Advances in technology continue to revolutionize policing in important ways. Three key advancements that are being used today are body-worn cameras, license plate readers, and gunshot detection syste

Advances in technology continue to revolutionize policing in important ways. Three key advancements that are being used today are body-worn cameras, license plate readers, and gunshot detection syste
Community support for licenseplate recognition Linda M. Merola, Cynthia Lum and Breanne Cave Department of Criminology, Law and Society, George Mason University, Fairfax, Virginia, USA, and Julie Hibdon Department of Criminology and Criminal Justice, Southern Illinois University, Carbondale, Illinois, USA Abstract Purpose– Although the use of license plate recognition (LPR) technology by police is becoming increasingly common, no empirical studies have examined the legal or legitimacy implications of LPR. LPR may be used for a variety of purposes, ranging from relatively routine checks of stolen vehicles to more complex surveillance functions. The purpose of this paper is to develop a “continuum of LPR uses” that provides a framework for understanding the potential legal and legitimacy issues related to LPR. The paper then analyzes results from the first random-sample community survey on the topic. Design/methodology/approach – Random-sample survey (n¼ 457). Findings – The paper finds substantial support for many LPR uses, although the public also appears to know little about the technology. The survey also reveals that the public does not regard the uses of LPR as equivalent, but rather support is qualified depending upon the use at issue. Originality/value – Previous research has not systematically categorized the wide variety of LPR uses, an oversight which has sometimes led to implicit consideration of these functions as if they are equivalent in their costs and benefits. To assist agencies concerned with community responses to LPR use, the paper points to a number of factors tending to decrease support for LPR, namely, the extent to which a use involves purposes unrelated to vehicle enforcement, the extent to which a function involves prolonged storage of individuals’ travel data, and the extent to which a use is perceived as impacting “average” members of the community. Keywords Police, Community relations, Public perceptions Paper type Research paper Introduction License plate recognition (LPR) technology is rapidly diffusing in policing. In a national survey of police agencies conducted by the authors in September 2009, it was found that 37 percent of large police agencies already used LPR and that nearly one-third of the remaining large agencies planned to acquire it within one year (Lum et al., 2010). Additionally, the technical capacities required for storing individuals’ travel data collected by LPR readers, as well as the ability to link this data with other databases, are similarly expanding. Within this climate of rapid adoption, however, a good deal of speculation exists over the legal and legitimacy implications of LPR use. For example, do certain uses of LPR without a warrant violate the guarantees of US Constitution? Additionally, will privacy concerns or the potential for the unauthorized disclosure of travel data lead to significant decreases in community legitimacy or job approval for law enforcement agencies that choose to adopt this technology? The current issue and full text archive of this journal is available at www.emeraldinsight.com/1363-951X.htm Received 12 July 2012 Revised 24 September 2012 8 November 2012 Accepted 25 November 2012 Policing: An International Journal of Police Strategies & Management Vol. 37 No. 1, 2014 pp. 30-51 r Emerald Group Publishing Limited 1363-951X DOI 10.1108/PIJPSM-07-2012-0064 The National Institute of Justice and the Department of the Navy/SPAWAR provided support for this research. Additionally, the authors would like to thank the Fairfax County Police Department for their assistance with this research. 30 PIJPSM 37,1 Despite the pressing need for knowledge about LPR, very few researchers or agencies have examined these issues. For example, as of February 2012, only a handful of law review articles have even briefly discussed the potential legal issues surrounding LPR use (Hubbard, 2008; Rushin, 2011). Even where assessments have been attempted, many are informal in nature and do not rely upon rigorous social scientific techniques of evaluation. In this paper, we conduct the first community survey related to LPR use in order to investigate the public’s views of the technology. Prior to discussion of the survey, however, the next section of this paper describes the technology in more detail, since all readers may not yet be familiar with LPR. Following this, we situate the project within existing research and also provide a brief overview of the relevant legal issues. Additionally, we introduce and discuss a “continuum of LPR uses.” We include this continuum for two reasons. First, it serves as a useful organizing scheme for analyses of LPR. Further, we develop this continuum because the wide variety of LPR uses have not been systematically categorized nor distinguished within the literature to this point. Rather, existing analyses have too often treated the numerous LPR uses as equivalent in their implications or ignored emergent uses which are made possible by increases in the technological capacity of law enforcement. Understanding the characteristics of the range of LPR uses can help researchers to generate more accurate tests of the technology, as well as aid the research and practice communities in thinking about the legal and legitimacy concerns that may arise from varying uses. LPR technology As an operational tool for law enforcement, the license plate reader is a straightforward and easily understood piece of sensory technology. LPRs scan the license plates of moving or parked vehicles while either mounted on a moving patrol car or attached to a fixed location, such as a toll plaza. Once a plate is scanned and its alphanumeric pattern is read by the LPR system, the technology compares the license plate against an existing database of plates that are of interest to law enforcement. Plates “of interest,” for example, might include those on vehicles which have been recently stolen, or whose registered owners have open warrants. When a match is made, a signal alerts the officer to proceed with further confirmation, investigation, and action. Hundreds of cars may be scanned and checked in very short periods of time. LPR technology thereby automates a process that, in the past, was conducted manually, slowly, tag by tag, and with a much discretion. In this manual approach, officers would see a car that appeared suspicious and provide the dispatcher with the plate number, who would then check the plate against a database such as National Crime Information Center to see whether the vehicle was stolen. The dispatcher would radio back to the officer with the status of the vehicle. Over time, it became routine to use in-car computer units – rather than a dispatcher – for this purpose. However, whether carried out through a dispatcher or via a check of in-car computer units, LPRs replace an ad hoc, tag-by-tag approach with an automated and speedy system. In addition to their quick scanning and matching capabilities, LPR is, in a broader sense, an information technology system. These systems can collect and store large amounts of data (most frequently, the time, date, and location of a vehicle’s observation and its plate number)[1] for future record management, analysis, and linking with other databases. For example, license plate numbers collected by a reader mounted on a toll plaza might be stored and then accessed in the future by police to confirm a suspect’s alibi or whereabouts at a particular date and time. Additionally, data might 31 Community support for LPR also be used for predictive purposes. For instance, LPR units could be used to scan and record vehicular activity in front of high-risk locations. Unusual patterns of traffic by one or multiple vehicles that emerge from analyzing collected data might alert agencies to a heightened risk or concern. In theory, with enough saved LPR data, longitudinal information related to places and the activities of individuals could be constructed over time.Because of the sheer volume of tags that LPR can scan in minutes and because of its information technology capabilities, LPR, in theory, can act as a force multiplier for law enforcement engaged in crime prevention and homeland security efforts. However, the effective use of LPR is primarily limited by three factors: the system’s ability to read license plates accurately, the quality and relevance of the other data that may be compared with scanned plates collected by LPR readers, and the way in which police departments deploy the machines. Thus, it follows that improvements and refinements in scanning, data access, and police deployment strategies could potentially improve LPR’s effectiveness in controlling and preventing crime. At the same time, as with many other police tactics, advances in each of these functions can challenge other equally important facets of policing. These include legal concerns about how long data can be stored, to what extent data might be mined or accessed appropriately, the balancing of community values of privacy with security, and the broader concern of continuing police legitimacy within communities. The current state of LPR research and the legal context Although a wide variety of agencies currently use LPR technology, relatively few social scientific studies have focussed on evaluating – or even understanding – its use. The most common type of LPR research has focussed on the functioning of the technology itself – demonstrating its effectiveness in scanning license plates and thereby detecting stolen automobiles in various settings, such as highways, parking lots, or toll booths (e.g. see Home Office, 2007; PA Consulting Group, 2003, 2004)[2]. In addition to these studies, two outcome evaluations measuring LPR’s deterrence effect on automobile and other crimes have recently been conducted in the USA (Lum et al. , 2011; Taylor et al., 2010). Though a limited number of interventions have been tested to date, these studies found no significant evidence of either a general or an offense-specific deterrent effect for LPR. As the LPR continuum discussed below suggests, the various uses of LPR can also present a number of legal and legitimacy challenges to the police. To this point, though, no empirical research has been published on the topic of legitimacy or community concerns related to LPR use. Even with respect to analyses or discussions of the legal issues related to LPR, only a few publications exist (discussed in greater detail below) (IACP, 2009; Hubbard, 2008). To date, a small number of courts have adjudicated cases tangentially involving LPR use, but these opinions represent first attempts by courts to grapple with situations where police have utilized LPR ( Green v. San Francisco, 2011; Machado v. City of New Haven , 2006;New York v. Davila , 2010;US v. Lurry , 2010). Though they exist, these opinions do not provide much guidance to agencies considering LPR adoption because only a very limited number of issues have been raised by the parties to these cases. Practically, this means that it will take some time for the law enforcement community to receive a more definitive answer to the legal questions surrounding LPR policy and use. However, some courts and scholars have examined related issues and technologies which can provide a foundation for our discussion. For example, with respect to the legal 32 PIJPSM 37,1 issues involved, US courts have repeatedly adjudicated challenges to manual license plate checks by individual police officers and consistently held these checks to be constitutionally permissible (US v. Ellison,2006;US v. Walraven ,1989;US v. Matthews , 1980). These decisions hold that individuals possess no “constitutionally protected reasonable expectation of privacy” in their license plates ( Katz v. US, 1967, p. 360), since driving is a public activity. While on the road, the license plate must legally remain in public view at all times, a factor which most courts have viewed as dispositive ( US v. Diaz-Castaneda , 2007, pp. 1150-1151; US v. Ellison, 2006, pp. 561-562; Olabisiomotosho v. Houston , 1999, p. 529;US v. Walraven, 1989, p. 974). At first glance, these arguments might also seem to resolve the constitutional issues related to privacy and the use of LPR. Yet, when the issues surrounding the use of automated LPR technology (as opposed to individual, manual license plate checks) are examined, the courts may raise additional concerns. Indeed, though one might argue that LPR technology simply automates a process that could be carried out legally by individual officers (IACP, 2009, p. 12; Hubbard, 2008, pp. 6-9), this assertion relies on the fact that there are no significant legal distinctions between individual officers checking license plates by hand and the use of LPR. Additionally, if translated into policy, this contention is also premised on the notion that the widespread use of LPR would not have disparate effects on police legitimacy, job approval, or community perceptions. On the contrary, a number of authors have argued that there are substantial differences between manual checks and the deployment of LPR, even with respect to the most common use of the technology, that of detecting stolen vehicles (Hubbard, 2008). For example, Hubbard (2008) argues that LPR use does not merely make an officer’s job more efficient and less costly, but that it also allows the police to acquire new abilities that no human officer could possess, such as “[reading] license plates at 60 mph and at night” (p. 34). Hubbard points to a number of Supreme Court cases in which the justices have expressed concerns about the use of increasingly invasive technologies by police ( Dow Chemical Co. v. US , 1986;Kyllo v. US , 2001). Courts may raise similar concerns in the context of future proceedings related to LPR. Moreover, while this “automation” argument might possibly resolve the constitutional issues involved with some uses of LPR, it does not fully address the act of linking LPR data to other databases or the preservation of individuals’ travel data for extended periods of time by police. This distinction provides a useful illustration of the importance of the continuum of LPR uses (presented below). A single check of a license plate and the widespread and varied uses of LPR may be viewed differently by the public and by future courts adjudicating LPR issues for the reasons discussed in the next paragraphs. The continuum represents a clearer framework for agencies considering LPR adoption and also underscores the potential for disparate legal and legitimacy implications connected with different uses. For example, as we will see, additional uses of LPR involve connecting a license plate to an individual’s motor vehicle records or connecting the license plate number with tertiary data unrelated to motor vehicles through the use of state motor vehicle databases. These functions may be viewed as distinct from other LPR uses because they involve linking LPR data to specific individuals and their records. As others have acknowledged, this may greatly increase the chance of harm to individuals in the community and may raise serious legitimacy issues if LPR data is misused or accessed illegally (IACP, 2009, pp. 11-12). Courts and members of the community may be unwilling to allow the connection of LPR travel information with the information 33 Community support for LPR contained in some other databases without any suspicion of wrongdoing by the individual.Indeed, though not a US Supreme Court case, there is some support for this notion in the case law. In State v. Donis(1998, p. 40), the New Jersey Supreme Court held that it was not permissible for police officers to use the mobile data terminals (MDT) in their patrol cars to obtain the registered owner’s personal information contained in the New Jersey Department of Motor Vehicles database without “reason to suspect wrongdoing.” Like the MDT searches that concerned the New Jersey Supreme Court, the linking of LPR data to other types of data involves the examination of personal data by the police and might be restricted by future court decisions if some individualized suspicion of wrongdoing is absent. Further, another distinct issue is raised by the collection and prolonged storage by law enforcement of a large quantity of data about citizens (many of whom have committed no crime). It is this momentum toward large scale, routine data storage by police that makes LPR truly unique in comparison with previous police activities. Significantly, as more data is amassed over time, data storage may also implicate the most significant risks to the community through unauthorized or improper disclosure (IACP, 2009, p. 17). And, the decision to save LPR data may involve some particularly nuanced privacy issues because data storage could eventually make it possible for police to recreate the daily activities of specific individuals through LPR data. These potential increases in the generalized surveillance capabilities of police would seem to implicate similar issues to those discussed very recently by the US Supreme Court. In US v. Jones (2012), the court unanimously determined that the Constitution prohibits the warrantless placement by police of a beeper on a private automobile in order to track a citizen’s movements over an extended period. Though this case related to tracking by beepers (and not tracking through the use of LPR data), five justices expressed specific concerns about recent advances in technology that may greatly increase the surveillance capabilities of the police ( US v. Jones, 2012, pp. 26-29, 47-51). And, they specifically noted the seriousness of these concerns, even while recognizing the many previous court decisions holding that individuals do not have a reasonable expectation of privacy on the roads ( US v. Jones, 2012, pp. 20-21). In future cases, if the Court wished to, it could affirm the use of LPR, by distinguishing its use from the warrantless use of beepers declared impermissible in Jones . To do this, the court would need to emphasize the fact that beeper tracking involves an actual trespass by police into an individual’s belongings (in order to place the tracker on the vehicle), while LPR does not. In fact, this point seemed important to the votes of four justices in US v. Jones. However, for five of the justices ( Justices Sotomayor, Alito, Ginsburg, Breyer, and Kagan), this did not seem to be a key distinction upon which the Jonesdecision turned. Rather, these justices expressed unease regarding potential violations of privacy and increased police surveillance capabilities, concerns that seem to translate into the LPR context. Indeed, these five justices could form a majority in future decisions related