviews in a small sample of Australian prisons found a variety of stated purposes for using CCTV in prisons (Allard et al, 2006). These included monitoring visitation, monitoring inmate behavior, monitoring inmates at high risk of suicide, providing evidence for incident investigations, improving safety, facilitating remote access control (for example, identifying who is requesting entrance to an area), viewing staff in personal distress or unsafe situations, providing perimeter security and coordinating movement of inmates. 2
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
Interestingly, cameras were not explicitly touted as a prevention tool in the same way it is often described in regards to public use areas. One case study of a British police station with holding cells examined the implementation of CCTV in individual custody cells and found that both officers and inmates supported the use of cameras (Newburn, 2002). While inmates disliked the presence of cameras in bathroom areas, they nonetheless perceived a greater benefit in protection from police abuse than the drawback of reduced privacy. Officers felt the primary benefits of the cameras were their ability to offer evidence against false allegations and to illustrate how busy they were if some procedure was missed. Other studies have also found staff and administrators to have positive attitudes toward CCTV (Travis et al, 1989; US Department of Justice, 1995). The first study to examine behavioral outcomes of CCTV in correctional environments was a survey of 220 inmates at a British prison for young offenders (see Allard et al, 2008). The study compared self-reported rates of victimization and offending between inmates living in units with CCTV in the shower areas and inmates living in units without cameras. It found that inmates in camera units reported lower rates of both victimization and offending, although there were no significant differences in feelings of safety between the two groups. Finally, another study (Allard et al, 2008) examined behavioral outcomes of surveillance cameras in Australian prisons. The researchers collected administrative data on 1116 incidents across four prisons during a 6-year period and compared the number of incidents occurring in camera-monitored areas with the number of incidents occurring in locations without surveillance. Overall, more incidents occurred in areas not observed by cameras (64 per cent) than in areas with camera surveillance (36 per cent). Furthermore, the study found that (i) non-violent offending and misbehavior were less likely to occur in areas with cameras than violent offending and (ii) incidents with weapons and multiple assailants (a possible indicator of more extensive planning) were less likely to occur in camera areas than incidents without weapons and with a single assailant. These results led the authors to conclude that cameras may not deter spontaneous behavior as well as other types of prison incidents. Although available research on CCTV use in correctional settings is limited, it nonetheless provides some preliminary evidence that cameras may be useful in deterring violence and other forms of inmate misconduct. However, given the small number of studies examining actual outcomes and the high variability of effects found in other CCTV research, it is important to conduct additional research on this topic to better understand how cameras impact inmate behavior. In addition, methodological features of past studies prevent confident interpretation. For example, the first outcome study relied solely on self-reported outcomes, which may be vulnerable to social desirability effects or fear of punishment for honest reporting. Allard et al also acknowledged the critical limitation that cameras were already implemented in the four studied prisons; consequently, they were only able to compare areas with cameras and without. However, these areas were dissimilar in terms of function and other location characteristics that, according to SCP, could potentially influence the likelihood of offending regardless of cameras. Furthermore, results from these studies may not be generalizable to American prisons or jails. The current study addresses some of these limitations by conducting an evaluation of CCTV in a correctional setting, using a mixed methods approach employing both administrative and self-report quantitative and qualitative data. Surveys examine inmate perceptions before and after camera implementation, while the quasi-experimental, longitudinal design employed to examine differences in © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
3
Debus-Sherrill et al
reported incidents allows for a more definitive interpretation of camera outcomes by analyzing actual behaviors over time. Furthermore, the study uses an American jail setting. Local jails, with their diverse inmate populations, rapid population turnover, connections with local communities and other distinctive characteristics are in need of their own jailspecific research. Specifically, this article explores the following research questions: 1. How do jail inmates perceive the value and impact of CCTV? 2. How do jail staff and administrators perceive the value and impact CCTV? 3. Does CCTV deployment in jail housing units impact inmate behavior, including physical violence, sexual violence and misconduct, and suicidal or self-harming behaviors?
Methodology The data employed for this analysis are part of a larger, National Institute of Justice-funded study conducted by the Urban Institute: the Jail Sexual Assault Prevention project.1 This action–research partnership study aided three county jail facilities located in three different US regions in identifying and addressing the factors driving violence in their facilities and evaluated the success of the selected interventions in reducing levels of violence. The project used a SCP theoretical framework and focused not only on sexual assault but on the interrelated issues of physical violence and suicide and self-harm. This article reports on the methodology and findings from one of the study sites that implemented a recording camera system within a portion of the jail’s housing units. The study consisted of three main phases. The first phase of the project entailed collecting and synthesizing data from multiple sources – site observations, administrative data, staff and inmate interviews, and other sources – to better understand violence in the facility and identify key contributing factors. Drawing from the results of these analyses, the second phase entailed the development of a series of recommendations for addressing the key issues. Researchers worked with the facility’s management to select an intervention that was promising, feasible and affordable: in this case, CCTV. In the final phase of