Hot Spot Policing With Actively Monitored CCTV Cameras

Introduction

The effect of closed circuit television (CCTV) cameras as a deterrent for violent crime is likely small or nonexistent (Welsh & Farrington, 2009). CCTV cameras as a tool to improve policing of hot spots with violent crime may however have a significant effect on reducing violence (Piza, Kaplan, Kennedy, & Gilchrist, 2015). Hot spot policing can broadly be understood as a tactic based in placing extra police resources on places with high crime rates (Braga & Weisburd, 2010; Telep & Weisburd, 2012). A number of studies have convincingly shown that hot spot policing can reduce violent crime (Koper, 1995; Rosenfeld, Deckard, & Blackburn, 2014; Sherman & Weisburd, 1995; see Braga, Papachristos, & Hureau, 2012 for meta-analysis), but relatively little research on the topic in a Swedish context exists with the exception of evaluations by the National Council for Crime Prevention (Brottsförebyggande rådet [BRÅ], 2014a, 2014b; Frogner, Andershed, Lundberg, & Johansson, 2013). On the combination of CCTV cameras and hot spot policing only one study exists in Sweden that finds no significant effect on violent crime, while two studies from the United States did find crime-reducing effects (BRÅ, 2014b; La Vigne, Lowry, Markman, & Dwyer, 2011; Piza et al., 2015). La Vigne, Lowry, Markman, and Dwyer (2011) found that CCTV cameras were considered meaningful by stakeholders such as police and that actively monitored cameras can reduce crime. Piza, Caplan, Kennedy, and Gilchrist (2015) found that actively monitored CCTV with directed patrols resulted in large reductions in violent crime and social disorder. This type of actively monitored CCTV differs from other CCTV operations through its real-time communication between an operator monitoring the CCTV and police officers who can respond swiftly to identified problems or situations.

The aim of this study is to assess whether implementation of CCTV to assist police patrols in the prevention of crime at a violent crime hot spot leads to fewer assaults, using a Scandinavian middle-sized town, Malmö, as case study. Malmö is an interesting case due to its fairly small size, about 300,000 residents, which may lead to differing results than those found in larger cities such as the Swedish capitol Stockholm (BRÅ, 2014b).

In the town of Malmö, the police have employed hot spot policing methods in parts of the town center entertainment district. Directed patrol combined with police-operated CCTV at weekend nights have been implemented and reported to be successful. The police have argued that the CCTV component is a key aspect of these hot spot policing efforts, and in this article, an attempt at evaluating such an effect will be made. The evaluation will be performed post facto by comparing the treatment hot spot areas with other nightlife areas and, in addition, by comparing weekend nights when the cameras are active (between 12.00 and 06.00 a.m.) with weekday nights when they are not. The aim of the present study thus is to evaluate whether the police-operated CCTV cameras to assist directed patrols are associated with a lower number of public environment assaults. Public environment assault is chosen for study, since directed police patrols should have a stronger effect in public rather than in private environments. This is an important topic to further our knowledge of efficient police work and more broadly to give a better understanding of the effect of actively monitored CCTV cameras as a tool for crime prevention. In Sweden, CCTV actively monitored by the police to reduce crime is a very rare phenomenon, and with the results from Stockholm showing nonsignificant effects of crime (BRÅ, 2014b), it is of particular interest to test whether this holds true for the town of Malmö as well.