specifically to limits on the use of LPR or other potential tools of warrantless surveillance. On a conceptual level, it seems more difficult to argue that LPR uses which involve data storage do not reflect a departure from our current notions of privacy. Police clearly do not presently store large quantities of data about citizens’ daily activities for extended periods of time. Authors writing about the privacy issues inherent in LPR have forcefully argued that the saving of LPR data should be viewed as distinct because these functions involve more than investigations of those suspected of 34 PIJPSM 37,1 breaking the law (Hubbard, 2008). Rather, these uses “track innocent people in the event that they may commit, or be involved in, a crime in the future [y] ” (Hubbard, 2008, p. 28). As we will see below, LPR systems may indeed be used for proactive or predictive purposes. Moreover, by storing data about the movements of individual vehicles, LPR systems may allow police to discern much more than the criminal activities of individuals. Stored travel data may convey a great deal of information about who an individual associates with, which doctors or religious services she visits, which protests she participates in, and even which political party she belongs to. Normally, these activities are “dispersed over space and time,” so police officers cannot see them all at once (Reiman, 1995, p. 29). However, the collection and storage of data may bring many of these bits of information together on one system or connected systems. This is a strong argument for considering the functions involving prolonged data storage to be – at the very least – conceptually distinct from those involving the immediate use and erasure of data. The IACP has cautioned that inaccurate data or even data taken out of context may yield an erroneous picture to law enforcement, an occurrence that may actually hinder investigations (IACP, 2009, pp. 12, 14; Solove et al., 2006, p. 522). Misleading data stored for an extended period may also be very difficult for individuals to refute, since people normally do not keep detailed records of their activities and may not remember their locations once time has passed. Finally, it is also possible that LPR could impact the exercise of other rights or lead to changes in individual behaviors as members of the community realize that their daily activities could be routinely recorded, preserved, or even used against them as evidence (IACP, 2009, p. 16). A goal of LPR is to discourage the commission of illegal acts, but widespread use of the technology may also lead individuals to suppress unpopular, unconventional, or embarrassing actions that are not illegal (Reiman, 1995, p. 35). For example, courts and community members may be concerned that it is difficult for individuals to exercise First Amendment rights, such as through participation in a rally or demonstration, without traveling to do so (IACP, 2009, p. 14). In this way, the implications of LPR may stretch beyond law enforcement or even privacy concerns to potentially influence a wider variety of legal behavior by members of the public. The continuum of LPR uses as a framework for analysis As indicated above, license plate readers have a range of functions; these include the scanning of passing cars to check if they are stolen, as well as the potential storage of data about vehicular movements to access locations of vehicles at a later date. Each potential type of LPR use may be associated with distinct benefits (such as deterrence and crime prevention) and distinct costs. Costs might include legal challenges or a reduction in the community’s view of police legitimacy. Since legal and legitimacy issues may be contingent upon the type of LPR use involved, potential benefits and costs need to be categorized by researchers and agencies in a way that can match uses with potential implications. This step is all the more crucial because agencies are currently acquiring LPR units quickly and at a substantial cost, and promulgating policy in a low-information environment. The development of a continuum of uses for LPR can provide a tangible framework for aiding agencies as they consider adopting and deploying LPR readers. Figure 1 presents one possible continuum of LPR use. Each category (or space on the continuum) represents a type of LPR use. As one moves farther to the right of the 35 Community support for LPR continuum, additional legal and legitimacy concerns seem likely to be raised by the uses of LPR located there. Moreover, the intensity of these concerns may increase exponentially as uses become more predictive in nature. Specific points along the continuum (1) Primary use: auto theft and cars of interest. This use of LPR involves an immediate check of a motorist’s license plate in order to detect whether that vehicle or license plate has been stolen or whether the particular vehicle is the subject of a search related to an investigation. We characterize this scenario as an “immediate” use of LPR because existing data that already identifies stolen vehicles is accessed, and the data collected from the LPR reader need not be stored for any length of time in order to perform this function. Research suggests that this is the way in which license plate readers are most frequently used by law enforcement agencies (Lumet al., 2010). According to a survey of agencies conducted by the authors in 2009, 91.4 percent of agencies with LPR use the technology for this purpose (Lum et al., 2010). Since this use of LPR does not necessitate data storage, it also seems reasonable to hypothesize that this function might raise the fewest privacy concerns or challenges to police legitimacy. Even with respect to this use of LPR, however, there are some arguments for considering this location on the continuum as conceptually distinct from traditional manual checks of license plates. For example, the argument has been made that the deployment of this technology represents more than simple automation or mere efficiency gains because the technology allows law enforcement to accomplish acts outside of human capabilities (Hubbard, 2008; Reiman, 1995). For example, the use of LPR allows officers to check license plates when it might be too dark outside for the human eye to see or even on the freeway when passing cars are going too fast for the human eye to register a license plate number (Hubbard, 2008). For these reasons, even the most common uses of LPR may be viewed by some members of the community as a departure from manual checks of license plates and, as a result, even these uses may have implications for police legitimacy. (2) Connection of LPR data with a secondary data source . The complexity of LPR use increases as one moves to the right of the continuum. The next likely use of LPR Primary Use Tertiary Data Mining PredictiveAnalysis Magnitude of Concern/Challenges Complexity of LPR Use Connection with Secondary Data Source Data Collection and Storage for Proactive Use Figure 1. Continuum of LPR uses 36 PIJPSM 37,1 involves the connection of scanned license plates to a secondary data source associated with those plates, usually the linking of LPR data with records from a state’s Department of Motor Vehicles. Therefore, at this step on the continuum, information from the LPR readers is connected for the first time to an actual individual (the registered owner of the vehicle) and then to that owner’s motor vehicle record. Unpaid parking tickets, lack of insurance, and other traffic-related delinquencies might be accessed. As a practical matter, the use of LPR technology at this step on the continuum also begins to raise issues of personal security for individuals in the community, since LPR data has now been linked with a specific individual. Prior to LPR systems, manual approaches also routinely required motor vehicle records to be accessed by police in the investigation of traffic and other offenses. The advent of LPR alters this only in the fact that these records would be accessed in an automated fashion and, perhaps over time, on a much larger scale as LPR use becomes more widespread. In addition to those suspected of a violation or a crime, the LRP data from law-abiding individuals might also be checked against existing databases of the types mentioned in the last paragraph. Most importantly, since LPR data would be linked with information about specific individuals via the state’s Department of Motor Vehicles database, this step on the continuum heightens the need for stringent standards for data handling. (3) Tertiary data mining . The third location on the continuum involves further data linking, this time connecting LPR data with tertiary databases not directly related to vehicle enforcement by using motor vehicle information to identify persons of interest. Prior to LPR, an investigation may have involved the police running a tag for a vehicle’s registered owner and then checking for the existence of an open warrant for the owner’s arrest. LPR certainly accelerates and automates this function. However, LPR use is not limited to checks for open warrants. Rather, the functions of license plate readers that fall into this third category can vary widely. For example, data that might be cross-checked against LPR data includes the license plates of vehicles owned by registered sex offenders, by individuals delinquent in the payment of child support, by recently released violent offenders, or even by individuals arrested for selling drugs around schools or public parks. In addition to helping to investigate or locate these individuals generally, vehicles with LPR may be deployed to patrol around specific locations, such as schools and parks. All of these LPR uses are similar, however, in that they involve the connection of LPR data to other data sources, but for law enforcement purposes unrelated to motor vehicles or vehicular enforcement . Novel legitimacy issues may arise precisely because the police have now used LPRs for functions other than law enforcement related to vehicles or traffic. Since LPR is not being employed as a technological tool for more efficient traffic or vehicular enforcement at this space on the continuum, members of the public could view these uses as promoting the acquisition of more generalized surveillance capabilities by the police. Thus, it seems a reasonable hypothesis that the use of LPRs for these functions may heighten the likelihood that LPR adoption will impact police legitimacy, community approval, or other facets of police-community relations. Moreover, even within this category, different uses may evoke varying responses. For example, members of the community may view sex offenses as grave enough to warrant the use of LPR to prevent individuals who have committed these crimes from entering school zones. Yet, the community might not tolerate other uses where the perceived benefits are too few or the perceived intrusions into the personal lives of community members seem too great. Though some authors writing on this topic have 37 Community support for LPR suggested hypotheses about the likelihood that some uses might be accepted over others, these hypotheses have not yet been tested through survey or other research. These differences are examined in greater detail in the community survey results, presented below.(4) Data collection and storage for proactive use . This step on the continuum involves the long-term preservation of data from LPR readers for future investigative purposes. For example, during an investigation, saved LPR data might demonstrate that a suspect’s vehicle traveled to a certain location. Alibis of suspects might also be corroborated or challenged from the information captured by LPR units placed at toll roads or near locations where an individual claimed to be. Yet, others have highlighted the potential negative consequences of this type of data retention, for example, that saved data may also prejudice the investigatory process against an individual. This is a concern because LPR information may be presumed to be correct even in instances when the data may be misleading. For instance, if an LPR unit records the presence of a vehicle at a particular location, this does not necessarily mean that the registered owner was driving the vehicle at the time. Additionally, after time has passed, it may be difficult for an individual to combat an assumption that the data presents an accurate picture of daily activities, since individuals do not normally keep detailed records of their day-to-day routines. Additionally, data storage raises an even more serious potential for abuse through either hacking or misuse; as a result, rigorous testing of policy in this area of the continuum is critical. Members of the community may also hold very strong opinions regarding whether or not this information should be considered private and even if data of this type should be collected and maintained by the police. The survey discussed below provides systematic evidence regarding one community’s response to these questions. (5) Predictive analysis . While stored LPR data might be relevant to ongoing investigations, searches for individuals, or the verification of alibis, LPR data may also be used for more predictive analysis, an extension of its proactive use. Predictive analysis involves the analysis of collected data to determine patterns of behavior and movements in order to anticipate and prevent future crime. One example might be the decision to place LPR units at locations around an arena prior to a major event. Unusual vehicular activity or multiple hits of particular vehicles in front of a location may be found by analyzing the saved data. Proactive investigations based upon this analysis might then be generated. Similar to these uses discussed above, vehicles might also be scanned for connection to other databases in order to anticipate problems for prevention purposes. This type of analysis logically would seem to offer special challenges to the legitimacy and legality of police actions. On the one hand, large amounts of data combining information from many incidents and individuals could be examined for overall patterns of behavior. This type of procedure is commonly used in intelligence analysis, where meaningful patterns may be found within large amounts of seemingly routine data. However, any type of predictive analysis runs the risk of false “positives” raising suspicions about innocent individuals. Anticipating and reducing the negative impact of false positives is an important crime prevention goal of democratic police agencies. Since predictive analyses utilizing LPR data may be undertaken in many different contexts, the reaction of the community may be dependent upon the context of such use. It is useful to gauge how such deployment of LPR units might be received by the community, an investigation which is begun in the survey discussed below. 38 PIJPSM 37,1 Data and methods Survey instrument and sampling The use of LPR may have important implications for police legitimacy and community-police relations, two factors that may further affect an agency’s ability to prevent and deter crime (Tyler, 1990). To explore these implications, we conducted a community survey in Fairfax County, Virginia, one of two locations in which the authors also later carried out experimental evaluations of the deterrence effect of LPR technology (Lumet al., 2011). Fairfax County is one of the large Northern Virginia suburban counties outside of Washington, DC, where many individuals who work in the metropolitan DC area reside. According to the US Census, the County has a population of 969,600 persons; approximately 62.7 percent are Caucasian, 17.6 percent are Asian, 9.2 percent are African American and 15.6 percent are Hispanic [3]. Fairfax County spans almost 400 square miles, with a population density of about 2,450 persons per square mile. The police department consists of approximately 1,370 sworn officers serving a relatively well-educated community (over 50 percent of residents have a college education) with a high home ownership rate (70 percent). To carry out this survey experiment, we randomly sampled 2,000 Fairfax households, from all residential units/households in Fairfax County [4]. To select only residential properties, we first used a zoning polygon file in ARCGIS, which represented all of the different land use zoning districts within Fairfax County (3,962 zones of a possible 7,496 zones). Then, using a building point file, we selected only the addresses that fell within areas that were zoned as residential. The result was 237,444 residential addresses from which we could randomly draw our sample of 2,000 possible respondents. Once the initial 2,000 residences were selected, each was checked individually against the County’s public real estate records to ensure that the residence was occupied, that we had the proper mailing address, and that there were no duplicate addresses. If the online database indicated that an address did not exist or referenced a non-residence (such as a church, school, etc.), the address was removed from the data set and replaced with another randomly sampled residence. In total, we replaced 106 cases. The survey was then mailed to respondents along with a consent document and an introductory letter. The survey instrument contained a mix of demographic questions, general questions about crime and police legitimacy, and questions targeted to specific points on the continuum of LPR uses presented in Figure 1. Participants were also asked separate questions about their support for the variety of LPR uses discussed above. Finally, respondents were asked whether or not LPR data should be considered private information. Response rate We sent out the first round of the survey to a sample of 2,000 households in May 2010. Respondents were given the choice to complete the survey in hard copy and return it by business-return envelope or to complete the survey online. Though the envelope containing the survey was addressed to the “current resident,” the consent document explicitly stated that the respondent had to be 18 years or older in order to complete the survey. Approximately every two weeks following the first mailing, we sent further reminders about the survey to those addresses from which we had not yet received a response. We did this until we ended data collection in mid-July 2010. The survey materials noted that the survey was being administered jointly by the George Mason University and by the Fairfax Police Department. 