the project, the intervention’s impact on violence and its cost effectiveness2 were evaluated through an analysis of administrative data, inmate surveys and staff interviews. Participating correctional facility The jail facility described in this article is located in a large Eastern US city and houses approximately 1100 maximum and medium security male inmates. The majority of inmates were African American (72 per cent) and were being held pretrial (80 per cent); a small number were juveniles being tried as adults. The facility practices direct supervision and unit management, an approach for dividing an institution into smaller, more manageable groups where inmates participate in most activities (for example, meals, recreation, medicine dispensing) within or near their own housing units. Each housing unit has an officer console, a central dayroom, an outside recreation yard, a shower area and two levels of doublebunked cells on tiers forming a Y-shape. While other areas of the jail had cameras, only one housing unit had cameras at the start of the study. 4
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
Selected intervention The research team combined the findings from all data collected during the first phase to identify key themes regarding violence in the facility and situational and environmental factors that were contributing to violence. Once the major factors contributing to violence were identified, the research team developed a set of site-specific recommendations for addressing these issues based on SCP, rational choice and other criminological theories. The process of selecting the most promising and practical interventions was a joint effort between the research team and jail administrators, in keeping with the action research design of the project. Out of the various recommendations made by the research team, the facility chose a recording camera system to eliminate blind spots, provide recorded evidence for incident investigations and improve staff conduct. In particular, cameras were needed to view blind spots at the rear of the bottom and top tiers of cells (the stem of the ‘Y’), areas identified by both staff and inmates as high risk for violence and suicide attempts. It was noted that the installation of cameras must be combined with increased staff vigilance, including prompt response to incidents that occur within view of the cameras. Jail administrators selected camera locations based on the findings from the first phase of the study, along with their own knowledge and expertise about the facility. With support from a US$25 000 subgrant to offset the costs of the intervention, the camera system was installed in September 2009 and later expanded in January and March of 2010. Ultimately, six maximum-security housing units known to have violence problems were equipped with either two or six recording cameras. Three units had two cameras overlooking the rear of the top and bottom tiers. The other three units had six cameras with the following view angles: (i) rear of top tier, (ii) rear of bottom tier, (iii) front of top tier, (iv) front of bottom tier, (v) officer console area and entrance to housing unit and (vi) dayroom. Even with six cameras, there were still a number of blind spots in the dayroom and other parts of the housing unit. Jail leadership preferred to keep a direct supervision focus within the facility and, therefore, officers in the auxiliary booth were not expected to monitor the cameras continuously (officers in the auxiliary booth also have additional responsibilities such as granting access to units). Instead, jail administrators emphasized the use of the cameras’ recording capabilities for viewing incidents after the fact. Policy mandated that video footage be reviewed for every incident occurring inside a housing unit, and reviews of investigation reports confirmed consistent use of this practice starting in Winter 2009. In November, jail leadership began a policy for shift commanders to conduct routine reviews of 1 hour of randomly selected video footage from their shift (initially weekly reviews and later monthly reviews). According to staff interviews and on-site observations, the facility experienced a more or less smooth implementation. However, the following challenges were encountered: (i) two of the cameras were initially placed in easy reaching distance, resulting in one incident of minor inmate vandalism of cameras before they were moved to a higher location, (ii) two cameras experienced intermittent functioning and (iii) cameras lacked visual clarity, which sometimes made it difficult to identify individuals because of blurriness. © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
5
Debus-Sherrill et al
Staff interviews The implementation process was documented through facility observations and regular communication (by phone and e-mail) with jail administrators. In addition, semi-structured interviews with jail management and staff, stratified by shift, (N = 14) occurred 1 year after the cameras were installed to obtain staff perceptions on the impacts of the intervention; all but 2 of the 14 staff had been working at the jail before camera deployment. Interview protocols covered topics such as perceived impacts of the intervention on inmates, staff and the facility; implementation challenges; satisfaction with the intervention; and areas for improvement. Urban Institute researchers incorporated information from site observations and staff interviews to document lessons learned from the implementation. Staff interviews were also qualitatively analyzed to identify staff perceptions of the impacts of the intervention. Inmate surveys Data on inmate perceptions of safety within the facilities were obtained through a survey instrument about physical violence, sexual victimization and self-harming behavior. The surveys were administered 16–19 months3 before (N = 110) and 13 months after (N = 101) the implementation of the new safety interventions. Within each general population housing unit where the camera system was used, inmates were randomly4 selected from a list provided by jail management that included all inmates 18 and older who had resided in the facility for at least 90 days. No attempt was made to survey the same individuals for both survey rounds; this would have been nearly impossible given the rate of jail population turnover. Surveys were administered in the jail’s gym in groups of 3–33 inmates. Respondents were asked about the likelihood of physical violence, suicide and self-harm, sexual assault and consensual sexual activity in the facility; the locations where these types of incidents typically occur; whether these incidents usually come to the attention of staff; inmate access to weapons and privacy, both of which can facilitate assaults; the incidence of gang