Theoretical Background

It has been shown that a small number of places have a large share of the crimes in a city, and such places have been labeled hot spots of crime (Braga & Weisburd, 2010; Sherman, Gartin, & Buerger, 1989). The concept of hot spot can be broken into its two core components of “hot,” indicating a high level of crime, and “spot,” reflecting the geographic concentration of a phenomenon in a particular place. In a theoretical discussion on hot spots by Sherman, Gartin, and Buerger (1989), place was defined as a space that could be seen from a central point (see also Sherman & Weisburd, 1995). A hot spot would then be a space that can be seen from a central point and that exhibits high levels of crime. In the early days of hot spot policing research, a hot spot could be defined as one or more addresses within half a block registered for multiple crimes (Sherman & Weisburd, 1988, cited in Sherman et al., 1989). Nowadays, a more common definition for a hot spot is a street segment, defined as “both sides of the street between two intersections” (Weisburd, Groff, & Yang, 2012, p. 23), with high concentrations of crime over time (Braga & Weisburd, 2010). A number of other geographical definitions have however also been employed, both smaller (e.g., Sherman et al., 1989) and bigger (e.g., Weisburd & Green, 1995). Theoretically, it has been argued that the smallest type units of analysis are more appropriate (Braga et al., 2012; Eck, 2002; Weisburd et al., 2012), but methodologically this raises issues, for example, in cases where police crime data are used with dubious spatial quality and reliability (Weisburd et al., 2012).

The basic premises of hot spot policing are that police can prevent and reduce crime in these spaces (Braga et al., 2012). In early studies of hot spot policing, emphasis was placed on police presence at the hot spot, with little concern for the potential of specific activities to be undertaken by the police to reduce crime (Sherman & Weisburd, 1995). An important study found that the optimal time spent on a hot spot was 15 min, after which the preventive effect of policing was diminished (Koper, 1995). Lately however, a larger emphasis has been placed on hot spot policing tactics and the activities of the police, sometimes in relation to problem-oriented policing (Braga & Weisburd, 2010; Telep & Weisburd, 2012). Braga and Bond (2008) noted that hot spot policing with a situational crime prevention component, which attempts to reduce opportunities for crime, had the strongest effect. They studied the effect of policing disorder, such as drunken people, abandoned buildings, or littering, to reduce crime and concluded that a problem-oriented approach in cooperation with the local community would be the most appropriate strategy. Two experiments in Jersey City found that a problem-oriented approach to crime hot spots could reduce crime (Braga et al., 1999; Weisburd & Green, 1995), but it has been noted that “aggressive disorder-enforcement tactics often were the general strategy” (Braga & Weisburd, 2010, p. 113). A similar approach was considered in Kansas City where traffic stops and searches were among the tactics employed to reduce gun crime (Sherman & Rogan, 1995). Although police presence and/or enforcement often appears to be the main tactic used, softer approaches have also been considered. Eck and Wartell (1999) found that police cooperation with land managers could help prevent drug sales on rental properties. As will be discussed below, the integration of CCTV surveillance and hot spot policing has recently received scholarly interest, although the topic is largely underresearched in Sweden.

The Case Study

Malmö is the third biggest town in the country with 307,758 residents (Malmö stad, 2014). In a Swedish context, it has a high crime rate, with on average 20,834 reported crimes per 100,000 citizens and year which is 40% higher than the national rate (BRÅ, 2012). Violent crime is clustered to the town center, and in the 3-year period of 2011–2013, a total of 273 public environment assaults were reported in the 200-m buffer treatment area of the study.

In 2011, the police in Malmö started working with a project on microplaces and crime, focusing on hot spots of violent crime. In total, 18 microplaces with elevated levels of violent crime were identified at highly localized areas or hot spots to receive additional police presence. The dosage of extra policing achieved in that effort is however unknown. Of those 18 microplaces, 6 were clustered on and around the square Stortorget, where several bars and nightclubs are located (Länsstyrelsen Skåne Län, 2012). In an effort to reduce violent crime in the Stortorget area, a more specific hot spot policing effort with directed patrols was implemented, focusing on those 6 microplaces rather than the full 18 originally identified. In practice, this means constant presence of at least four policemen in the vicinity on weekend nights. According to the police, there have only been a few occasions when the target of four police officers present has not been met due to extraordinary events (Deputy Head of Police in Malmö Karlsson, personal communication, December 3, 2014). In order to strengthen the policing effort around Stortorget, the police on August 17, 2012, decided to implement CCTV cameras on and around Stortorget which were monitored by police officers. The CCTV was to be operated only between midnight and 6 a.m. on Saturday and Sunday when the levels of violent crime were considered to be at their worst. The aim of the CCTV was explicitly connected to improving the policing effort in the area, “in real time discover disturbances and to act on such information by distributing resources [police officers] to prevent crime or to stop crimes in action” (Länsstyrelsen Skåne Län, 2012, p. 4). The Stortorget square thus on weekend nights has four police officers on the ground and one police officer actively, and exclusively, monitoring the CCTV cameras at the square. When the monitor notices a situation that may escalate into a crime, the police officers on the ground are notified directly. An important focus is thus on de-escalation of situations before they erupt into violence, or to stop ongoing violence from escalating further. In essence, the goal is to reduce provocation rather than to deter crime from a situational crime prevention perspective (Cornish & Clarke, 2003). This sets the use of CCTV cameras in this instance apart from CCTV cameras used to prevent crime through deterrence and makes the connection between CCTV and hot spot policing efforts explicit.