39 Community support for LPR At the conclusion of data collection, 457 Fairfax residents had completed the survey, yielding a response rate of 22.9 percent, a relatively high response rate for a mailed survey. In terms of gender, the response pool included 48.9 percent female and 51.1 percent male respondents. With respect to race, the respondents indicated that they were 85.8 percent Caucasian, 3.7 percent African-American, 3.4 percent Latino, and 7.1 percent Asian/Pacific Islander. The divisions reported with respect to political party identification were 33 percent Democrat, 30 percent Republican, and 37 percent Independent. We used GIS software to link Census block-group information to addresses in our sample, and then compared respondents and non-respondents on their block-group estimate means. We compared block-group levels of poverty, unemployment, median family income, home ownership, linguistic isolation, and racial neighborhood composition. t-Tests of means did not indicate that those who responded to the survey were significantly different (with regard to social, economic, and demographic factors) than those who did not respond. However, despite this, there are some limitations to the postal survey methodology which should be acknowledged. As with all postal surveys, though the households receiving the surveys were randomly selected, it is not strictly possible to ensure the randomization of individuals within each household (since the survey was addressed to the “current resident,” rather than to specific individuals). It is possible that individuals with a greater tendency to be community-minded or even pro-police opted to answer the survey at higher rates than did other individuals in the community. If so, this might produce a response pool containing answers more supportive of police or more accepting of new or innovative police technologies. Additionally, along similar lines, it should be noted that relatively small numbers of African-Americans and Latinos elected to respond to our survey. It is possible that a lowered response rate among members of these communities might also translate into a sample more predisposed to support the police. Results Familiarity with LPR technology The survey suggests that many members of the community (62.8 percent) have heard of LPR, but that residents currently do not know much about the technology. In fact, almost 90 percent of respondents were willing to admit that they did not know whether or not their local police used LPR. When asked a factual question such as this, there is sometimes a tendency on the part of survey respondents to be unwilling to admit a lack of knowledge. However, in this case, the fact that nearly 90 percent of respondents selected “I don’t know” rather than attempting to answer (or “guess”) may underscore the degree to which residents are not yet familiar with LPR. These respondents seem to have felt little social stigma attached to a lack of knowledge about LPR. These results are consistent with the fact that public discourse on this issue has been extremely limited to this point. Moreover, as we shall see, this impression is confirmed by examining the results of this survey as a whole, as there are a number of questions where significant percentages of respondents expressed no opinion regarding various LPR issues. Primary and other “immediate ”uses of LPR Figure 2 demonstrates that respondents were generally supportive of both the primary use of LPR (detecting stolen autos) and of what we have termed the other “immediate” uses of LPR (those at continuum points (2) and (3), not requiring prolonged data 40 PIJPSM 37,1 storage). Specifically, when referencing the retrieval of stolen vehicles, 79.9 percent of respondents indicated that they would “strongly support” or “support” a decision by their local police to use LPR in this manner. This figure represents a very high level of support; in fact, only 10.7 percent of respondents indicated that they would oppose or strongly oppose a decision by the police to use LPR to detect stolen vehicles. In total, 9.4 percent of individuals indicated that they were neutral with respect to this question, leaving only a very small population of potentially undecided individuals (Figure 2).Many of the remaining “immediate” uses of LPR described in Figure 2 are also supported by the majority of the community. For each of these survey items, respondents were asked to assess the decision to use LPR when the technology is employed to detect crime at the moment that the data is collected. For example, 76.6 percent of respondents either “strongly supported” or “supported” the use of LPR to check passing vehicles to see if registered owners are wanted for crimes. Similar to the primary use of LPR, then, community support for checking outstanding warrants was quite high. This is in marked contrast to the community response to a survey item that asked about checking all passing vehicles for unpaid tickets and parking violations. Though directly related to traffic regulation and conceptually close to the primary use of LPR, this item represents the only immediate use of LPR that was supported by less than a majority of respondents (48.1 percent). Potential explanations for this result are discussed in detail below. 0% 10% 20% 30% 40% 50% Check all passing vehicles for stolen status (n =448) Check all passing vehicles for parking violations and unpaid tickets (n =449) Check if vehicle owners are wanted for a crime (n =449) Check if vehicle owners are sex offenders (n =447) Strongly Oppose Oppose Neutral Support Strongly Support Investigate all vehicles to prevent terrorism (n =451) Figure 2. Community responses to primary and immediate uses of LPR 41 Community support for LPR As expected, community support for the tertiary uses of LPR was also generally high (space 3 on the continuum), though not as high as for the primary use. Specifically, 66.7 percent of respondents either “strongly supported” or “supported” using LPRs to check if registered owners of passing vehicles are sex offenders. Similarly, 70.1 percent of respondents either “strongly supported” or “supported” utilizing LPR to investigate all vehicles passing or parking near important places or buildings for the purposes of terrorism prevention. Interestingly, despite the fact that suspected terrorists and child molesters are among some of the most despised categories of individuals, support for these uses was somewhat lower than support for the use of LPR to retrieve stolen vehicles. As discussed above, this may result from the fact that these uses are not directly related to vehicle enforcement. It seems logical that members of community – while still very supportive of these uses – might view these LPR functions as more intrusive and farther removed from the primary use of LPR. Community reactions to the storage of LPR data Figures 3 and 4 present the results of the survey relating to LPR data storage. To begin, we asked respondents to specify whether they considered the four elements of LPR data (date of vehicle observation, time of observation, license plate number, and the location of the observation) to be public or private information. Surprisingly, despite the fact that those in the sample were generally quite supportive of LPR use, the majority of respondents (53.4 percent) considered LPR data to be private information. In designing the survey, we purposely placed this question prior to any other survey items regarding specific uses of saved data. This was done in order to guard against any potential bias that could be introduced through concern or reflection over specific uses of saved data. In addition to the majority that responded that LPR data should be considered private, 17.3 percent of the respondents expressed neutrality with respect to this question. Like some of the other survey items, this reflects a fairly large percentage of undecided individuals. Once the public becomes more familiar with 40.00% 35.00% 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% Strongly Private (5) Private (4) Neutral (3) Not Private (2) Strongly Not Private (1) Note: (n=451) Figure 3. Do you believe that information from LPR readers should be considered private? 42 PIJPSM 37,1 LPR and experiences its use within the community firsthand, the opinions of these individuals may be altered.In comparison, the results are about evenly split when respondents are asked how long LPR data should be saved. As a response to this question, participants were permitted to select one of four options: that the data should be not be saved; that it should be saved for about one week; that it should be saved for about six months; or that it should be saved for as long as the police wish to save it. In the end, 30 percent of respondents opted for the six-month retention period, while approximately 23 percent of respondents opted for each of the remaining categories. This result could reflect a small preference for a data storage period of approximately six months, but the fact that the responses are so evenly split across all options more likely reflects a lack of developed opinion on this issue. Moreover, when we examine the findings of the survey targeted to the later steps on the continuum, we find varied results (Figure 4). The use of saved LPR data to learn about the past activities of vehicles suspected of a crime yielded the highest levels of support (87.6 percent). Additionally, the use of saved data to investigate the activities of suspected terrorists was also associated with very high levels of support (79.0 percent). Support then declined slightly when respondents were asked about the use of saved data to examine the activities of individuals suspected of a crime (71.1 percent) and the past activities of sex offenders (66.7 percent). In contrast, proposals to utilize the same data to investigate “all vehicles which drive around an important place or 0% 10% 20% 30% 40% 50% 60% […] to find the last location of a vehicle connected with a crime? (n=444) […] to investigate all vehicles which drive around an important place or building? (n =437) […] to learn about the past activities of a suspect who is under investigation for a crime? (n =443) […] to learn about the past activities of a person suspected ofterrorism? (n=442) […] to learn about the past activities of sex offenders? (n =447) Strongly Oppose Oppose Neutral Support Strongly Support […] to learn about the activities of parents who don’t pay child support in order to force these parentsto appear in court? (n =439) Figure 4. “The police should be able to use saved LPR data [ y]” 43 Community support for LPR building” prompted only 53.1 percent of respondents to mark “strongly support” or “support.” And, an item asking about the use of LPR to learn about the activities of parents who do not pay child support garnered the least support of the “saved” data uses (51 percent). Discussion The survey conducted here provides the first evidence regarding public support for the adoption of LPR technology by law enforcement agencies. Although our survey results suggest relatively robust support for many uses of LPR, this finding must be moderated by the fact that, to date, only one community survey has been conducted on this topic. Despite this, the results presented above are helpful to researchers and law enforcement agencies interested in LPR for a number of reasons. First, at present, so little evidence exists as to community views of the technology that even one systematic study in an American community represents a significant advance in the evidence base. The systematic investigation of community opinion related to emerging technologies, like LPR, is important because alterations in the conduct of police functions can influence agency legitimacy and community approval, which in turn can also impact willingness to abide police and to obey the law (Tyler, 1990).Second, given the specific characteristics of the community selected, these findings may translate to similar suburban communities, particularly where support for the police is relatively robust. Of course, these results may differ in communities where lowered levels of police legitimacy are found. As a first study of LPR, however, this research provides a foundation for testing of further hypotheses about public opinion. Lastly, in addition to the survey findings, a main contribution of this paper was to delineate a conceptual framework for future testing and analysis. To fill this gap, the continuum of uses was detailed above and then explored using survey results from one community. Indeed, when the survey results are viewed through the organizing scheme of the continuum of LPR uses, it becomes apparent that the public does not regard the various LPR uses to be equivalent, but rather makes significant distinctions when considering the ways in which LPR may be used. Indeed, support for LPR uses varied from 48.1 to 87.6 percent, depending upon which function was contemplated by respondents. As such, the survey results suggest that distinctions between LPR uses are not merely conceptual in nature, but seem to be associated with significant variations in the level of support reported by members of the community. To explore the hypothesis that members of the community view different uses of LPR as distinct in their implications, we targeted individual survey items to specific locations on the continuum presented above. Future studies would also benefit from considering the different uses of LPR as distinct, both in their potential benefits and also in their potential implications in the eyes of the community and of the courts. One of the notable characteristics considered by the authors to distinguish conceptually the various points on the continuum was the extent to which a particular LPR use relates to vehicle enforcement. For the most part, expectations that this factor would impact support for specific LPR functions are reflected in the survey results. Community support for LPR uses closely related to vehicular enforcement appears to be high. This includes both the primary use of LPR (retrieval of stolen vehicles) (79.9 percent support) and the secondary uses specifically related to vehicles (step 2 on the continuum), such as utilizing LPR to check outstanding warrants (76.6 percent support). The fact that support for these immediate, vehicle-related uses is so high is 44 PIJPSM 37,1 consistent with their location at the start of the continuum. Members of the community seem to feel comfortable with the implications of these uses and seem confident in their judgments about the potential tradeoffs involved with these functions.In fact, a survey finding suggesting that nearly 80 percent of the public supports any policy or government initiative is actually quite striking. One explanation may be the perception that automated license plate checks (without data storage) are similar to the manual license plate checks already conducted by police. Thus, this finding also appears consistent with previous judgments of US courts, emphasizing that individuals do not generally find license plate checks to be too intrusive ( US v. Diaz-Castaneda , 2007, p. 1151;US v. Walraven , 1989, p. 974). Of course, in communities with lower levels of police legitimacy, support for even these uses of LPR technology may correspondingly be reduced. Future research should aim to discern the exact magnitude of this impact on the categories of LPR use found in the continuum. In this case, however, as the “immediate” uses of LPR raised by the survey became less closely tied to the regulation of vehicles, the results suggest that community support for these uses began to erode, at least slightly. The results with respect to investigating vehicles to prevent terrorism (70.1 percent support) and using LPR to check if vehicle owners are sex offenders (66.7 percent support) provided examples of this. To be precise, though a substantial majority of the community still supported each of these uses, some additional members of the public appear to have questioned the tradeoff between enhanced enforcement and the expanded reach of LPR when connected with tertiary data sources. This result is particularly interesting, given the fact that the two items included on the survey – utilizing LPR to aid in the prevention of terrorism and of sexual offenses – are generally considered to be exceedingly important law enforcement purposes. In fact, no doubt many would argue that these purposes are actually more critical than those related to vehicular enforcement, detailed at continuum steps 1 and 2. Yet, respondents supported the connection of LPR data with tertiary data bases – even for these very serious purposes – in lower numbers than they supported using LPR for vehicle-related enforcement. Moreover, the lowest support for any LPR use mentioned in the survey is found when a police decision to use LPR to check passing vehicles for unpaid tickets and parking violations is mentioned. Interestingly, though this function is related directly to vehicle enforcement, this use was not supported by even a majority of respondents (48.1 percent support). The finding that support for this use is significantly lower among members of the community raises a crucial point for agencies considering the adoption of LPR. The most likely explanation for this lack of support is that respondents were easily able to recognize and to weigh a tangible (and even personal) cost that might result from more efficient police enforcement of parking ordinances (that they might be forced to pay fines more frequently). Simultaneously, individuals may also have had a difficult time associating the payment of parking tickets with tangible benefits to themselves or to the community. And, for those individuals sophisticated enough to weigh privacy concerns against potential law enforcement benefits, the enforcement of parking tickets may not have been viewed as an important enough community concern to merit grappling with the potential privacy implications of widespread or expanded LPR use. These issues merit careful consideration by agencies in formulating LPR policy. Individuals in the community may be much less supportive of LPR uses that seem likely to affect them personally or to affect other “innocent” members of the community (such as when LPR is used to give parking tickets). The community can understand the tradeoffs of these uses with relative ease because they are close to an average citizen’s 45 Community support for LPR experiences and individuals can likely envision the possibility of experiencing negative consequences personally. It is not surprising, then, that community support for such uses would drop precipitously in comparison to other LPR functions. Such a result may also raise the possibility that individuals could be receptive to some of the arguments of privacy advocates and others, suggesting that LPR can be problematic because it targets “innocent” citizens in addition to those guilty of criminal behavior (Hubbard, 2008). The survey findings suggest that support for uses perceived as targeting “innocent” citizens seem, indeed, much less likely to be popular.A related caution is raised by examining the survey item suggesting that the majority of respondents deemed LPR data to be private information. This sentiment could provide a platform for alterations in public support for LPR use, particularly should events cause individuals to begin to assess the tradeoffs in terms benefits or costs of LPR differently. As events intervene, if they tip the scale in a manner which suggests that the crime control benefits