involvement among inmates; and inmate access to mental health care. Respondents were also asked about their own experiences of physical violence and self-harming behaviors in the previous 30 days. The response format for these questions generally followed a varying5 4-point Likert scale with response options scored as (−3, −1, 1, 3) with more positive values indicating safer perceptions. Inmate survey data were analyzed with independent sample t-tests and χ2 tests of independence to determine whether inmate perceptions of safety changed after the intervention was implemented. Correlations were run to determine whether perceptions of safety varied by respondent characteristics; significant correlations were only reported when present in both the pre and post samples. Incident data Data on reported incidents were collected from the jail’s hard copy files for the period of 1 January 2005 (56 months before the project began) through 30 September 2010 (12 months after the interventions were implemented). Data included all instances for which an inmate or staff member suffered physical violence or injury, including all incidents of sexual assault, 6
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
suicide, self-harm and physical violence (for example, physical fights, attacks, assaults) that come to the attention of staff. The data also included all seizures of weapons and contraband and any use of physical force by correctional staff. Data on the content of the incident (for example, incident type, weapon usage, resulting harm) and circumstances surrounding the incident (for example, time, location, presence of witnesses) were recorded for each report. The following pre-established categories of incidents were collected for analysis: physical assaults, sexual incidents, self-harm, weapon/contraband seizure, insubordinate/threatening inmates, staff use of force and staff misconduct (see Table 1). These data were used to measure changes over time with the exception of sexual incidents, self-harm, incidents with insubordinate/threatening inmates and staff misconduct categories, which did not have sufficient sample sizes to reliably analyze with time series methods. While the incident type was coded from the primary, or most serious, incident of the event, the use of force and contraband analyses were completed using flags for every incident indicating whether these factors were present. Incident types include allegations, attempts and completions by either inmates or staff, unless specified otherwise.6 We employed structural break analyses to examine whether the incidence of these facility events was impacted by the implementation of CCTV. Structural break analysis detects when significant changes in incidents occur during the time period, and then these time points can subsequently be compared with known milestones. Such models can detect both (i) immediate shifts beginning the week after an intervention was implemented and continuing until the start of a new event; and (ii) time-variant intervention effects, which can change over time (for example, the initial deterrent effect of cameras degrades over time as inmates learn that blind spots exist).7 While largely atheoretical and less conservative, structural break is a well-documented econometric approach for evaluating programs with inexact implementation dates (Piehl et al, 2003). ARIMA time series analysis was also considered, but ARIMA does not lend itself well to multiple intervention points with limited time periods in-between. Owing to limitations including gradual implementations of the intervention, inexact dates for other facility events and changes, and the presence of overlapping events,8 structural break analysis was viewed as a more appropriate technique than ARIMA, which relies strongly on specific intervention and event dates. All models controlled for inmate-to-staff ratios. Monthly data were used for all incident types except for: all incidents, inmate assaults and use of force, which were counted weekly. Table 1: Analyzed incident types Incident type (1 January 2005–30 September 2010)
Number of incidents
All incidents Main incidents (including physical assaults, sexual incidents, suicide/self-harm) Physical assaults (combined assaults on inmates and staff) Inmate-on-inmate assaults Inmate-on-staff assaults Sexual assaults and other sexual incidents Suicide/self-harm Contraband seizure Insubordinate/threatening inmates Use of force by staff Staff misconduct
N = 1254 N = 664 N = 539 N = 353 N = 182 N = 33 N = 101 N = 231 N = 148 N = 521 N = 39
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
7
Debus-Sherrill et al
In addition, the following categories had enough incidents to examine whether patterns in incidents differed for intervention areas compared with other areas: All Incidents, Main Incidents and Physical Assaults. For these incident types, both the number of incidents in non-intervention housing areas and in non-housing areas were included as covariates. Incident patterns similar to the time series data for the intervention areas would have significant coefficients, while differing patterns would be non-significant. It would be expected that non-housing areas would have a different pattern of incidents since these include areas with very different functionalities; however, the non-intervention housing areas can serve as a quasi control area. The comparison is not a perfect one, as the intervention units were maximum-security units selected specifically for their identified violence; however, the analysis still provides useful information. The comparison units included two mental health units, administrative segregation, protective custody, two juvenile units and one regular medium-security unit.
Evaluation Findings Staff interviews Interviews with staff revealed that management and jail leadership found the intervention to be a useful tool for both investigating inmate incidents and for monitoring staff behavior. Respondents reported only a small number of staff incidents detected on camera within the study period; these were being addressed through training and discipline. At the end of the implementation period, jail administrators hoped to expand the system to cover more blind spots and additional housing units. Line officers, on the other hand, had more mixed opinions, with criticisms centering around the use of cameras to monitor their own behavior instead of inmate behavior. Many officers reported that the system did not impact officer behavior, expressing that officers did their jobs to the best of their abilities before and after the implementation and would not have time to think about cameras in critical situations. Respondents had mixed views on whether cameras deterred inmates, especially for those who were intent on fighting. They reported that inmates were skilled at discovering blind spots and could initiate violence in these areas. In fact, within a month of implementation, multiple incidents had occurred in blind spots, while only one incident had been recorded in the camera viewing area. Instead, line staff pointed to the usefulness of the cameras for investigations of inmate incidents. According to one respondent, inmates also appreciated this role of cameras and often requested that staff view the cameras to corroborate their story.