Research Design

Data

In the present study, data documenting reported assaults, involving the use of physical violence against another person, in public environments from the Malmö police is used as outcome variable. The choice to only include public environment incidents was made since the effect of directed police patrol should be bigger in public rather than in private environments. As noted by the police, it is difficult to precisely pinpoint where a crime has been committed in open public environments such as squares, since crimes are registered on the closest address (Länsstyrelsen Skåne Län, 2012). Buffers have therefore been implemented to reduce the risk of missing relevant data, both in respect to where the hot spot policing effort actually will take place and in relation to the spatial imprecision prevalent in the recording of crime. It should be noted that the inclusion of buffer areas, while reducing the risk of unreliable geographical crime data, also reduces the validity of the evaluation, since parts of nonexperiment areas are included in the analysis. Since the vast majority of crimes have been registered just outside the treatment area, the benefits have been considered to outweigh the risks, but it is important to consider the effect of such a choice in interpreting the results of this article. The data include both common and more severe (aggravated) assaults. Crimes that are registered outside the municipal borders have been excluded from the analysis as have crimes with no registered start time. Descriptive data for the areas used are reported in Table 1.

OPEN IN VIEWER

Table 1.

Descriptive Data.

Area	Number of Nightclubs (2013)	Assault (2011)	Assault, 2012 (% Change)	Assault, 2013 (% Change)
Treatment	5	6	4 (−33)	2 (−50)
Treatment, 50-m buffer	9	93	52 (−44)	59 (+13)
Treatment, 100-m buffer	17	99	54 (−45)	66 (+22)
Treatment, 200-m buffer	24	126	67 (−47)	80 (+19)
Control	9	58	38 (−34)	42 (+11)
Control, 100-m buffer	11	96	72 (−25)	65 (−10)
Rest of the town	52	1222	901 (−26)	964 (+7)

Additional geographical data, such as premises with permits to serve alcohol, municipal borders, buildings, and roads, were also retrieved from the municipality of Malmö. Since CCTV cameras are only operating between midnight and 6 a.m. on Saturday and Sunday, the main analysis of this study compares public environment assaults committed during that time period relative to control area and control time. The cameras were activated on the August 17, 2012, and in the analysis, the year prior to that date is compared to the following year.

Hot Spot Treatment and Control Areas

In order to evaluate the effect of the operator-controlled CCTV cameras to improve the effect of hot spot policing, the treatment area as defined in the decision to authorize the use of CCTV (Länsstyrelsen Skåne Län, 2012) has been mapped, with an added buffer area of 50 m (Figure 1). Data for 100-m and 200-m buffers not reported in the article are available on request.

OPEN IN VIEWER

Figure 1.

Experiment area with buffer zones and nightclubs (2013).