are not as robust as previously imagined or the costs to individuals in the community greater, a basic belief that this information ought to be considered private may serve as the foundation for changes in opinion. Even if significant changes in opinion do not occur, an erosion of police legitimacy may result from the perception that police are becoming too involved in collecting or storing private information. An experimental test of the legitimacy implications of LPR represents a future direction for research planned by the authors. Legally speaking, the question of whether or not LPR data is to be considered public or private information is also important because several of the court cases referenced earlier in this paper have held that individuals do not have a reasonable expectation of privacy in their license plates while driving ( US v. Diaz-Castaneda, 2007;US v. Ellison , 2006; US v. Walraven , 1989;US v. Matthews , 1980). As we have seen, however, the resolution of this issue may involve broader qu estions than the constitutional protection afforded during a single license plate check . While the courts may not view individual license plate checks as a violation of privac y, it is possible that LPR data may be viewed as distinct because an individual’s daily activities, preferences, or opinions might eventually be capable of being recreated or ascertained through saved LPR data. Similar concerns were echoed by a majority of justices in the Supreme Court’s most recent pronouncement on the issue of vehicle tracking by police utilizing an electronic beeper ( US v. Jones , 2012). Indeed, the survey provides some further indications that the issue of data storage also represents an important factor in how these respondents reacted to the survey. Overall, much like the uses of LPR located on the left side of the continuum, those in the sample generally expressed support for the uses of saved LPR data located on the right side of the continuum. Yet, in a number of instances, perceptible erosion in support can also be detected when compared with the corresponding “immediate” uses of LPR. For example, 76.6 percent of respondents supported using LPR data to check outstanding warrants immediately following the collection of the data, while 71 percent of respondents supported using stored data to do so. Admittedly, this does not represent a large alteration in support; over two-thirds of respondents still supported the saved data use. However, at this point in the life cycle of LPR technology, agencies and researchers should be attentive to even slight alterations in support, as the public has encountered almost no negative or critical information about LPR to this point. Since LPR is an emerging technology, public opinion related to the topic is likely to be in its infancy. Further, as this survey represents the first study of its kind, one goal of this research was to begin the process of establishing a cross-sectional measurement 46 PIJPSM 37,1 of “baseline” support for LPR. For this reason, no information regarding privacy concerns was provided to respondents. While support for LPR appears to be robust, a cautious approach merits leaving open the possibility that support could change significantly if instances of hacking, misuse or harm to community members resulted from LPR uses involving saved data.Additionally, it is important to note that only a bare majority of respondents (53.1 percent) supported the use of saved LPR data to investigate vehicles passing or parking in front of important places. In this age of concern about terrorism, it is surprising that this use did not produce a greater consensus. This finding is even more striking when it is considered that the sample was drawn from residents of Fairfax, Virginia, a community in close proximity to the nation’s capital, an area with a tangible connection to terrorism and many famous landmarks. A similarly lowered level of support was found when respondents were asked about the use of saved LPR data to investigate individuals delinquent on child support payments (only 48.4 percent support). In both of these cases, it seems theoretically plausible that concerns about data storage moderated otherwise high levels of support for the perceived efficiency increases associated with LPR. Once again, we also find evidence that support was qualified by the perceived likelihood that a particular use (this time, of saved LPR data) might impact “average” or “innocent” members of the community (as opposed to those engaged in criminal behavior). The uses of saved LPR data discussed in the last paragraph represent functions that would impact “average” members of the community because those survey items mentioned using LPR to investigate “all vehicles” or “parents” delinquent on child support payments. Inherent in these questions, then, were references to either “average” members of the community or to “all” members of the community, and support declined. In contrast, support remained high when survey items specifically referenced “sex offenders,” or “individuals suspected of a crime.” This is likely because members of the public have a difficult time associating potential alterations to the privacy rights of “average” members of the community with survey questions related to intensified investigations of individuals engaged in such criminal activity. In fact, outside of the LPR context, researchers investigating the underpinnings of public support for rights, such as individual privacy, have found similar results. Specifically, a frequent finding of this research has been that members of the community – who are generally supportive of individual rights in the abstract – tend to default more frequently to a pro-security (or pro-crime prevention) stance when asked about specific public policies that might limit the rights of members of disliked groups (Marcus et al., 1995; McClosky, 1964; McClosky and Brill, 1983; McClosky and Zaller, 1984; Prothro and Grigg, 1963; Stouffer, 1955). Sex offenders and other criminals certainly qualify as members of disliked groups. Members of the public often have a difficult time connecting and weighing potential limitations to the rights of small, disliked groups (such as “sex offenders”) with possible erosions of broad rights guarantees (such as privacy) that may ultimately impact “average” citizens. In short, it is difficult for most members of the community to perceive that these changes might impact them personally in the future when they do not envision concrete impacts immediately. And, the survey results provide one further example that is worth discussing. The community exhibited particularly high levels of support for the use of saved LPR data when used specifically to find the last location of “a vehicle involved in a crime.” Though it involves saved data, this function of LPR was actually associated with the 47 Community support for LPR highest levels of support within this community of any function listed on the survey. A number of factors might account for this result. First, this LPR function relates directly to vehicles, a factor which appears to be associated with increased support for the use of LPR. Perhaps equally important, though, the survey question does not actually refer to any specific individual or owner (but rather only to finding the location of “a vehicle”). This phrasing may have led many respondents to assume that only vehicles (and not persons) were being investigated as a consequence of this function. As a result, it would not be surprising if individuals neglected to consider the privacy implications related to saved data in responding to this particular survey item. Like the uses discussed in the last paragraph, it would be difficult for respondents to envision negative impacts to “average” members of the community stemming from this LPR function. Future discussions of privacy concerns in the media or well-publicized instances of unauthorized data access might make it easier for the public to connect the range of LPR uses to potentially negative, real-world implications for “average” members of the community.We began this paper by underscoring the fact that law enforcement agencies are acquiring LPR readers at a rapid pace and with significant cost, but in a low-information environment. This study represents the first to systematically examine community opinions about LPR and is one of only a small number of studies related to emerging data storage technologies more generally. Though preliminary, this study is useful in that it highlights the importance of considering community sentiment and also points to a number of specific factors which should be considered by agencies in configuring LPR readers, as well as by researchers in generating future hypotheses to test. LPR readers have been shown to function well technologically, but we know relatively little about either the potential crime control benefits or the possible legitimacy costs of their deployment. For evaluation to progress, however, the discussion must move beyond the consideration of all uses of LPR as if they are equivalent. It is our hope that the continuum of uses detailed above may be a useful tool for understanding the varied costs and benefits along the range of LPR uses. Notes 1. Some of the sources that we discovered have considered LPR systems that also record digital images of distinguishing vehicle features (such as damage to the vehicle or bumper stickers) or a digital image of the vehicle’s driver and passengers (International Association of Chiefs of Police (IACP), 2009). These possibilities may raise additional legal or constitutional implications not discussed here. 2. The authors conducted a full discussion of the research related to LPR functioning in a companion article to this piece (Lum et al., 2011). 3. These percentages sum to 4100 percent due to multiple-race reporting. 4. In order to estimate our sample size, we set as our goal a confidence interval (or margin of error) of between 4 and 5 percent. For the size of Fairfax County, this would require a final sample of between 384 and 600 respondents. Based upon the resources available for the project and an estimated response rate of between 20 and 30 percent, we chose to sample 2,000 residents of the County in order to acquire a sample of the appropriate size. References Dow Chemical Co. v. US (1986), 476 US 227. Green v. San Francisco (2011), “US Dist. LEXIS 108617”, available at: www.lexisnexis.com. mutex.gmu.edu/hottopics/lnacademic (accessed November 8, 2012). 48 PIJPSM 37,1 Home Office (2007),Police Standards Unit: Evaluation of Automatic Number Plate Recognition 2006/2007 , PA Consulting Group, London. Hubbard, T.E. (2008), “Automatic license plate r ecognition: an exciting new law enforcement tool with potentially scary consequences”, Syracuse Science and Technology Law Report , Vol. 2008 No. 3. International Association of Chiefs of Police (IACP) (2009), “Privacy impact assessment report for the utilization of license plate readers”, available at: www.theiacp.org/LinkClick. aspx?fileticket ¼N%2bE2wvY%2f1QU%3d&tabid ¼87 (accessed May 24, 2010). Katz v. US (1967), 549 US 956. Kyllo v. US (2001), 533 US 27. Lum, C., Merola, L., Willis-Hibdon, J. and Cave, B. (2010), License Plate Recognition Technologies for Law Enforcement: An Outcome and Legitimacy Evaluation, SPAWAR and National Institute of Justice, Washington, DC. Lum, C., Hibdon, J., Cave, B., Koper, C. and Merola, L. (2011), “License plate reader (LPR) police patrols in crime hot spots: an experimental evaluation in two adjacent jurisdictions”, Journal of Experimental Criminology , Vol. 7 No. 4, pp. 321-345. McClosky, H. (1964), “Consensus and ideology in American politics”, American Political Science Review , Vol. 58 No. 2, pp. 361-382. McClosky, H. and Brill, A. (1983), Dimensions of Tolerance, Russell Sage Foundation, New York, NY. McClosky, H. and Zaller, J. (1984), The American Ethos: Public Attitudes Toward Capitalism and Democracy , Harvard University Press, Cambridge, MA. Machado v. City of New Haven (2006), “Conn. Super. LEXIS 2709”, available at: www.lexisnexis. com.mutex.gmu.edu/hottopics/lnacademic (accessed November 8, 2012). Marcus, G.E., Sullivan, J.L., Theiss-Morse, E. and Wood, S.L. (1995), With Malice Toward Some: How People Make Civil Liberties Judgments , Cambridge University Press, Cambridge. New York v. Davila (2010), 901 N.Y.S.2d 787. Olabisiomotosho v. Houston (1999), 185 F.3d 521 (5th Cir.1999). PA Consulting Group (2003), “Engaging criminality – denying criminals the use of the road”, available at: www.paconsulting.com/es/ NR/rdonlyres/8BEBFD54-94DC-4274-8283- EE1A66D0E09E/0/ANPR_report_24_Oct.pdf (accessed July 21, 2007). PA Consulting Group (2004), “Driving crime down – denying criminals the use of the road”, available at: http://police.homeoffice.gov.uk/news-and-publications/publication/operational- policing/Driving_Crime_Down_Denyin1.pdf?view ¼Binary (accessed July 21, 2009). Prothro, J.W. and Grigg, C.M. (1963), “Fundamental principles of democracy: bases of agreement and disagreement”, Journal of Politics, Vol. 22 No. 2, pp. 276-294. Reiman, J.H. (1995), “Driving to the panopticon: a philosophical exploration of the risks to privacy posed by the highway technology of the future”, Santa Clara Computer and High-Technology Law Journal , Vol. 11 No. 1, pp. 27-44. Rushin, S. (2011), “The judicial response to mass police surveillance”, University of Illinois Journal of Law, Technology, and Policy , No. 2, pp. 281-328. Solove, D.J., Rotenberg, M. and Schwartz, P.M. (2006), “Information privacy law”, available at: http://docs.law.gwu.edu/facweb/dsolove/Information-Privacy-Law/files/IPL-Update-2007.pdf (accessed May 5, 2012). State v. Donis (1998), 723 A.2d 35. Stouffer, S. (1955), Communism, Conformity, and Civil Liberties , Doubleday, New York, NY. Taylor, B., Koper, C. and Woods, D. (2010), Combating Auto Theft in Arizona: A Randomized Experiment with License Plate Recognition Technology , Police Executive Research Forum, Washington, DC. 49 Community support for LPR Tyler, T.R. (1990),Why People Obey the Law , Yale University Press, New Haven, CT. US v. Diaz-Castaneda (2007), 494 F. 3d 1146 (9th Cir. 2007). US v. Ellison (2006), 462 F.3d 557 (6th Cir. 2006). US v. Jones (2012), “US LEXIS 1063”, available at: www.supremecourt.gov/opinions/11pdf/10- 1259.pdf (accessed March 4, 2014). US v. Lurry (2010), “US Dist. LEXIS 118456”, available at: www.lexisnexis.com.mutex.gmu.edu/ hottopics/lnacademic (accessed November 8, 2012). US v. Matthews (1980), 615 F.2d 1279 (10th Cir. 1980). US v. Walraven (1989), 892 F. 2d 972 (10th Cir. 1989). Appendix. Survey question wording If your local police agency decided to use LPR to check all passing vehi cles to see if any have been stolen, would you support this decision? CIRCLE ONE: An officer should be able use LPR technology in order to: Strongly Support Support Neutral Oppose Strongly Oppose … check all passing vehicles for parking violations and unpaid tickets. … check if the registered owners of all passing vehicles are wanted for a crime. … check if the registered owners of all passing vehicles are sex offenders. … investigate all vehicles passing or parking near important places or buildings to try to prevent terrorism. Strongly supportSupport Neutral Oppose Strongly Oppose … learn about the past activities of sex offenders? Strongly Support Support Neutral Oppose Strongly Oppose License Plate Recognition (LPR) technology may be used in many other w ays. Please tell us which other uses of LPR you would support by marking one box on each line below: LPR systems take a photograph of a vehicle’s license plate number, wh ich can then be linked to the vehicle’s registered owner. The system can also be set up to record the date, time and exact location of a vehicle at the moment the photograph is taken. Do you believe that this information should be considered private? Please circle one number along this range. Private Information (1) (2) (3) (4) (5) Not Private Information If the police decide to save the LPR data (license plate number, date/time, location of the vehicle), the police will be able to look at the saved data in the future. Please tell us which uses of saved LPR data y ou would support by marking one box on each line below. The police should be able to use saved LPR data in order to: … find the last location of a vehicle connected with a crime? … investigate all vehicles which drive around an important place or building? … learn about the past activities of a suspect who is under investigation for a crime? … learn about the past activities of a person suspected of terrorism? … learn about the activities of parents who don’t pay child support in order to force these parents to appear in court? 50 PIJPSM 37,1 About the authors Dr Linda M. Merola is an Associate Professor of Criminology, Law and Society and the Co-Director of the Criminal Justice Policy Research Program in the Center for Evidence-Based Crime Policy, George Mason University. Dr Linda M. Merola is the corresponding author and can be contacted at: [email protected] Cynthia Lum is an Associate Professor of Criminology, Law and Society and the Deputy Director of the Center for Evidence-Based Crime Policy, George Mason University. Breanne Cave is a PhD Candidate in the Department of Criminology, Law and Society at the George Mason University. Dr Julie Hibdon is an Assistant Professor of Criminology and Criminal Justice, Southern Illinois University, Carbondale, IL. To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints 51 Community support for LPR
Advances in technology continue to revolutionize policing in important ways. Three key advancements that are being used today are body-worn cameras, license plate readers, and gunshot detection syste