Inmate surveys Both the pre (N = 110) and post (N = 101) samples were similar in terms of demographics and background (see Table 2). The sample included only male inmates, and respondents were primarily Black/African American and had a mean age of 30–31 years. The post sample had a significantly smaller number of Black/African-American respondents (58 per cent) than those surveyed before the intervention (73 per cent). However, correlational analyses of responses by 8
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
Table 2: Sample characteristics Variable Mean age Percentage of female Percentage of Black* Percentage of Latino/Hispanic† Percentage of White/non-Hispanic Percentage of heterosexual Percentage of violent offense Percentage of property offense Percentage of drug offense Percentage of other offense Mean number of convictions Mean number of months served at jail† Percentage of maximum security Percentage of first time at current jail Percentage of people at another jail in past Status Sentenced (in percentage) Awaiting trial (in percentage) Other (in percentage)
Pre
Post
30.9 0.0 72.6 12.1 13.9 99.0 49.0 15.7 20.6 14.7 3.4 12.3 63.8 63.3 76.1
30.0 0.0 58.1 23.0 13.9 98.9 54.6 11.3 21.6 12.4 3.3 10.2 62.9 68.7 80.8
30.3 68.2 0.90
25.3 72.7 2.0
*P < 0.05; †P < 0.1
race did not yield consistence differences in survey responses, and the sample size was too small to support weighting to control for any pre-existing differences between the two groups. Similarly, fewer respondents reported being Latino or Hispanic in the post (23 per cent) sample compared with the pre (12 per cent) sample, but this difference was only marginally significant. Nearly all respondents (99 per cent) reported a heterosexual orientation. The largest proportion of respondents were charged with violent offenses (49–55 per cent), followed by drug offenses (21–22 per cent). Drug offenders were about evenly split between possession (48 per cent) and selling (52 per cent) offenses. Respondents had prior experience with the criminal justice system, with an average of three convictions in their past. The majority (76–81 per cent) had been at a different jail previously,9 and around onethird (31–37 per cent) had been confined at this same facility before. At the time of the survey, inmates had spent about 10–12 months at the jail, and the majority was classified as maximum security (63–64 per cent). While most respondents were awaiting trial (68–73 per cent), 25–30 per cent were already sentenced. Table 3 displays findings from the survey on various safety risks and related facilitators. The majority of sexual items have a positive mean value on the 4-point Likert scale (−3, −1, 1, 3) described earlier, indicating that inmates do not strongly perceive sexual incidents to be of high risk overall. However, the mean values on the Likert scale were negative in the pre sample for likelihood of sex with an officer (−0.15), likelihood of consensual sex occurring in cells (−0.43) and likelihood of staff finding out about both forced (−0.15) and consensual (−0.49) sex, showing some perceptions of higher risk for these particular items. After implementing cameras, there were significant increases in mean scores, indicating that far fewer respondents thought that sex with an officer, sex in exchange for something and sexual assault were likely. © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
9
Debus-Sherrill et al
Table 3: Perceptions of safety risks and related facilitators Variable Perceptions of sexual behaviors Likelihood of sexual assault** Likelihood of exchange sex** Likelihood of consensual sex Likelihood of sex with officer** Perceptions of sexual assault Likelihood in cell** Likelihood in dayroom Likelihood on tiers* Likelihood in showers Likelihood in rec area* Likelihood in closets** Likelihood staff find out Comfort of reporting assaulta Perceptions of consensual sex Likelihood in cell* Likelihood in dayroom Likelihood on tiers Likelihood in showers Likelihood in rec area Likelihood in closets Likelihood staff find out Perceptions of physical violence Likelihood of fight/attack Likelihood in cell** Likelihood in dayroom Likelihood on tiers Likelihood in showers Likelihood in rec area Likelihood in classroom/library Likelihood staff find out Number of inmates in gangs† Perceptions of self-harm and health care Ease of getting meds* Ease of getting mental health care Likelihood of inmate hurting self Likelihood staff find out Respondent experiences in past 30 days Tried to hurt self in past 30 days Threatened by inmate in past 30 days† Hurt by other inmate in past 30 days Fought in past 30 days† Perceptions of contraband and privacy Number of inmates w/weapons Ease of getting weapon Number of inmates using drugs Ease of getting drugs Ease of getting privacy†
Pre mean
Post mean
Pre (in percentage)
Post (in percentage)
1.31 0.98 0.98 −0.15
2.17 1.80 1.39 0.69
20.4 30.9 33.0 55.8
4.5 13.8 23.0 35.7
0.02 2.11 1.97 1.59 2.03 0.86 −0.15 N/A
0.86 2.33 2.36 1.81 2.43 1.58 −0.38 0.41
50.0 3.0 8.1 12.1 5.2 33.7 43.8 N/A
32.2 3.4 2.3 11.6 3.4 19.8 40.7 40.8
−0.43 1.88 1.96 1.51 2.00 0.76 −0.49
0.40 2.17 2.08 1.50 2.07 1.21 −0.44
60.6 12.2 10.2 18.4 8.2 39.8 34.0
42.9 2.4 4.8 17.9 4.8 28.6 34.6
−1.80 −0.85 −1.36 −1.27 0.88 −1.60 0.60 0.85 0.09
−1.63 0.06 −1.15 −0.91 0.79 −1.59 0.55 0.75 0.65
87.9 67.3 83.0 79.0 29.8 84.9 38.5 71.4 50.5
88.0 47.2 78.3 77.8 29.9 85.7 37.5 71.6 41.9
−0.29 0.22 −0.46 0.71
0.33 0.60 −0.05 1.09
43.9 63.8 67.0 69.9
57.0 67.0 62.5 77.9
N/A N/A N/A N/A
N/A N/A N/A N/A
0.9 24.3 13.3 20.0
4.3 12.2 12.4 9.8
−1.02 −0.96 −0.61 0.05 −0.28
−0.77 −0.48 −0.71 0.53 0.25
75.7 70.2 73.2 54.5 51.5
67.5 61.0 76.5 38.2 44.6
*P < 0.05, **P < 0.01, †P < 0.1 Notes: The item Number of Inmates in Gangs, Number of Inmates with Weapons and Number of Inmates Using Drugs has a scale of ‘Most Inmates’ to ‘None’. Percentages shown are for respondents endorsing ‘Some Inmates’ or ‘Most Inmates’. The items, Ease of Getting Meds, Ease of Getting Mental Healthcare, Ease of Getting Weapons, and Ease of Getting Drugs have a scale of ‘Very Easy’ to ‘Very Hard’. Percentages shown are for respondents endorsing ‘Easy’ or ‘Very Easy’. a The item Comfort of Reporting Assault was only administered for the post survey and has a scale of ‘Very Comfortable’ to ‘Very Uncomfortable’.