The hot spot treatment area is compared to a control area consisting of the other major nightlife district in the town. The control area was constructed ex post facto based on density of nightclubs and bars. This is motivated by the strong association between violent crime density and nightlife in Sweden (BRÅ, 2011). The operationalization of nightclubs and bars is based on permits to serve alcohol, and only permits that are valid later than 1 a.m. have been counted. The decision to base the definition of control area on nightclub density rather than, for instance, on density of violence departs from the argument by Nakaya (2000) that the aggregation of geography should be based on independent variables to reduce the risk of biased results. However, the treatment and control areas are also the two main violent crime hot spots of the town (data not shown), with comparable numbers of public environment violent crimes in the two areas (Tables 1 and 2; Figure 2). The nightlife control area in the present study was defined by calculating a kernel density (Chainey, Reid, & Stuart, 2002) of nightclubs based on 200 m radius and then locating parts of the raster that are registered for more than 6 standard deviation higher density of nightclubs than the town mean (Figure 3). As the map shows, the 200-m buffer area of the experiment covers a large part of high nightclub density area, and apart from that, there are just three additional areas, the biggest of which was chosen to act as the control area (Figure 3). See Appendix for statistical procedures.

OPEN IN VIEWER

Table 2.

Police Reported Assaults in Public Environment, Weekend Nights 12.00–06.00 a.m. for Selected Time Periods.

Area	August 17, 2011–August 16, 2012	August 17, 2012–August 16, 2013	Difference (%)	Relative Effect Size [95% CI]
Treatment (experiment area with 50-m buffer)	35	28	−20.0
Control area	32	20	−37.5	0.78 [0.34, 1.78]
Control time	16	21	+31	1.64 [0.66, 4.05]

OPEN IN VIEWER

Figure 2.

Monthly reports of weekend night assault in public environment in treatment (200-m buffer), control area (100-m buffer), and rest of the town 2010–2013.

OPEN IN VIEWER

Figure 3.

Nightclub density, experiment area, and control area.

Results

The results of this study show no effect on public environment assaults of actively monitored CCTV directly linked to police officers at a hot spot of violent crime. Violent crime in public environments fell between 2011 and 2012, before the CCTV effort was introduced (Figure 2). The police have suggested it may partly be due to an increased number of police officers available. In 2012, the Malmö police received a substantial number of extra resources as a result of a high number of shootings, and the extra police officers were used both for visibility through foot patrol and for targeting known criminals. Whether the extra police had such an impact on crime is unknown, but for the present study, it is of interest to note that violent crime fell rapidly before CCTV cameras were implemented in August 2012. In addition, there is a large monthly variation in reported public environment assaults, with a yearly spike in the hot spot treatment area in August due to the town festival taking place in the middle of the month (Figure 2).

As shown in Table 2, the implementation of operator-controlled CCTV cameras during weekend nights did not lead to fewer reports of assault in the town of Malmö during the time the cameras were used. Assaults in the public environment fell in the experiment area but much less than in the control area. As compared to a control time, there is however a reduction in assaults during the treatment time. Such a difference could be considered indicative of a possibility that the operator-controlled CCTV cameras helped curb an overall trend of increasing assault rates in the treatment area during times they were active. The potential for such an effect has been discussed in relation to the evaluation of hot spot policing efforts before (Wikström et al., 1997). The relative reduction is however not significant. Overall, the nonsignificant and differentially directed results suggest the cameras have not had a major impact on crime. There is no evidence for a reduction in public environment assaults associated with operator-controlled CCTV cameras to assist directed police patrols in the area. The same substantial results are achieved for three different sizes of the experiment area and two different sizes of the control area (data available on request). With both confidence intervals containing 1, no effect is considered to be significant.

Discussion of Results

Although the findings of this study differ from those of Piza et al. (2015), there are several possible explanations for such a discrepancy. One important difference is that the variance between treatment and control areas identified by Piza et al. (2015) was based on the level of active monitoring and directed patrols, while in this study the difference is based on the presence or absence of actively monitored CCTV. A key point when compared to previous studies on this topic is the number of people involved in the CCTV-assisted directed patrol experiment in the present study. With two police patrols in a single area and one police officer acting as CCTV operator for just six cameras, both the police presence and the police surveillance is greater than in previous studies finding significant effects (La Vigne et al., 2011; Piza et al., 2014; Piza et al., 2015). It should however be noted that the police officer monitoring the cameras is taken from the pool of active officers, thus meaning one less police officer is on the streets when cameras are active. Even if other differences between the United States and Sweden are disregarded, direct comparison of the studies is not possible because the studies analyze different things.