R ealizing the Potential of Technology in Policing A Multisite Study of the Social, Organizational, and Behavioral Aspects of Implementing Policing Technologies by Christopher S. Koper , George Mason University (PI) Cynthia Lum , George Mason University (PI) James J. Willis , George Mason University (Co -PI) Daniel J. Woods, Police Executive Research Forum Julie Hibdon, Southern Illinois University Supported by the National Institute of Justice (201 0-MU-MU-0019) Realizing the Potential of Technology in Policing 1 This study was supported by Nati onal Institute of Justice Grant # 2010 – MU – MU – 0019 The authors shown below used federal funds provided by the U.S. Department of Justice and prepared the following final report: Realizing the Potential of Technology in Policing A Multisite Study of the Social, Organizational, and Behavioral Aspects of Implementing Policing Technologies Christopher S. Koper, George Mason University (PI) Cynthia Lum, George Mason University (PI) James J. Willis, George Mason University (Co -PI) Daniel J. Woods, Police Executive Research Forum Julie Hibdon, Southern Illinois University December, 2015 This report updates an earlier version, dated “January 2015” Opinions or points of view expressed are those of the authors and do not necessarily reflect the official position or policies of the U.S. Department of Justice. Realizing the Potential of Technology in Policing 32 3. Key Technologies in Law Enforcement To better understand the impacts of technological changes in policing, we sought to examine the social, organizational, and behavioral implications of a range of relatively new and significant policing technologies that have diffused into law enforcement. Our intent was also to focus on technologies that are critical to primary police functions and central to evidence-based strategies and practices designed to reduce crime and/or enhance police legitimacy. To select these technologies, we reviewed academic and nonacademic research literature on police technology as well as other technology reports, guides, and needs assessments produced by government agencies and policing organizations (notably, NIJ, the Community Oriented Policing Services (COPS) Office, the International Association of Chiefs of Police, and PERF). 4 We examined the technologies featured in these studies and considered experts’ assessments of the impacts and importance of these technologies to policing. We also examined how commonly police use various technologies as reported in the Bureau of Justice Statistics’ Law Enforcement Management and Administrative Statistics (LEMAS) surveys and other surveys of police agencies (Burch, 2012; Hickman and Reaves, 2006a, 2006b; Koper et al., 2009; Lum et al., 2010; Reaves, 2010). In so doing, we sought to select technologies that are well developed and in relatively common use (with regard to the latter, we considered both current use and trends in the adoption of various technologies). Finally, we considered the existing evidence on evidence-based strategies to enhance police effectiveness and fairness (e.g., Braga, 2007; Eck and Weisburd, 2004; Lum et al., 2011; Skogan and Frydl, 2004) and identified technologies that have logical relevance to implementing or enhancing strategies and practices supported by policing research (e.g., Lum, 2010b). For instance, what technologies have the most potential to facilitate evidence-based practices such as hot spots policing and problem-oriented policing (Braga 2007; Lum et al.,2010; Skogan and Frydl 2004; Weisburd, Telep, Hinkle, and Eck, 2010)? Which have the most potential 4 Project staff examined the contents of 140 reports by government and policing organizations and reviewed several dozen academic and nonacademic works discussing theory or research on police and technology. Realizing the Potential of Technology in Policing 33 to improve police legitimacy by increasing transparency, accountability, and/or responsiveness to the community? Based on these assessments, the research team identified the following categories of police technologies as particularly central to everyday police work and successful practices:  Information technologies for the collection, management, and sharing of data;  Analytic technologies such as GIS and crime analysis;  Communications technologies including those related to dispatch (e.g., next generation 911 and computer-aided dispatch with GPS tracking of patrol cars) and those for disseminating information to personnel in the field (e.g., mobile computers and wireless access systems);  Surveillance and sensory technologies (e.g., CCTV networks, LPRs, and patrol car cameras); and  Identification technologies (e.g., DNA testing and other forensics equipment). From among these categories, we then selected the following specific technologies to aid us in understanding the impact of technology on law enforcement:  Information technologies (IT), defined broadly as intra- and interagency sy stems for managing, sharing, and analyzing data, including mobile computers and wireless access systems for sharing information with officers in the field;  Crime analysis, defined to include analytic processes and products of crime analysis as well as the mechanisms for disseminating results throughout the agency;  License plate readers (LPRs);  Patrol car video cameras; and  DNA testing technology. Note that while we cannot make any absolute claims that these technologies are the most important in law enforcement based on objective assessments, one can reasonably argue that these technologies are particularly worthy of study in view of prior research and theory, expert opinion, and usage patterns and trends. In the sections below, we discuss contemporary use of these technologies in policing and briefly review prior research on their impacts. Realizing the Potential of Technology in Policing 34 3.1 Information Technologies Information technologies (IT) within police agencies include a wide array of databases and data systems (and their supporting hardware and software) for storing, managing, retrieving, sharing, and analyzing information both within and across agencies. Common IT components in police agencies include records management systems (RMS) that capture criminal incident records, comput er-aided dispatch systems that record and assign calls for service, and various other databases that may contain information and/or intelligence on persons, groups, personnel, and other matters. Police agency websites used to exchange information with community members constitute another important part of police IT systems (Rosenbaum, Graziano, Stephens, and Schuck, 2011). Finally, our definition of IT also includes mobile computers and data terminals that give officers wireless access to information in the field and that allow them to file reports remotely. (Mobile computers may also be viewed as communication technologies.) Developments in IT have enhanced records management, data sharing, crime analysis, and performance management in police agencies in many ways over the last few decades. According to the 2007 LEMAS survey, half or more of local police departments and sheriffs’ offices use computers for records management, crime investigation, personnel records, information sharing, and dispatch (Burch, 2012: 15; Reaves, 2010: 22). Indeed, computers are now used for these functions in a majority of all but the smallest police agencies. Agencies also use computers to support functions like automated booking, fleet management, and resource allocation. As of 2003, the majority of police agencies maintained electronic data on incident reports, arrests, calls for service, stolen property, and traffic citations (Hickman and Reaves, 2006a: 31; 2006b: 31). Other data that agencies often maintain in electronic form include warrants, criminal histories, traffic accidents, and summonses. In addition, more than half of local agencies reported having in-field computers or terminals for their officers as of 2007 (Burch, 2012: 16; Reaves, 2010: 23). 5 More than 90% of local police departments serving populations of 25,000 or more now have such capability, as do more than 85% of s heriffs’ offices serving populations of at least 100,000. Agencies with in-field computers or terminals 5 Since 1990, there has been more than a 12-fold increase in the percentage of local police departments with in-field computers and terminals (Reaves, 2010: 24). Realizing the Potential of Technology in Policing 35 typically have 40– 50 such devices for every 100 officers. Most agencies use their in- field computers and terminals for writing reports, and a majority of agencies serving larger jurisdictions also use them for other communications. Information commonly accessible to officers through these computers and terminals, particularly in larger jurisdictions, include motor vehicle records, warrants, calls for service, criminal histories, protection orders, interagency information, the Internet, and, to a somewhat lesser extent, crime maps. The development of IT systems for sharing and analyzing data within and across agencies has also been emphasized in recent years. In many agencies, various types of records maintained by different units are now integrated and are easily accessible and searchable for officers, often remotely. Police have long had the ability to access national data systems like the Federal Bureau of =nvestigation’s (FB=) National Crime Information Center (NCIC). More recently, however, law enforcement practitioners have developed more extensive systems for sharing a wider variety of data across federal, state, and local agencies. Spurred in part by concerns over terrorism, the Department of Homeland Security has established fusion centers around the country (78 as of 2013) 6 to share information and intelligence among federal, state, and local agencies. Similarly, the Naval Criminal Investigative Service launched the LInX initiative in 2003 to promote more information sharing between law enforcement agencies at multiple levels. Currently, nine regional LInX systems involving over 760 partner agencies have been established across the United States. 7 T he FB=’s Law Enforcement National Data Exchange, or N-DEx, allows agencies to search and analyze data using powerful automated capabilities designed to identify links between people, places, and events . 8 In sum, current state- of-the-art systems provide many agencies with sophisticated capabilities for linking and querying databases within and across agencies. 9 For example, officers may query things like nicknames or see linkages of offenders, suspects, victims, and associates across multiple databases. As stated above, IT is arguably the technology with the most potential to impact policing, as it affects almost all aspects of police work and manageme nt. IT may enhance various dimensions of police efficiency and effectiveness, such as: the 6 See http://www.dhs.gov/fusion-center-locations-and-contact-information , accessed June 23, 2013. 7 See http://www.ncis.navy.mil/PI/LEIE/Pages/default.aspx , accessed June 22, 2013. 8 See http://www.fbi.gov/about-us/cjis/n-dex , accessed June 22, 2013. 9 A 2008 survey of agencies affiliated with the Police Executive Research Forum (PERF) suggests that most larger police agencies already have systems linking them to regional or national systems (Koper et al., 2009). Realizing the Potential of Technology in Policing 36 speed and accuracy of crime reporting; the amount of time officers spend in the field; the ability of officers to identify persons, vehicles, and places of interest (thus enhancing both reactive and proactive field work and improving officers’ ability to identify potential safety threats); the ability of detectives and officers to identify and locate suspects in criminal investigations; the capacity of managers to identify and respond to crime patterns and trends, monitor organizational performance, and assess the work and conduct of individual officers; the problem-solving capabilities of officers and managers; information exchange with the public; and the speed of ad ministrative processes (Groff and McEwen, 2008). These benefits might be offset to some degree, however, by technical difficulties and complexities in use of the IT systems, additional time and resources devoted to maintaining the systems and meeting reporting requirements, reduced interaction with citizens (i.e., officers may become more engrossed in working with technology and less engaged with people), and (as alluded to previously) the inability or disinterest of officers and managers to capitalize on the strategic uses of IT. Many police researchers have recognized the centrality of IT to police work and organizational change more generally (e.g., Boudreau and Robey, 2005; Chan, 2001, 2003; Ericson and Haggerty, 1997; Harris, 2007; Manning, 1992a; Mastrofski and Willis, 2010 ). Accordingly, it has been studied more extensively than other forms of police technology. Yet, this body of research has produced complex and often contradictory findings on =T’s impact. Some of the broadest assessments of the impact of IT on policing have come from studies of the federal Community Oriented Policing Services (COPS) program, which provided hundreds of millions of dollars in grants to state and local agencies for the acquisition of technologies during the 1990s. COPS grantees used much of their funding to obtain various forms of IT, including mobile and desktop computers (79% o f grantees had acquired funding for the former by 1998, making it the leading type of COPS-funded technology), computer-aided dispatch systems, booking and arraignment technologies, and telephone reporting systems (Roth et al., 2000). Although grantees reported substantial benefits from these grants, largely in the form of officer hours redeployed into the field (Koper et al., 2002; Koper and Roth, 20 00), studies of the COPS program have yielded mixed results as to whether the technology grants actually helped police reduce crime (e.g., U.S. Government Accountability Office, 2005; Zhao, Scheider, and Thurman, 2002, 2003 ). And even the most optimistic assessments suggest that the crime reduction benefits of the technology grants were less than those of grants for innovative programs and hiring Realizing the Potential of Technology in Policing 37 officers. 10 Hence, while technology may bring tangible benefits to police agencies, it doesn’t necessaril y provide a cost-effective alternative to additional officers or innovative strategies. Similarly, in a national study of large police agencies over the period of 1987- 2003, Garicano and Heaton (2010) found that increases in the application of IT were not associated with reductions in crime rates, increases in clearance rates, or other productivity measures (IT that facilitates better crime reporting actually generated the appearance of lower productivity). However, they also found evidence that IT was linked to improved productivity when complemented with organizational and managerial practices, like Compstat, that reflect more strategic uses of IT (see also Nunn, 2001 for related findings). Other studies, which have consisted largely of case studies and which examined a number of attitudinal and objective outcome measures, have also yielded mixed findings with respect to the effects of IT on officer productivity, case clearances, proactive policing, community policing, problem solving, and other outcomes, though officers have generally shown positive attitudes towards IT improvements (Agrawal, Rao, and Sanders, 2003; Brown, 2001; Brown and Brudney, 2004; Chan et al., 2001; Colvin, 2001; Danziger and Kraemer, 1985; Ioimo and Aronson, 2003, 2004; Nunn, 1994; Nunn and Quinet, 2002; Palys, Boyanowsky, and Dutton, 1984; Rocheleau, 1993; Zaworski, 2004). We examine many of the issues raised by these studies throughout our investigation. Note that we devote particular attention to IT in our case studies, given its centrality to policing and the myriad ways in which it can affect police organizations. Despite the mixed findings of prior research, we noted earlier that important innovations like hot spots policing and Compstat have been linked to advances in IT. Strategic use of =T capabilities by police are thus likely key to realizing =T’s full potential. One strategic use with demonstrated promise for improving the effectiveness of police is =T’s application to crime analysis, a form of analytical technology highlighted next. 10 An analysis by the U.S. Government Accountability Office suggests that each dollar spent on COPS grants for technology reduced index crimes by 17 per 100,000 persons (U.S. GAO, 2005). In contrast, each dollar spent on grants for hiring new officers or innovative community policing programs reduced index crimes by 29 and 88 per 100,000 persons, respectively. Realizing the Potential of Technology in Policing 38 3.2 Crime Analysis Crime analysis is the main analytic technology used by police today. As de scribed by Taylor and Boba (2011 : 6), “crime analysis involves the use of large amounts of data and modern technology —along with a set of systematic methods and techniques that identify patterns and relationships between crime data and other relevant information sources —to assist police in criminal apprehension, crime and disorder reduction, crime prevention, and evaluation.” While the collection of Uniform Crime Report (UCR) statistics and counts of crime might be considered an early stage of crime analysis, the activities and analyses that fall under the umbrella of “crime analysis” are wide ranging. Common duties for crime analysts involve assisting detectives, mapping crime, identifying crime patterns, conducting network analysis, and compiling data for UCR reporting and managerial meetings (Taylor and Boba, 2011). The development and adoption of crime analysis has been an important trend in policing over the last few decades. In a recent national survey, Taylor and Boba (2011) found that 57% of police agencies have staff whose primary responsibility is conducting crime analysis, and 89% of agencies have personnel w hose primary or secondary responsibility is conducting crime analysis. Similarly, the 2007 LEMAS survey showed that the use of computers for crime analysis is quite common, particularly among larger police agencies (Burch, 2012; Reaves, 2010). 