10
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
In contrast to perceptions about sexual violence, the majority of items about physical violence are negative at both waves, indicating that inmates feel at high risk for physical violence. Overall, the majority of inmates (88 per cent of pre and post inmates) reported fights or attacks were likely (‘Likely’ or ‘Very Likely’), and there was no significant change in the perceptions of risk for physical violence in general. There was also no significant change in perceived likelihood of self-harm, which was viewed as likely by the majority of respondents (63–67 per cent). Respondents were asked if they had any of the following experiences in the past 30 days: (i) hurt self, (ii) threatened by an inmate, (iii) hurt by an inmate or (iv) participated in a fight. Small percentages of respondents reported experiencing these four types of events (see Table 3); hurting oneself was particularly rare. The number of respondents who reported being threatened by another inmate and becoming involved in a fight decreased from the pre to post survey, although these differences were only marginally significant. A prime tenet of SCP theory is that specific contexts are taken into account when considering offending events. As this project used a SCP framework, it was important to measure not only the overall likelihood of these different safety risks, but also their likelihood in various situations. With regard to physical location, respondents reported that cells and closets were the most likely locations for sexual behavior and that the dayroom, tiers and cells were the most likely locations for physical violence. There were locationspecific decreases for perceived likelihood of sexual assaults in cells (also for consensual sex in cells), the tiers, the recreation area and closets. Significantly fewer respondents thought physical violence was likely in cells in the post sample, although there was very little change in perceptions of physical violence for other areas. The dayroom and cells were the most common locations for violence experienced in the past 30 days, with incidents also occurring in the recreation area, tiers, library or classrooms and closets. The dayshift (7:00–15:00) was the most common time of day for being hurt by another inmate, while the evening shift (15:00–23:00) was the most common time for fights; however, victimization and fights occurred during every shift. Slightly more than half of injured respondents (55–64 per cent) reported the use of weapons when an inmate hurt them. Disrespect was the most common cause of fights, with telephones, street issues, stealing, gangs, television, gambling, sexual orientation, defense of other inmates, contraband, frustration, newspapers, owed money, racial/ethnic division and sexual ‘favors’ as other reported causes. Health care, weapons, drugs and access to private locations can influence violence and self-harm indirectly. Respondents had mixed perceptions of the jail’s health care with split numbers of respondents viewing acquiring medications (44–57 per cent) or mental health care (64–67 per cent) as easy (‘Easy’ or ‘Very Easy’). A large number of inmates reported they personally had experienced difficulty obtaining medications or mental health care in the jail (56 per cent of pre and 42 per cent of post respondents who reported needing health care). Respondents reporting personal challenges with the jail’s health care were more likely to think that obtaining medications and mental health care were difficult. However, overall, there was a significant improvement in perceptions of the ease of accessing medication. This did not appear to affect the perceived likelihood of self-harm, though. The majority of respondents reported inmates had access to weapons and drugs. There were shifts in perceptions of the accessibility of contraband and privacy, although only the change for privacy approached significance. Although there was a substantial decrease in the © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
11
Debus-Sherrill et al
number of respondents who thought it was easy to get drugs in the post sample (38 per cent compared with 55 per cent), this change is not significant because of a smaller sample size.10 Inmates had inconsistent views regarding the presence of gangs in the jail with about half (51 per cent) of inmates in the pre sample believing that ‘Some’ or ‘Most’ inmates were in gangs. These perceptions only experienced a marginally significant decline for post respondents (42 per cent). Perceptions of the likelihood of staff finding out about incidents varied by incident type. While the majority of inmates believed staff would learn of self-harm (70–78 per cent) or physical violence (71–72 per cent), fewer respondents thought staff would discover consensual (34–35 per cent) or forced sex (41–44 per cent). Less than half (41 per cent) of post11 respondents said they would be comfortable reporting a sexual assault to correctional staff. The most common reasons for inmates’ discomfort in reporting (N = 31) were that they distrusted officers (officers would tell other inmates, were disrespectful/unprofessional or would not care) (34 per cent), they felt that sexual assault is a personal matter or not other people’s business (25 per cent), they were ashamed (16 per cent) or they feared retaliation (9 per cent). In an attempt to further investigate the role of cameras in changed perceptions, two-way ANOVAs were run to examine the interaction between survey administration and whether the respondent’s housing unit had cameras installed. However, because of sampling difficulties regarding lockdowns and inmate schedules, the research team was only able to survey one comparison unit (N = 47) without cameras across the two survey groups compared with five units (N = 136) with cameras across the two survey groups. In addition to this limitation, the single comparison unit without cameras was substantially different from the five units with cameras. While all five of the units with cameras were maximumsecurity housing units with known violence problems, the comparison unit held mediumsecurity inmates with few documented problems. Nonetheless, these comparisons were explored to determine if they offered any additional insight to the survey findings. Only two significant interaction effects were found: inmates in units with cameras felt it was more difficult to find privacy after the cameras were installed, while inmates in the comparison unit without cameras thought it was easier to find privacy at the second survey administration. There was also a significant interaction for the likelihood of consensual sex, with both types of units reporting reduced likelihood from the pre to post administrations, but the comparison unit without cameras displaying a larger change. While these findings may shed further light on whether the cameras played a strong role, they should be interpreted with caution because of the substantial limitations of the available comparison group. In addition to replicating questions on the pre survey, we included items (following a 4-point Likert scale with response options ranging from ‘Strongly Disagree’ to ‘Strongly Agree’) on the post survey to learn about inmate perceptions of the cameras (see Table 4), as well as questions asking about the specific cameras in their housing unit. Findings revealed that many respondents did not have a clear understanding of the camera system. Over onethird (37 per cent) of inmates reported an incorrect (some, vastly incorrect) number of cameras in their housing unit.12 Less than half (42 per cent) knew the cameras recorded, with 6 per cent believing the cameras did not record and a little over half (52 per cent) reporting they did not know whether or not the cameras recorded. Although less than half (48 per cent) of inmates stated that cameras make the jail more safe, the majority of respondents believed the cameras contributed to increased safety, including violence being less likely with cameras (74 per cent), cameras making 12
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
Table 4: Perceptions of intervention Variable a
Cameras make jail more safe Violence less or much less likely to occur with camerasb Cameras make investigations faira Cameras help verify inmate concerns about staff
Post mean
Post (in percentage)
−0.15 0.85 0.55 0.41
48.2 74.4 65.5 66.7
The items Cameras Make Jail More Safe and Cameras Make Investigations Fair have a scale of ‘Strongly Disagree’ to ‘Strongly Agree’. Percentages shown are for respondents endorsing ‘Strongly Agree’ or ‘Agree’. a
b The item Violence Less or Much Less Likely to Occur with Cameras has a scale of ‘Much More Likely to Occur’ to ‘Much Less Likely to Occur’. Percentages shown are for respondents endorsing ‘Much Less Likely’ to ‘Less Likely’.
investigations fair (66 per cent) and cameras helping to verify inmate concerns about staff (67 per cent). Those who had spent time in other jails were less likely to believe that the cameras improved safety, and Black respondents were less likely to believe that cameras lead to fairness in investigations.