In addition, the statistical power of the present study is very limited with the relatively low frequency of public environment assaults in the treatment and control areas. To reach significance, a very large relative change, amounting to a relative effect size of 2.45, must be registered. In practice, this would mean that the number of violent crimes in the treatment area/time would have needed to drop from 35 to 14 if no change would be registered in the control area. Larger buffer sizes were tested too, but the difference was not large enough to yield significantly better statistical power, and the more valid small areas reported here were therefore used. Such a large difference is fairly uncommon, and it may thus be worthwhile to further investigate potential sources for nonsignificant change in both the treatment and control areas to determine whether the operator-controlled CCTV cameras have contributed to a less violent public environment. This is of particular importance for the police that take one police officer off the streets to monitor the cameras every weekend night. In addition, the efficacy of CCTV surveillance is of interest in relation to concerns that have been raised over privacy issues of surveillance. It has been argued that increased surveillance can be detrimental to peoples control over how information about themselves is communicated (Taylor, 2010). If CCTV surveillance is not efficient, it could be suggested that such concerns should point to a reduction in the use of CCTV.

There are a number of potential shortcomings in the present study. Police recorded data in general suffer from unreliability, and more specifically, a recent trend toward a higher willingness to report crimes which may have an impact on results over time has been discussed (Granath, 2013). Although there is no specific reason to believe willingness to report would differ between control and experiment area, it remains a possibility, not least in relation to potential differences in sociodemography between the treatment and the control areas (data not reported), something which in previous studies have been shown to influence reporting (Goudriaan, Wittebrood, & Nieuwbeerta, 2006).

Furthermore, there is a lack of documentation available to researchers from the police over the specifics of the hot spot policing effort, and it cannot be ruled out that there have been changes made to police procedures in the treatment area, the control area, or both that may have had an impact on results. There has also been a discussion as to whether hot spot policing leads to more reports of crime due to increased police availability at the scene (Granath, 2013). As discussed previously in the article, there was already a hot spot policing effort in place when CCTV cameras were implemented, and the increase in police availability due to operator-controlled CCTV cameras should thus be limited.

This is also related to a more general shortcoming based on the reliance of police data in this study. Crimes can and will be committed anywhere on the square where CCTV cameras were implemented in the town of Malmö, but they will be registered to the nearest address. This means that the inclusion or exclusion of an address in the treatment area can have a fairly large impact on the numbers, but given that different buffer sizes yield the same substantial results, there is no such bias present in this study. Although employment of buffers means a reduced risk of faulty data having an impact, it also means that the actual treatment area isn’t studied per se. It is possible that CCTV cameras may register something within the treatment area, which later leads to crime outside of the treatment area, but in general, most of the effect of the CCTV cameras should be exclusive to the treatment area rather than to the buffers. In the future, it would be beneficial if the police could attempt to improve the geographic reliability of their data to facilitate evaluation of experiments and more generally of crime trends at places.

Finally, it should be noted that the substantial decrease of assault in public environments recorded between 2011 and 2012 (Table 1) that preceded the introduction of CCTV cameras in part may be related to the more general hot spot policing effort the police started in 2011. Unfortunately, there is little documentation over that effort, and a more formal evaluation of such an effect is difficult to do when no known start date or other crucial information is available.

Conclusions

The results presented in this study suggest that the implementation of operational CCTV cameras at the square Stortorget in the town of Malmö to improve directed patrol efforts has not had a significant crime-reducing effect on assaults. Although not significant, there is in fact, as compared to the control area, a relative increase in reported assaults in the area where operator-controlled CCTV cameras were implemented to assist police efforts at curbing crime. Comparing weekend nights when the cameras are active to weekday nights when cameras are inactive shows an insignificant decrease of assaults in the experiment area. In the treatment area, there was an increase in reported assaults on weekday nights when the cameras were not active but a reduction when cameras are active.

The main contribution of this article was to show that actively monitored CCTV cameras with directed patrols at hot spots of violent crime may be less effective in a Swedish context than it appears to be in the North America. Further research must show where, and under what circumstances, actively monitored CCTV with hot spot policing can be effective at reducing crime.