11 This development of crime analysis has been facilitated by the improvement of police data systems and the development of computer software for specialized applications such as geographical and intelligence analyses. Indeed, Weisburd and Lum (2005) found that computerized crime mapping is an innovation that has spread widely in policing. The 2007 LEMAS found that more than 80% of local police departments serving populations of 50,000 or more use computers for crime analysis and crime mapping. The majority of these agencies also use computers for identification of hot spots (small areas of crime concentration). The majority of sheriffs ’ offices in jurisdictions of 100,000 or more people also use computers for 11 Am ong large police agencies (those with 100 or more officers), 78% had crime analysis personnel as of 2000, and 72% of those agencies had specialized crime analysis units (O’Shea and Nicholls, 2003). There is also an international organization of crime analysts (see http://www.iaca.net/index.asp) which provides training, conferences, and support in advancing the use of crime analysis in law enforcement. Realizing the Potential of Technology in Policing 39 crime analysis and crime mapping. Roughly half of sheriffs serving very large jurisdictions (500,000 or more) do hot spot identification. Crime analysis has great potential for improving the effectiveness of police. While it has perhaps been linked most prominently to hot spots policing and Compstat, crime analysis is also used heavily for investigative work and can be a valuable component of problem-oriented policing (see Taylor, Koper, and Woods, 2011a ). However, with the exception of its role in supporting hot spots policing, we are not aware of any evidence demonstrating a clear link between the use of crime analysis and lower rates of crime (Lum, 2013). Although this may reflect a lack of study (for example, we have seen no before-and-after assessments evaluating the impact of establishing crime analysis units), it is also likely that, as with other technological and analytical innovations, the potential impact of crime analysis is limited by outside factors. One such factor is that the sophistication of crime analysis capabilities and work varies considerably across agencies. Though dated, a survey conducted with larger police agencies (those having 100 or more officers) in 2000 found that crime analysis personnel in many agencies did not have sophisticated software applications, made limited or no use of databases from outside their agencies (e.g., non-law enforcement data or data from other law enforcement agencies), and/or conducted only simple (i.e., counting) forms of analysis (O’Shea and Nicholls, 2003). Important predictors of the range and sophistication of crime analysis include the availability of hardware and software, data collection capabilities, training, and structural characteristics such as whether an agency has a specialized crime analysis unit (O’Shea and Nicholls, 2003). At the same time, obstacles to effective use of crime analysis can lessen its impact. These may include a police culture that doesn’t value analytical work, the reactive nature of policing, and a disregard for crime analysis that is done largely by civilians (Lum, 2013; Taylor and Boba, 2011). In practice, officers may not use products like maps and may find them of little value in their work (Cope, 2004; Cordner and Biebel, 2005; Paulson, 2004). Indeed, crime analysis is largely produced for police managers, and while they tend to be its heaviest users (O’Shea and Nicholls, 2003; Taylor and Boba, 2011), they often focus largely on criminal ap prehension and tactical short-term planning rather than long-term strategic p lanning (:arris, 2007; O’Shea and Nicholls, 2003). Realizing the full potential of crime analysis requires more emphasis on long-term strategic planning, more attention to developing analytical products of value to officers, and proper training, Realizing the Potential of Technology in Policing 40 coaching, support, and reinforcement at all levels in the agency. Stronger management support and appreciation by target audiences have been shown empirically to have a positive impact on crime analysis functions and sophistication (O’Shea and Nicholls, 2003). 3.3 License Plate Readers License plate readers (LPRs) are high-speed camera and information systems that read vehicle license plates in real-time using optical character recognition technology. Plates are checked instantaneously against databases that may contain license plate information on stolen vehicles, vehicles linked to fugitives and criminal suspects, and other vehicles of interest (e.g., vehicles linked to sex offenders, parking violators, and drivers with suspended licenses). LPRs can be assigned to mobile patrol units or deployed at fixed locations. When an LPR finds a match, it sounds an alarm or provides another type of notification. While LPRs serve an important surveillance function, they can also be viewed as information technologies, as the data they collect can be stored, analyzed, and searched for in vestigative purposes. LPR technology has been used since the 1980s in Europe to prevent crimes from vehicle theft to terrorism (Gordon, 2006). LPR use is particularly extensive in t he United Kingdom; all police forces in England and Wales now have LPR capability (PA Consulting Group, 2006). In the United States, LPR use is growing rapidly. About a quarter of U.S. police agencies were using LPRs as of 2009 (Roberts and Casanova, 2012) , and more than a third of agencies with 100 or more officers were using them (Lum et al., 2010; also see Koper et al., 2009). Upwards of 50% of agencies having 500 or more officers used them (Roberts and Casanova, 2012), and many additional agencies were interested in acquiring them (Koper et al., 2009; Lum et al., 2010). Lum et al. (2010) have suggested that the diffusion of LPR has been quite rapid, even in comparison to other popular policing technologies such as computerized crime mapping (see Weisburd and Lum, 2005), in-field cameras, or forensic tools. At the same time, the vast majority of agencies using LPRs —86% according to one survey — had no more than 4 of the devices as of 2009 (Lum et al., 2010). This is likely due in p art to the cost, which generally runs from $20,000 to $25,000 per unit. LPR systems provide officers with the ability to scan and check hundreds of license plates in minutes, thereby automating a process that in the past was Realizing the Potential of Technology in Policing 41 conducted by officers manually, tag- by-tag, and with much discretion. As an information technology system, LPRs can collect and store large amounts of data (plates, dates, times, and locations of vehicles) for potential use in criminal investigations, homeland security operations, and other crime prevention efforts. Visible deployment of LPRs may also have some deterrent value. 12 Given these characteristics, LPR has the unique potential to improve police effectiveness. Although police have tended to use LPR primarily to reduce auto theft (Lum et al., 2010), they seem to be considering its use for a wider range of applications (Roberts and Casanova, 2012; Lum et al., 2010; PERF, 2012). Prior studies of LPR conducted in the United Kingdom and North America have focused largely on the accuracy and efficiency of the devices in scanning license plates and on their utility for increasing the number of arrests, recoveries of stolen vehicles, and seizure of other contraband (Cohen, Plecas and McCormack, 2007; Maryland State Highway Authority, 2005; Ohio State Highway Patrol, 2005; PA Consulting Group, 2003; Patch, 2005; Taylor, Koper, and Woods, 2011b, 2012). H owever, the studies found limited evidence on whether LPR use actually reduces crime. Studies of LPR use and its effects on crime have tested small-scale deployment of LPRs with patrol units. One study that spanned two suburban jurisdictions in Virginia found that 30-minute LPR patrols conducted once every few days (on average) in selected crime hot spots for a period of two to three months did not reduce auto-related or other forms of crime in the targeted locations (Lum et al., 2010, 2011). In contrast, a study conducted in Mesa, Arizona, found that short- term deployment of an LPR team (using four of the devices) to high-crime street segments produced reductions in drug offenses at those locations that lasted for several weeks beyond the intervention (Koper et al., 2013; also see Taylor, Koper, and Woods, 2012). Other findings from that study suggested that LPR deployment might also help to reduce auto theft and personal offenses at hot spots, depending on exactly how officers use the devices. Both studies were limited, however, by the short duration or low dosage of the intervention, the small numbers of LPRs available, and the limited data fed into the LPR devices (the data consisted largely or entirely of manually downloaded information on stolen vehicles and license plates). Updated studies are needed to examine larger-scale LPR deployments and LPR operations conducted with access to more extensive data systems. 12 For discussions of the deterrent value of surveillance cameras more generally, see Welsh and Farrington (2008) and LaVigne et al. (2011). Realizing the Potential of Technology in Policing 42 Further assessment is also neede d of other ways that police might use LPRs. For example, data collected by LPR units have been used to identify vehicles (and thus suspects) that were near a crime scene at a given time or to determine the whereabouts, and/or confirm the alibi, of potential suspects or witnesses. In major crises, LPR data can be used to recreate vehicular movement around high-risk locations. Some agencies have also used LPR to scan and record all vehicles in and around a crime scene shortly after a crime occurr ed. In terms of our study, we are particularly interested in how LPR affects not only efficiencies related to investigative activities and case clearances, but also how this technology changes the way in which officers patrol their beats or detectives investigate cases . Police adoption of LPR also has implications for community perceptions of police legitimacy insofar as it raises issues of surveillance and privacy. In their study of LPR use in Virginia, Lum et al. (2010) surveyed community residents in one of the study jurisdictions and found that while there was strong support for LPR use in general, this support varied depending on the types of LPR applications under consideration (e.g., using the devices to detect stolen automobiles received much more community support than using them to detect parking violators). Survey results also suggested that citizens prefer to have some external controls (e.g., court orders or consultation with attorneys or the community) on police storage and use of LPR data (see Merola and Lum, 2013; Merola, Lum, Cave, and Hibdon, forthcoming). Finally, it remains to be seen how officers and agencies will adapt to LPR as its use expands. For example, do officers like using LPR technology and how does it affect the way they conduct everyday patrol and other activities? Does it increase their job satisfaction or personal motivation? Does it prompt them to be more proactive and strategic in their actions? And how do supervisors assign and monitor LPR deployment and use for its fullest effect? 3.4 In-Car Video Cameras In -car video (ICV) systems are devices used to create video and audio records of selected events and encounters experienced by officers. The cameras are mounted within the patrol vehicle, and officers wear a wireless microphone that transmits audio signals to the system. The devices are typically activated Realizing the Potential of Technology in Policing 43 automatically when officers put on their flashing lights or exceed a certain speed. Officers can also activate them manually. ICV systems serve a number of purp oses (e.g., see Maghan, O’Reilly, and 😮 Shon, 2002; Schultz, 2008). Most notably, they can be used to monitor the legality and professionalism of officer conduct in various contexts. In this way, ICV systems can help guard against excessive use of force, illegal searches, racial profiling, and other forms of illegal, unprofessional, or abusive behavior by officers. Indeed, some agencies have adopted ICV systems in the wake of controversial use of force cases or in response to accusations of other problematic conduct by officers such as racial profiling (Maghan et al., 2002). At the same time, ICV systems also protect officers from false allegations of unlawful or unprofessional conduct, and there have been many accounts of ICV systems exonerating officers in court cases and misconduct investigations. Further, ICV systems can provide evidence for police and prosecutors in certain types of criminal cases (e.g., cases involving driving under the influence or assaults on officers). Recordings from ICV systems can also be valuable in training officers about professionalism, safety, lawful searches, and other issues. ICV systems have been in use since at least the 1990s (Maghan et al., 2002), and their use has grown considerably since that time. As of 2007, roughly two thirds of local police agencies reported using cameras in their patrol cars (Burch, 2012: 15; Reaves, 2010: 21). Use of these systems is common among agencies of all sizes, though the largest agencies are somewhat less likely to use them, due likely to the expense of equipping their large automobile fleets. 13 Overall, local police agencies reported having nearly 100,000 cars equipped with cameras in 2007, which amounted to about a quarter of all cars they operated (calculated from Burch, 2012 and Reaves, 2010). Further, in a 2008 survey of agencies affiliated with PERF, nearly all agencies using car cameras found them to be effective, and almost half reported no significant challenges to their use (Koper et al., 2009). The main challenges agencies did id entify, noted by 25% of users, were “economic and political.” With respect to political challenges, agencies may face the greatest obstacles from within their agencies. Anecdotal accounts suggest that officers often resist ICV technology out of concern th at managers will use it to “spy” on them and overly scrutinize their behavior (Maghan et al., 2002). Training on the potential benefits of ICV systems to officers may help overcome this resistance, as may policies about how (and for how long) the videos will be saved and the circumstances under which 13 Only 38% of agencies serving populations of 1 million or more reported using ICVs in 2007, as did slightly less than half of agencies serving populations of 250,000-499,999 (Reaves, 2010: 21). Realizing the Potential of Technology in Policing 44 they will be used by supervisors. The fact that the cameras are typically activated only in certain types of situations also means that officers need not feel that they are under continuous surveillance. ICV systems would seem to have much potential for affecting police- community interactions and community perceptions of police fairness and legitimacy. Both police and citizens can be expected to regulate their behavior more carefully when they know that they are being recorded by ICV systems, thus potentially preventing or diffusing volatile encounters. In places where police use this technology, community members can have greater assurance that police will be held accountable for misconduct, and they may be better informed about the veracity of complaints made against the police when cases get publicized. Yet beyond anecdotal accounts (e.g., Maghan et al., 2002), there has been little, if any , systematic research on how ICV systems affect outcomes such as complaints against the police, community views of the police, use of excessive force, and the like. Nor has there been research on how, if at all, ICV systems affect the ability of police to reduce crime. One could speculate, for instance, that ICV systems might influence the inclination of police —one way or the other —to engage in more intensive traffic enforcement or order maintenance policing. On the one hand, officers might feel inhibited by ICV systems; on the other hand, they might feel more protected against complaints. Officers in the field may also devise ways to use ICV systems for different forms of surveillance, though this might sometimes raise legal issues, depending on local eavesdropping laws (Maghan et al., 2002), and/or raise public concerns about intrusive surveillance and privacy. As the technology improves, police will also likely have more options for transmitting recordings from ICV systems and for integrating these systems with LPRs and facial recognition systems (Maghan et al., 2002). 3.5 DNA Testing Law enforcement agencies use a variety of forensics technologies to assist them in the identification of criminal offenders. One of the most important enhancements to these capabilities in recent decades has been the development of identification tests using deoxyribonucleic acid, commonly known as DNA. DNA tests identify unique individual genetic codes from DNA samples that are extracted from biological evidence such as blood, semen, hair, and saliva. Developed in the 1980s, Realizing the Potential of Technology in Policing 45 DNA testing has become a common method of identification, particularly for sex crimes and other violent offenses, and it is widely viewed as the state of the art in offender identification (National Research Council, 2009). In the United States, DNA testing is mostly used in violent crime cases due to its expense, but its use for property crimes is also expanding (Roman et al., 2008). Police may collect and use DNA evidence in a number of ways. They may use D NA testing to determine whether a particular suspect can be linked to physical evidence from a particular crime scene. They may use recovered DNA evidence from a crime scene to identify suspects, though it seems that many agencies do not understand or take advantage of this potential DNA application (Strom et al., 2009). Fin ally, police and other criminal justice agencies take DNA samples from convicted offenders and in some states from arrestees to test them for matches to evidence from unsolved crimes and for use in future investigations. The DNA Identification Act of 1994 authorized the FBI to establish a national DNA database with indexes for persons convicted of crimes, missing persons (and relatives of missing persons), samples recovered from crime scenes, and samples recovered from unidentified human remains (Roman et al., 2008: 13-14). This national database is combined with state and local DNA databases in a system named CODIS (for the Combined DNA Index System). By the late 1990s, all 50 states had passed legislation requiring convicted offenders to provide DNA samples (Samuels, Davies, and Pope, 2013; Schwabe, 1999). As of 2009, 47 states collected DNA samples from all convicted felons and 37 collected samples from those convicted of certain misdemeanors (DNA Resource, 2009, as cited in Wilson, Weisburd, and McClure, 2011: 8). In addition, 28 states have laws authorizing the collection of DNA evidence from all or subsets of felony arrestees (and sometimes from misdemeanor arrestees) prior to conviction (Samuels et al., 2013). The collection of DNA from arrestees has expanded considerably since 2005 following federal legislation allowing for such information to be uploaded into CODIS. 14 Nearly 10.4 million DNA profiles were in CODIS as of 2011, up from 1.2 million in 2002 (Samuels et al., 2013: 4). Although the submission of DNA from arrestees has been interrupted in some states by recent court cases challenging the constitutionality of this procedure, the United States Supreme Court upheld the practice in the case of Maryland v. King, which was decided in June 2013. 14 State laws provide for expunging this evidence if the arrestees are not convicted, but many states leave the burden of initiating these procedures on the arrestees (Samuels et al., 2013). Realizing the Potential of Technology in Policing 46 According to a recent survey, only 8% of local agencies have a local lab to conduct DNA testing, 88% send evidence to state labs for testing, and the remaining agen cies use federal, private, or other types of labs (Strom et al., 2009: 3-12). :owever, many of the nation’s largest agencies (which are responsible for large numbers of cases) have their own crime labs (counted above as local labs) and may thus have their own DNA testing capabilities. In principle, greater use of DNA evidence should help police solve a greater number of crimes and improve the likelihood of convictions in those cases. This, in turn, should reduce crime through incapacitation of offenders and potentially through deterrence of those who have had their DNA taken (but see Bhati, 2010 for mixed assessments on the latter point). Further, DNA testing may be particularly helpful in identifying the most active repeat offenders who commit disproportionate numbers of crimes. Evidence on how DNA testing impacts police performance and crime is rather limited (Wilson et al., 2011). However, a randomized experiment involving five jurisdictions in the United States found that the use of DNA evidence greatly enhanced outcomes in property crime cases, namely, residential and commercial burglaries and thefts from automobiles (Roman et al., 2008). Compared to traditional investigations, cases involving the use of DNA evidence resulted in twice as many suspects being identified, twice as many suspects being arrested, and more tha n twice as many cases being accepted for prosecution. Compared to the use of fingerprints, the use of DNA was also at least five times more likely to result in the identification of a suspect. Moreover, suspects identified through DNA evidence tended to be more serious offenders; overall, they had at least twice as many felony arrests and convictions as did suspects identified in other cases. 