Incident data analysis Across the nearly 6-year period, there were 1254 incidents reported at the jail with the following distribution for the primary incident associated with each event:13 43 per cent physical assaults, 18 per cent contraband, 12 per cent insubordination/threats, 8 per cent selfharm and 3 per cent sexual incidents (see Table 1). Staff force was used in 42 per cent of incidents, and 19 per cent of contraband incidents included weapons. The most common types of locations for main incidents (including sexual incidents, physical assaults and selfharm) were cells (40 per cent), the dayroom (25 per cent) and tiers (11 per cent). Most incidents occurred in housing units, particularly one of the maximum-security units and both the maximum- and medium-security mental health units. Incidents most often occurred during the evening shift, particularly during the hours of 16:00–18:00. Using structural break analyses that controlled for inmate-to-staff ratios, no significant breaks were found for any of the incident types examined (see Figure 1). Comparisons with non-intervention housing units and non-housing areas also did not support an effect of cameras. Inmate-to-staff ratio, surprisingly, also did not appear to have any significant impact on the number of incidents, even though there was a large decrease in the number of inmates toward the end of the study period. Overall, there was no evidence that the camera system had a direct impact on the number of incidents reported by staff, once controlling for the inmate-to-staff ratio.
Discussion Revisiting our original research questions, we find that the effects of CCTV among both inmates and staff are somewhat mixed. Results from inmate surveys indicate that fewer inmate respondents believed consensual and forced sexual behaviors were likely to occur © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
13
Debus-Sherrill et al
Figure 1:
Inmate-to-staff ratio (Graph 1) and adjusted residuals for all incidents (Graph 2)
after camera implementation. Violence was also perceived as less likely to occur in cells, and a smaller percentage of respondents reported being threatened or involved in fights in the past month. In addition, post respondent inmates thought it was easier to access medications, which could also reduce inmate violence. These changes in perceptions are promising; however, it is surprising that they would occur when some areas viewed by the cameras (such as the dayroom) did not experience any changes in perceptions of violence. Cameras would be expected to have a more holistic impact on perceptions of safety, particularly with regard to physical violence, the ability of staff to detect incidents and the presence of violence in the specific locations of the dayroom and tiers where these cameras directly overlook. Therefore, it is unclear whether these improvements are because of the camera intervention or other reasons. For instance, the inmate-to-staff ratio declined throughout the study, which may have improved inmates’ feelings of safety. Inmates also had mixed views on the degree to which cameras increased safety overall, although the majority of respondents believed the cameras reduced violence. In addition, the majority believed cameras had procedural benefits, such as supporting fair investigations and verifying inmate complaints about staff misconduct. While inmates’ views of the camera system were somewhat positive, staff and administrator perceptions of CCTV were mixed, mostly by rank. Staff at the management and leadership levels were more likely to see value in the cameras, whereas line-level staff expressed concerns about their purpose in monitoring officer activity. Finally, analyses of actual incidents helped to answer the question of how CCTV impacts inmate behavior, although conclusions were not definitive. No significant changes in incidents followed the implementation of the camera system – a finding that could be because of displacement of violence beyond the view of cameras, masked impacts from a combination effect of greater deterrence (decreasing actual incidents) and detection (increasing identification and reporting of already existing incidents), and/or limitations in data and analyses. Indeed, the structural break analyses employed are ill-suited to detect short-term changes occurring in less than 7 months. However, it is questionable whether correctional leaders would be interested in investing in expensive14 camera systems if the 14
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
potential benefits were so short-lived. Given that recent research of public surveillance cameras suggest active monitoring is critical to the technology’s effectiveness (La Vigne et al, 2011a), the absence of stronger effects may also be because of the lack of integrating these interventions with additional staff training, an accompanying inmate behavior management strategy or stronger monitoring. To be sure, cameras have been found to be effective in other studies (for example, Allard et al, 2008; La Vigne et al, 2011a), and it may be that the manner in which these cameras were deployed and used limited their effectiveness. Although the jail is credited with implementing mandatory video review for every incident and monthly random reviews of video footage, there was little direct monitoring of cameras and random reviews only covered a relatively small proportion of time. While not necessarily reflecting actual risk, inmate perceptions are important indicators of fears and concerns about safety and could be related to future violence if inmates feel the need to preemptively attack other inmates in order to prove their ‘toughness’ or prevent feared assault by others (Dumond, 1992; Kunselman et al, 2002). Most inmates thought physical violence was likely to occur, particularly in the recreation area, in the dayroom, on the tiers and in cells. However, only small numbers reported actual victimization or fights in the past 30 days. Responses also displayed a large gap between self-reported incidents of self-harming behavior in the past 30 days and perceptions of the likelihood that inmates would hurt themselves. These discrepancies may illustrate how fears of correctional violence out measure actual rates, or may be due to the limited reporting period (30 days) or to dishonest reporting. Overall, while perceptions of safety improved in some areas, there still appear to be widespread perceptions among inmates that violence, particularly physical violence, is likely to occur at the jail. The study’s findings have important implications for practice. The results do not bolster confidence in the ability of cameras to deter violence and other types of risky behaviors in correctional settings. However, cameras may still provide important benefits in terms of better evidence for incident investigations and making inmates feel safer. The study also illustrates how SCP may be used to guide intervention decisions and better understand the contextual factors and opportunity structures surrounding violence within correctional settings. Finally, certain limitations should be noted. First, the evaluation design relies on analyses of changes in incidents and inmate survey responses across time. Since the researchers were unable to use a randomized experimental design, it is not possible to determine definitively whether any changes are because of the implemented intervention or to other changes at the facilities. Attempts at including similar control groups were made, but differences in housing populations make such interpretations weak. Second, statistical power is limited for incidents analyzed monthly as opposed to weekly, as this reduced the sample size of time units. In addition, the facility’s environment was far from stable, and some of these other changes took place over extended periods of time, making it difficult to pinpoint when the impact of these changes would be experienced by the inmate population (for example, changes to locks in housing units made over a period of several months). The lack of a stable baseline made detection of intervention effects more challenging. As mentioned, the structural break analyses controlled for the notable change in inmate-to-staff ratios but were unlikely to detect short-term impacts that occur over less than 7 months. There are also limitations associated with the inmate surveys. Although attempts were made to select participants randomly, the research team experienced challenges with random selection and passive refusals, which may have resulted in some selection bias. Inmate © 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
15
Debus-Sherrill et al
surveys rely on self-reported perception data as a measurement of change in the jail. These findings should not be interpreted as direct measures of the true prevalence of incidents or behavior and instead should be examined in the context of other evidence. Moreover, the difference in composition of inmates in the pre- versus post-camera deployment samples with regard to race and ethnicity may have some bearing on survey results. For example, Blacks, who were under-represented in the post-deployment sample, may have different perceptions and experiences with victimization and perpetration than their White counterparts that were erroneously attributed to camera implementation. Unfortunately, the relatively small sample size precluded the use of weighting to create more balanced preand post-implementation samples. In addition, the inmate pre-intervention surveys were administered many months (16–19 months) before the intervention because of unexpected and extensive delays in the procurement process for the interventions, allowing for more time for other changes to occur in the interim. Owing to these limitations and the inherent difficulty in measuring impacts for detection systems that can simultaneously decrease incidents while increasing detection, it is difficult to know with confidence whether the findings offer a negative judgment on correctional cameras or just an inconclusive one. As correctional environments often have substantial changes across time in policies, populations, staffing and equipment/infrastructure, future studies should use a randomized experimental study if possible to provide the best opportunity for strong evidence regarding camera impacts. Future work should also examine the contextual factors surrounding correctional violence to better understand the facilitators and barriers for certain risks, as well as consider how inmate perceptions might be affected in different ways than actual incidents. With the correctional field’s current reliance on camera technology (in the absence of a consensus of scientific evidence), it is critical to continue conducting research to develop a better understanding of the cameras’ impacts in jails and prisons.
Acknowledgements The Jail Sexual Assault Prevention Project was sponsored by the National Institute of Justice (NIJ), US Department of Justice. The authors wish to thank the thoughtful peer reviewers whose comments helped to improve the study’s dissemination, and the many Urban Institute staff who have shared their expertise, reviewed earlier report versions and otherwise worked on the project team, including Diana Brazzell, Lisa Brooks, Megan Denver, Brian Elderbroom, Robin Halberstadt, Paula Heschmeyer, Shalyn Johnson, KiDeuk Kim, Akiva Liberman, Samantha Lowry, Dwight Pope, Darakshan Raja, Tracey Shollenberger, Bogdan Tereshchenko, Brian Wade, Charlie Zamiskie and Janine Zweig. We are particularly indebted to our expert consultant on this project, Laura Maiello of Ricci Greene Associates, for providing her expertise in jail design and jail management issues and for reviewing earlier report versions. Finally, we would like to thank the project’s supportive grant monitors, Marilyn Moses and Andrew Goldberg, at the National Institute of Justice. The late Andrew Goldberg was extremely dedicated to building knowledge on the topic of sexual assault in correctional facilities; his commitment to his work and to this topic has had a significant and lasting impact on the field. 16
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
CCTV in jail housing
Notes 1 The Prison Rape Elimination Act (PREA), passed in 2003, was designed to improve practices associated with the prevention of sexual assault in correctional settings. This included the provision of funds for research on this topic (see the National PREA Resource Center http://www.prearesourcecenter.org/about/prison-rapeelimination-act-prea). 2 Please see full report, La Vigne et al (2011b), for details on cost-effectiveness analysis. 3 The procurement process was much lengthier than anticipated, causing significant delays in the implementation. Owing to these lengthy delays, the inmate surveys were administered far in advance of actual implementation. 4 Exceptions to randomized selection occurred in two units during the pre-intervention surveys. Inmates were only selected from half of all possible inmates in two units, because only half of the list was screened for eligibility by the jail. This resulted in inmates only being selected from the lower tier of cells within these particular units. Furthermore, there were entire units that were unavailable for surveying because of being locked down for violent incidents. These restrictions prevented the researchers from surveying from all eligible housing units. There were additional challenges with passive refusals where substantial numbers of selected inmates refused to come to the gym to hear about the study (where all inmates were offered the option to refuse participation). Therefore, the sample may be vulnerable self-selection bias. 5 For example, one question might ask about the likelihood of an attack occurring (Very Likely to Very Unlikely), while another question might ask about the number of inmates who are in gangs (Most Inmates to None). Although the response options differ across the different types of questions, all 4-point Likert scale items were coded in the same way: (−3), (−1), (1), (3). 6 Self-harm events include both self-harm and suicide, as well as hunger strikes and injuries from suspected selfharm. Contraband was only coded if officers were able to seize it (that is, an incident would not be coded as contraband if an assault occurred with a weapon, but the weapon was not recovered). Insubordinate or threatening inmates includes inmate threats toward others (not threats of self-harm), miscellaneous discipline or insubordination and intentional flooding. Includes inmate threats (physical or sexual, but not self-harm), miscellaneous discipline or insubordination and intentional flooding. 7 It is important to note structural break analyses require the analyst to define the minimum length of a ‘regime’ or potential period of time related to hypothetical events. This is often set to 15 per cent of the total series length (approximately 11 months in this case). However, because of the rapid and frequent changes occurring in jail environments, a 10 per cent regime length was used instead (approximately 7 months). Changes that occur for a shorter period of time than this would be unlikely to be detected. 8 Other changes unrelated to the study were occurring in the jail during the study period, including policy changes disallowing suicidal inmates to be left alone, providing automatic mental health referrals for suicidal inmates, locking cell doors during the day, program and population changes to one of the intervention units, among others. The jail administrators were unable to confirm start dates for many of these changes. 9 The larger number of respondents reporting time served time at another jail may be because the study jail is part of a large complex of jails, one of which is a main intake facility, and inmates may count that as a stay in another jail. 10 Three drug items were included in 43 per cent of the surveys in an attempt to create multiple versions of the survey instrument to dissuade inmates from trying to view other inmates’ surveys. Because of this, there is a smaller sample size (N = 90 across both pre and post samples) and less power for these comparisons. 11 This question was only asked at the post wave. 12 There was a high non-response rate for this particular item (N = 65 of 101 total post respondents), likely also because of inmates’ unfamiliarity with the camera system. 13 Incidents were coded by the most serious type of incident occurring for each event. For instance, if there was an event where an inmate attacked another inmate, then attacked a staff member, a knife was recovered, pepper spray was used, and the inmate threatened the nurse who was treating his injuries, this would be coded as an assault on staff. (For coding purposes, assaults on staff were considered more severe than assaults on inmates, as there was an extra security risk involved with staff victimization.) 14 Actual expenditure costs of implementing the system within this jail were $54 740, although the cumulative cost, including economic and opportunity costs, was $85 000 over the 1-year period (with an expected average monthly cost of $2193 in marginal labor costs moving forward). Please see full technical report for more details about the cost-effectiveness analysis.