15 Similarly, a study examining criminal cases in New South Wales, Australia, f rom 1995 through 2007 found that the expansion of a DNA database for imprisoned offenders started in 2001 led to increases in case clearances and cases resulting in charges for sexual assault, robbery, and burglary (Dunsmuir, Tran, and Weatherburn, 2008). However, these outcomes did not improve for assaults and motor vehicle crimes, nor did the development of the DNA database improve conviction rates for any of the offenses studied. A few other studies have also reported improvements in 15 These findings are also consistent with evidence from the United Kingdom, where there has been a national program to expand the use of DNA evidence in property crimes. Research there indicates that the suspect identification rate in burglary cases with DNA evidence is 41% as compared to 16% in other cases (Home Office, 2005, cited in Roman et al., 2008: 7). Realizing the Potential of Technology in Policing 47 case outcomes stemming from the use of DNA evidence, but methodological weaknesses in these studies preclude definitive conclusions (see review in Wilson et al., 2011). Moreover, no studies have yet examined the impact of DNA testing on crime rates. Expanding the use of DNA evidence also raises a number of organizational issues for police agencies and crime labs with respect to equipment and staffing needs and the establishment of DNA testing policies and procedures (e.g., Samuels et al., 2013). Expanded DNA use is adding to already substantial backlogs of cases wi th untested forensics evidence. In a 2007 survey, police agencies in the United States reported that they had handled 31,570 homicide and rape cases and over five million property cases with unanalyzed forensics evidence over the previous five years (Strom et al., 2009); roughly 40% of the homicide and rape cases in question had unanalyzed DNA evidence. Yet that report also showed that many cases went unanalyzed because police had not identified suspects in the cases. This suggests that many agencies are missing out on the potential of DNA testing to help identify leads in criminal cases. Hence, additional training and policy changes will be required for agencies to fully capitalize on the potential of DNA testing technology. Problems with resources and backlogs may also ease somewhat as DNA testing procedures improve, reducing the time and cost of DNA tests. For example, although they do not yet appear to be in common use, portable devices for the collection and testing of DNA evidence have been developed that may alleviate backlogs in DNA testing and greatly reduce the cost of such tests (Nunn, 2001). How DNA testing might affect other aspects of police work and organizations (e.g., the everyday activities and decisions of police officers and managers) has received little attention to date. As noted by Bayley and Nixon (2010), for instance, DNA evidence allows a greater number of cases to be solved without witnesses or confessions. This could substantially change the nature of detective work and potentially reduce the reliance of the police on community cooperation (which is likely to have pros and cons) in investigating crimes. There is also the issue of how DNA testing might affect perceptions of police fairness and legitimacy, particularly in minority communities that are likely to be disproportionately impacted by expanded DNA collection. On the one hand, DNA offers the possibility of exonerating defendants who have been wrongly accused or convicted. On the other hand, might DNA arrest policies lead to greater use of pretextual arrests as an excuse to collect DNA from suspects, a charge that has been leveled in the United Kingdom (Stanglin, 2009)? At the same time, public Realizing the Potential of Technology in Policing 48 perceptions might put greater pressure on police to collect DNA in a wider range of ca ses if people come to expect the availability of DNA evidence as the norm in proving criminal cases (what is often referred to as the “CS= effect” 16 ). It remains to be seen how and to what degree these considerations will affect police agencies. 3.6 Summa ry These five technologies —information technology systems, crime analysis, LPRs, in-car video, and DNA analysis —are major technologies in use by many police agencies today. They reflect common types of technology used in policing more generally (i.e., informational, analytic, communications, surveillance, and forensics technologies) and could potentially have a number of intended and unintended effects in policing. In our study, we used these technologies as a starting point to prompt personnel in four law enforcement agencies to think about the role, function, and impacts of technology on their organizations and their daily lives and activities. By asking about specific types of technologies and their impacts on various aspects of the police agency, we were able to gain a stronger understanding of technology’s impacts on law enforcement more generally. =n the next section, we describe our approach before providing the results of the various studies we conducted. 16 This phrase was derived based on a popular television series dramatizing the work of forensic- specialist crime scene investigators (CSIs ).
Advances in technology continue to revolutionize policing in important ways. Three key advancements that are being used today are body-worn cameras, license plate readers, and gunshot detection syste
Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=gpas20 Policing and Society An International Journal of Research and Policy ISSN: 1043-9463 (Print) 1477-2728 (Online) Journal homepage: https://www.tandfonline.com/loi/gpas20 Strategic policing philosophy and the acquisition of technology: findings from a nationally representative survey of law enforcement Joshua A. Hendrix, Travis Taniguchi, Kevin J. Strom, Brian Aagaard & Nicole Johnson To cite this article: Joshua A. Hendrix, Travis Taniguchi, Kevin J. Strom, Brian Aagaard & Nicole Johnson (2019) Strategic policing philosophy and the acquisition of technology: findings from a nationally representative survey of law enforcement, Policing and Society, 29:6, 727-743, DOI: 10.1080/10439463.2017.1322966 To link to this article: https://doi.org/10.1080/10439463.2017.1322966 Published online: 07 May 2017.Submit your article to this journal Article views: 778View related articles View Crossmark dataCiting articles: 2 View citing articles Strategic policing philosophy and the acquisition of technology: findings from a nationally representative survey of law enforcement Joshua A. Hendrix, Travis Taniguchi, Kevin J. Strom, Brian Aagaard and Nicole Johnson Policing, Security, and Investigative Science, Center for Justice, Safety, and Resilience, RTI International, Research Triangle Park, NC, USA ABSTRACTPolice departments that emphasise certain strategic models (e.g. community-oriented policing, problem-oriented policing) may adopt specific types of technology to better achieve their core missions. A contrasting theory is that police agencies do not invest strategically in technology; rather, they adopt technology in a‘black box ’without a larger plan for how a particular technology fits within the agency’ s guiding philosophy or operational goals. Despite the importance of this discourse, very little research has been conducted to address these claims. Using survey data from a large and nationally representative sample of police agencies in the United States ( N= 749), we examine whether strategic police goals are associated with technology use for six core technologies (crime mapping, social media, data mining software, car cameras, license plate readers (LPRs), and body-worn cameras (BWCs)). Nationally, across the sample of all US law enforcement agencies, we find little relationship between strategic goals and technology. Agency size, rather than policing philosophy was a more important determinant of technology use. However, stronger relationships between strategy and technology emerged when the analysis was limited to a subsample of larger agencies (250 or more sworn officers). Specifically, community and hot spot policing strategies were positively associated with the use of geographic information system technology, social media, and LPRs. Agencies who emphasised hot spot policing were also more likely to have used BWCs. Implications of these findings are discussed. ARTICLE HISTORYReceived 29 September 2016 Accepted 20 April 2017 KEYWORDSLaw enforcement; technology; policing strategy Advancements in computers and communication tools over the last several decades have made numerous technologies available to law enforcement that were virtually unheard of even a few decades ago. Many departments are implementing technology to increase efficiency and improve outcomes, especially in times of diminished resources and enhanced public scrutiny of law enforce- ment tactics. Despite the accelerating use of technology in policing and speculative connections between technology and policing tactics, it is not well understood how technology is selected by police agencies. It is also unclear as to whether the acquisition of technology among law enforcement is calculated and rationally driven or based on a constellation of ad hoc factors that do not necessarily relate to the fit between an agency ’s goals and the technology it implements. Importantly, police agencies vary in philosophy, culture, management strategies, and goals (Weiss 1997), yet there is limited information on the extent to which the devotion to particular strategic models is linked to the use of particular technological devices. © 2017 Informa UK Limited, trading as Taylor & Francis Group CONTACT Joshua A. Hendrix [email protected] POLICING AND SOCIETY 2019, VOL. 29, NO. 6, 727 –743 https://doi.org/10.1080/10439463.2017.1322966 Given that technology can have a dramatic impact on how policing is done, how successful com- munity relations are, and to what extent public safety is ensured, it is imperative that police execu- tives, elected and appointed officials, and policy-makers have sound empirical evidence regarding the factors that affect technology acquisition. We drew upon data from a nationally representative survey of police agencies ( N= 749) to address the following research question: To what extent are different technological advancements associated with policing strategies that are designed and implemented to control crime? Our analysis focused on six technologies: crime mapping, social media, data mining software, car cameras, license plate readers (LPRs), and body-worn cameras (BWCs). Theoretical framework Technological advances in recent years have changed the nature of policing so significantly that many methods from just a decade ago have become antiquated and incompatible with current tech- nology (Goodison et al.2015 ). Some of these advances include location-monitoring devices for the tracking of high-rate offenders, predictive analytics and crime mapping software for the deployment of officers into crime hotspots, crime scene technology that enhances the collection and processing of evidence, and interoperable Web-based and other communication devices that facilitate connec- tions between police and the communities they serve. Although these advances have enhanced police capabilities (Danziger and Kraemer 1985, Roth et al.2000 , Ioimo and Aronson 2004, Roman et al .2008 ), it is not clear that they have made police more effective (Sherman and Eck 2002, Lum 2010 ). For example, despite advances in DNA technology and computer databases for handling for- ensic data, clearance rates for violent and property crime have remained relatively stable since the mid-1990s (Federal Bureau of Investigation 1996,2011 ). Regardless, it is indisputable that technologi- cal innovations have created seismic shifts in law enforcement tactics and can have a major effect on what police do on a day-to-day basis (Harris 2007). It is therefore important to understand key factors that influence the adoption of police technology. Generally speaking, the processes by which law enforcement technology is acquired are not well understood. One commonly invoked theoretical perspective for understanding technology acqui- sition within organisations is the diffusion of innovations model, which classifies adopters of technol- ogy as innovators, early adopters, early majority, late majority, and laggards (Rogers 1962). Although this taxonomy has intuitive appeal, it is limited in its ability to describe how technologies are acquired by police departments. For example, the categories are not mutually exclusive in practice. Police agencies do not easily fit into one subgroup when considering a specific type of technology, let alone across different types of technology. An agency could be considered both an early adopter and a laggard when it comes to geographic information systems (GIS) technology if mapping is done at an aggregate level but without incident-based geocoding. Additionally, the same agency may be a laggard in regard to LPR usage but an innovator in the use of BWCs. Thus, a more comprehensive conceptual framework is needed to understand whether strategic models are related to technology acquisition. The existing literature on organisational choice provides a useful starting point and an overarching theoretical framework for the present study, as it describes four perspectives for understanding how organisations identify and achieve agency goals. The rationalperspective posits that organisations behave rationally by identifying official goals, designing strategies to accomplish those goals, and then implementing technology that supports the strategies they have designed (Cyert and March 1963 , Simon 1997). The contingency perspective suggests that organisations operate in particular environments and key decisions depend on external factors and events (Lawrence and Lorsch 1967 ). The institutional perspective argues that organisations have their own interests as well –sur- vival, status and prestige, maximising resources, and protection from threats (Scott 2008). Finally, the entropic perspective depicts organisations more as anarchies than as well-oiled machines, often iden- tifying solutions before they have a specific problem demanding to be solved (Cohen et al.1972 ). 728 J. A. HENDRIX ET AL. According to this perspective, organisational options (such as technology) are frequently just lying around waiting for an opportunity to be adopted.The rational and entropic perspectives are most relevant for the present study, as they are diame- trically opposed in the ways they view the process by which technology is adopted. Whereas the rational perspective would anticipate connections between strategy and technology, the entropic perspective would assume that decisions regarding technology are made ad hoc, independently of strategic goals. Our study tested these competing viewpoints. Below, we describe prominent poli- cing strategy models, summarise what is known regarding factors that predict technology acqui- sition, and anticipate connections between strategic models and technology according to rational and entropic perspectives. Background and literature review Police strategy Significant debate exists among practitioners and researchers regarding the labels used to identify policing strategy models (Moore and Trojanowicz 1988, Weisburd and Braga 2006). This discourse has suggested the existence of at least eight strategic models that include professional, community, problem-oriented, intelligence-led, hot spot, offender targeting, predictive, and broken windows/ zero-tolerance policing. Although strategies are not mutually exclusive, each one emphasises differ- ent activities or objectives (e.g. improving police-community relations), which in turn might lead to an emphasis on different types of technology (e.g. intelligent use of social media). The professional policing model emphasises rigid hierarchical organisational structures, limits the use of discretion, and prioritises efficient response times. Community policingpromotes organis- ational strategies, including the systematic use of partnerships and problem-solving techniques, to proactively address the immediate conditions that give rise to public safety issues. The problem- oriented policing model subjects discrete pieces of police business to microscopic examination in hopes that what is learned about each problem will facilitate discovery of more effective strategies for dealing with it. Intelligence-led policing is a business model and managerial philosophy in which data analysis and crime intelligence are pivotal to an objective decision-making framework that facili- tates crime and problem reduction, disruption, and prevention through strategic management and enforcement strategies that target serious offenders. The hot spot policingmodel prioritises the identification and targeting of specific locations that generate the most calls for police service. Offen- der targeting policing emphasises the importance of identifying and prioritising repeat offenders. Pre- dictive policing uses predictive analytics and other techniques to pinpoint specific geographic locations most susceptible to crime. Finally, broken windowsorzero-tolerance policing is based on Wilson and Kelling ’s(1982 ) seminal article suggesting that targeting minor forms of social and phys- ical disorder will reduce more serious crime. Technology acquisition As noted, the knowledge base pertaining to factors that influence police departments to acquire and implement technology is underdeveloped. Som e evidence suggests that the size and location of an agency affects the number and types of technology it acquires (Mamalian et al.1999 , Chamard 2002,2003 ,2006 ). The impact of other agency- or conte xtual-level factors on technology acquisition is less understood. According to Schuck ( 2015), the adoption of technology can be understood as a complex interact ion between characteristics of the technology, organisational culture, and features of the larger social-structu ral environment. Using data from the Law Enforce- ment Management and Administrative Statistics, Schuck examined