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18
17
Debus-Sherrill et al
References Allard, T., Wortley, R. and Stewart, A. (2006) The purposes of CCTV in prison. Security Journal 19(1): 58–70. Allard, T., Wortley, R.K. and Stewart, A.L. (2008) The effect of CCTV on prisoner misbehavior. The Prison Journal 88(3): 404–422. Cameron, A., Kolodinski, E., May, H. and Williams, N. (2007) Measuring the Effects of Video Surveillance on Crime in Los Angeles. CRB-08-007. Sacramento, CA: California Research Bureau. Clarke, R.V. (1992/1997) Situational Crime Prevention: Successful Case Studies. Reprint New York: Harrow and Heston. Clarke, R.V. and Homel, R. (1997) A revised classification of situational crime prevention techniques. In: S.P. Lab (ed.) Crime Prevention at a Crossroads. Cincinnati, OH: Anderson. Dumond, R.W. (1992) The sexual assault of male inmates in incarcerated settings. International Journal of the Sociology of Law 20(2): 135–157. Gill, M. (ed.) (2006) CCTV: Is it effective? In: The Handbook of Security. New York: Palgrave Macmillan, pp. 438–461. Gill, M. and Loveday, K. (2003) What do offenders think about CCTV? Crime Prevention and Community Safety: An International Journal 5(3): 17–25. King, J., Mulligan, D.K. and Raphael, S. (2008) CITRIS Report: The San Francisco Community Safety Camera Program. Berkeley, CA: University of California Center for Information Technology Research in the Interest of Society. Kunselman, J., Tewksbury, R., Dumond, R.W. and Dumond, D.A. (2002) Nonconsensual sexual behavior. In: C. Henley (ed.) Prison Sex: Practice and Policy. Boulder, CO: Lynne Rienner Publishers, pp. 27–47. La Vigne, N.G., Lowry, S.S., Markman, J. and Dwyer, A. (2011a) Evaluating the Use of Public Surveillance Cameras for Crime Control and Prevention. (Technical Report) Washington DC: The Urban Institute. La Vigne, N.G., Debus-Sherrill, S., Brazzell, D. and Downey, P.M. (2011b) Evaluation of a Situational Crime Prevention Approach in three Jails: The Jail Sexual Assault Prevention Project. (Technical Report) Washington DC: The Urban Institute. Newburn, T. (2002) Introduction of CCTV into a custody suite: Some reflections on risk, surveillance and policing. In: A. Crawford (ed.) Crime and Insecurity: The Governance of Safety in Europe. Portland, OR: Willan Publishing. Norris, C., McCahill, M. and Wood, D. (2004) The growth of CCTV: A global perspective on the international diffusion of video surveillance in publicly accessible space. Surveillance and Society 2(2/3): 110–135. Piehl, A.M., Cooper, S.J., Braga, A.A. and Kennedy, D.M. (2003) Testing for structural breaks in the evaluation of programs. Review of Economics and Statistics 85(3): 550–558. Ratcliffe, J. (2006) Video Surveillance of Public Places. Washington DC: US Department of Justice, Office of Community Oriented Policing Services. Ratcliffe, J. and Taniguchi, T. (2008) CCTV camera evaluation: The crime reduction effect of public CCTV cameras in the city of Philadelphia. PA Installed during 2006. Philadelphia, PA: Temple University. Taylor, E. (2010) I spy with my little eye: The use of CCTV in schools and the impact on privacy. The Sociological Review 58(3): 381–405. Travis, L.G., Latessa, E.J. and Oldendick, R.W. (1989) The utilization of technology in correctional institutions. Federal Probation 53: 35–40. US Department of Justice. (1995) Technology Issues in Corrections Agencies: Results of a 1995 Survey. Longmont, CO: National Institute of Corrections. Welsh, B.C. and Farrington, D.P. (2002) Crime Prevention Effects of Closed Circuit Television: A Systematic Review. Home Office Research Study 252. London: Home Office, Research, Development and Statistics Directorate. Welsh, B.C. and Farrington, D.P. (2004) Surveillance for crime prevention in public space: Results and policy choices in Britain and America. Criminology and Public Policy 3(3): 497–526. Welsh, B.C. and Farrington, D.P. (2008) Effects of closed circuit television surveillance on crime. Campbell Systematic Reviews 17: 2–73. Wortley, R. (2002) Situational Prison Control: Crime Prevention in Correctional Institutions. New York: Cambridge University Press.
18
© 2014 Macmillan Publishers Ltd. 0955-1662
Security Journal
1–18