factors that could explain why agencies adopt dash and mobile cameras, including characteristics of the technology (design, func- tionality, and congruency with agency goals), organisational traits (hierarchical structure, POLICING AND SOCIETY 729 formalisation, spatial differentiation), characteristics of the community (income and demographic composition), and features of the local political environment. While the strongest predictor of mobile camera adoption in large agencies was local c rime, organisational size and spatial differen- tiation (sprawl) were positively associated with mobile camera adoption in smaller and medium- sized agencies. A study by Weisburd and Lum ( 2005) found that the adoption of some types of technology, such as computerised crime mapping, was related to the ‘cosmopolitanness’of the agency. That is, early adopters had a tendency to employ officers with more sophisticated levels of knowledge of research related to crime mapping, GIS, and hot spot policing. Skogan and Hartnett ( 2005) found that technol- ogy adoption and use in law enforcement are largely independent of one another and motivated by separate processes. Whereas the most important predictors of adoption were whether the agency was involved in ‘cosmopolitan networks ’, experience levels for using enforcement databases, and the degree of human capital in the organisation, key factors predicting usage were organisational resources and experience levels for using the system. Leong and Chan ( 2014) found that number of employees, the extent to which the department used digital information, and number of hours of training spent on new recruits were the key predictors of the adoption of web-based mapping. Finally, Randol ( 2014) reported that vertical height (i.e. number of vertical ranks and other aspects of complexity in the organisation ’s command structure), functional differentiation (i.e. number of specialised units with personnel dedicated to their functions), formalisation (i.e. number of written policies and procedures), and community policing were all associated with the likelihood of adopting crime analysis technology. Because of the limited research on technology a cquisition specifically, we also look to other areas of change experienced by law enforcement agencies. Some research has assessed the influ- ence of organisational and contextual factors of police departments on adoption of new policing strategies (Katz 2001,Morabito2008, Darroch and Mazerolle 2013). Examining whether insti- tutional processes, including mimesis (i.e. emula tion of other agencies), publication, and profes- sionalisation influence the adoption of intelligence-led policing, Carter ( 2016)foundthat institutional pressures, along with perceived crime control benefits had a positive effect on the adoption of intelligence-led policing. Similarly, Giblin ( 2006) examined the role of mimesis, exter- nal funding, and national accreditation as institu tional pressures on the formation of crime analy- sis units, finding that the most likely predictors of agencies forming a crime analysis unit were agency size and accreditation. Willis et al.(2007) explored agency adoption of COMPSTAT through the lens of the rational versus institution al organisational perspectives, and found that institutional pressures to ‘appear progressive and successful ’influenced agencies to adopt COMP- STAT more so than technical considerations. Ot her scholars have also examined community and macro-contextual influences on police strateg y implementation and organisational change (Bayerl et al.2013 , Darroch and Mazerolle 2015). Theoretical connections between strategy and technology Despite the frequency of discussions within the criminal justice arena pertaining to policing models and their impact on law enforcement activities (Moore and Trojanowicz 1988, Weisburd and Braga 2006 ), our review finds virtually no research that speaks to associations between different orien- tations towards common policing strategies and the acquisition of technology. This research ques- tion is important, given that it could provide context to previous studies suggesting that the increasing use of innovative policing technology has not necessarily led to more effective outcomes (Lum 2010). Specifically, if agencies are adopting technology without a strategic understanding of how that technology integrates within their overarching goals, acquisitions may be limited in their impact on key agency outcomes. As noted, relationships between technology and strategy can be anticipated by viewing them through the lens of the rational and entropic organisational perspectives. 730 J. A. HENDRIX ET AL. The rational lens The rational organisational perspective views organisations as calculating entities that make strategic decisions to complement the global goals of the organisation. This viewpoint anticipates that police departments will implement technologies that directly support and facilitate their respective over- arching strategic missions. Although we were unable to identify past research that has directly tested this assertion, it is logical to believe that an agency’s devotion to a particular model would influence the types of technology it uses for achieving certain outcomes. Technology could make a new strategy possible, provide a new tool for an existing strategy, or allow for a combination of both scenarios. For instance, the 911 system has been described as a major force that has shaped and reinforced reactive policing (Sparrow et al.1990 , Skogan and Frydl 2004). Drawing from the technology literature and using the rational organisational lens, we anticipate several connections between police strategy and technology. First, we expect relationships between crime mapping software and intelligence-led, hot spot, predictive, and community policing. The soft- ware ’s ability to identify geographic locations where crimes cluster and the potential for this infor- mation to inform crime reduction and prevention (Sherman and Weisburd 1995, Mazerolle et al. 1997 , Braga et al.1999 , Mamalian et al.1999 , Braga and Bond 2008, Braga et al.2012 ) lend themselves to both intelligence-led and hot spot policing models. The use of advanced mapping techniques, such as risk terrain modelling, may help departments to anticipate where future crimes will occur and to inform deployment decisions (Caplan et al.2011 ); therefore, this technology may also be com- patible with the predictive policing model. Likewise, crime mapping has been described as central for community policing, as it allows departments to produce maps that can be accessed by citizens to increase their awareness of local threats to public safety (Dunworth et al.2001 , Randol 2014). Second, we hypothesise connections between social media and community, predictive, and intel- ligence-led policing models. Although rarely examined empirically, the available information suggests that social media is perceived to enhance departments ’abilities to interact positively with the community (Burger 2013); thus, we see connections with the community policing model. With a vested interest in community outreach, departments can use social media to post information about suspects, crime prevention, or other public safety issues, and active social media use can huma- nise officers and ultimately enhance police-community trust (Stevens 2010). Recent surveys also find social media to be helpful for investigating crimes and for anticipating future crimes, which lends itself to intelligence-led and predictive policing models. In a 2013 survey of agencies by the Inter- national Association of Chiefs of Police (IACP), 80% of the sample reported that social media was a valuable investigative tool and helped them to solve crimes (IACP 2014). A 2014 survey by LexisNexis revealed that 67% of respondents perceived social media to be an effective investigative tool and platform for anticipating future crimes (LexisNexis 2014). Third, we anticipate relationships between data mining and intelligence-led and predictive poli- cing. Intelligence-led policing emphasises the use of data to facilitate crime reduction and preven- tion. Similarly, the predictive policing strategy is based on the logic that future crimes can be better anticipated, responded to, or prevented using intelligence collected from a variety of data sources. Since the 9/11 terrorist attacks, law enforcement agencies have been under pressure to become more data-driven in their daily operations. These data can take numerous forms; for example, they may derive from the agency ’s records management system, census databases, mobile resources (e.g. smartphones), LPRs, or social media. Data mining technology was designed to address needs related to handling prolific quantities of data from diverse sources (Fayyad and Uthurusamy 2002). Analysts may use data mining software to mine text data, visualise crime net- works, identify possible suspects, or recognise crime patterns and characteristics associated with them to guide the deployment of officers (Chau et al.2002, Hauck et al.2002 , Pearsall 2010). The potential for data mining software to uncover underlying causes of crime trends and patterns that can then inform the allocation of police resources as a crime prevention strategy is therefore consist- ent with the basic premises of intelligence-led and predictive policing. POLICING AND SOCIETY 731 Fourth, we expect to find relationships between the use of car cameras and community policing. Although the diffusion of dash cameras throughout American law enforcement was initially a conse- quence of increased attention on drinking and driving in the 1980s, allegations of racial profiling against the police, and demands for greater officer safety (Westphal 2004), research has documented the potential for car cameras to improve community relations. For example, in 2004, the IACP studied the use of in-car camera systems among 47 state police agencies that had received funding under the In-Car Camera Incentive Program in the late 1990s. Results indicated that officers perceived numer- ous benefits of the in-car camera systems, including improved community perceptions, agency accountability, and enhanced professionalism. Interviews with patrol officers suggested that in-car cameras also augmented officer safety and facilitated more harmonious relationships with the com- munity, as the presence of a camera can de-escalate confrontational situations when citizens are informed of being recorded (IACP 2004). Fifth, we anticipate a relationship between LPRs and intelligence-led and offender targeting models of policing. Automatic LPRs are high-speed cameras paired with character recognition soft- ware that can document thousands of license plates per minute while also recording the date, time, and geographic location of every scan. Some police departments, such as those in New York and Sacramento, have reported increases in arrests or reductions in reports for auto thefts as a function of implementing LPRs (see Hsu 2014). Additionally, 68% of agencies from a study by Roberts and Casanova ( 2012) reported that LPRs had enabled them to increase stolen vehicle recoveries, and 55% reported that auto-theft-related arrests had increased. 1The use of LPRs for vehicle recovery and other purposes seems most compatible with the intelligence-led and offender targeting models of policing, given that it is a form of innovative data collection that can help to prevent or resolve problems in the community and also to target high-risk offenders. Finally, we may expect a relationship between the use of BWCs and the community policing model. Multiple studies have indicated that the use of BWCs can help to reduce the number of com- plaints filed against local police departments, which may ultimately enhance police trust of the com- munity they serve. For example, the Rialto, California, police department found that shifts where BWCs were not deployed had more than twice as many use-of-force incidents as shifts which used them, and complaints against the police had decreased from 24 complaints filed during the 12 months before the study to 3 during the 12 months of the study (Barak et al.2014 ). In a 2015 study of the Phoenix, Arizona, police department ’s use of BWCs, complaints dropped sharply (23%) among the BWC group, compared with a 10.6% increase in the comparison group (Katz et al .2015 ). Given the national attention on BWCs, it is possible that willingness to adopt BWCs is a latent reflection of a department ’s intentions to keep up with expectations for modern policing (Byrne and Marx 2011). Therefore, BWCs may also be associated with more progressive and strategic models such as hot spot, problem-oriented, predictive, or intelligence-led policing. The entropic lens In contrast to the rational perspective, the entropi c perspective does not anticipate clear or consist- ent connections between police s trategy and technology. Specifically, this perspective suggests that law enforcement agencies adopt technology a ccording to trivial or external factors that do not relate systematically to the agency ’s central or long-term goals. Indeed, some research has suggested that law enforcement agencies select, implement, and integrate technology indepen- dent of existing empirical evide nce or concern for how these sys tems affect departmental oper- ations, strategic decisions, or crime outcomes. Byrne and Marx ( 2011) argue that empirical research that documents the effectiveness of a gi ven technology typically plays a minor role in the decisions to adopt or continue using that tec hnology. In essence, it is argued that law enforce- ment adopts technology as a ‘black box ’(Weisburd and Neyroud 2011). Therefore, notwithstanding the connections between strategy and technology anticipated when viewing these domains through the rational lens, the entropic perspecti ve expects few or inconsistent relationships 732 J. A. HENDRIX ET AL. between strategy and technologybecause it does not perceive law enforcement agencies to make decisions about technology in a tactical and deliberate way. Agency size To assess the extent to which associations exist between technology use and belief in particular stra- tegic models, we drew upon survey data from a nationally representative sample of state and local law enforcement agencies, including appropriate proportions of small, medium, and large agencies. However, research has indicated that the use of technology, policing activities, and other agency characteristics (e.g. style of leadership) may vary significantly by agency size (Mamalian et al.1999 , Chamard 2002,2003 , Lum et al.2010 , Schuck 2015). Larger agencies tend to have more slack resources to invest in new technologies, and they tend to have greater diversity of job functions and more specialised units that require the adoption of more technologies (Nunn 2001, Mastrofski et al .2003 , Skogan and Hartnett 2005, Randol 2012). Larger agencies with more specialised job func- tions may be more ‘in the know ’of the newest research, practices, and technologies available to inform agency goals (Weisburd and Lum 2005). Because the vast majority of agencies in the United States have fewer than 250 full-time sworn officers, analytical models that have been adjusted to represent the overall population of police departments will provide information reflective of small agencies and will say little about larger departments. Therefore, in addition to the primary analysis of the nationally representative sample, we also assessed the relationship between strategy and tech- nology for a subset of large agencies. Methods The study was conducted collaboratively by the P olicing, Security, and Investigative Science Program at RTI International and the PERF. To a ssess our key research question, we developed a questionnaire and administered it to a nationa lly representative sample of law enforcement agencies. The content of the survey was develop ed with input from an expert panel, which was convened in June 2013 in Washington, DC. The panel consisted of nine criminal justice pro- fessionals and civilians who had expertise worki ng in law enforcement and experience selecting and implementing technology in police agencies. Once the survey instrument was finalised, the panel reviewed and provided recommendations to ensure that the team was able to meet the pro- ject ’s goals and objectives. The sampling frame for the survey was developed using the 2012 National Directory of Law Enfor- cement Administrators (NDLEA), an electronic list obtained from the National Public Safety Infor- mation Bureau (NPSIB). The 2012 NDLEA is composed of contact information for 15,847 law enforcement agencies in the United States. On the basis of a power analysis, our initial goal was to obtain a minimum of 949 surveys. Assuming a 74% completion rate, this would have required a sample of 1283 law enforcement agencies. To ensure adequate representation from each type of agency in the survey responses, the sample included all tribal ( n= 69) and state agencies ( n= 49). 2 The remaining desired sample count was stratified to ensure adequate representation across census regions (Northeast, Midwest, South, and West) and further stratified by agency type and agency size (1 –99, 100 –249, 250 –499, and more than 500 sworn officers). The required sample size, after subtraction of the tribal and state agencies, was evenly dispersed across the 32 cells created by the cross-tabulation of region/type by size. It was clear that low cell counts would affect agencies with more than 250 sworn officers. As a result, all agencies with 250 or more sworn officers were included in the sample ( n= 360). The remaining count ( n= 707) was dis- tributed across the other 16 cells of the cross-tabulation (as sizes 250 –499 and 500+ were removed). Using this process, we needed to randomly select 45 agencies within each remaining cell. Some cells within the 100 –249 size range had fewer than 45 agencies and were fully sampled. A random selec- tion of agencies was then generated within each stratum. POLICING AND SOCIETY 733 Data collection Survey respondents were contacted and prompted via nonresponse follow-up through multiple mailings and phone calls. The survey was initially distributed in February 2014, followed by two reminder letters sent three and six weeks after the initial survey distribution. Next, a mailed notifi- cationletterfromtheprojectofficerwassentinApril 2014. To address nonresponse, an email was sent to nonresponding PERF (general) members in May 2014, followed by a mailed reminder letter in May 2014 to all nonresponding agencies. In an effort to boost response rates among small agencies, an additional mailed reminder letter was sent with a targeted explanation of the impor- tance of the project and its relevance to small age ncies. After the mailed survey prompts, we con- ducted two waves of reminder phone calls to the 350 nonresponding agencies in June and July 2014. At the conclusion of the data collection period, we had obtained a response rate of 60.5% ( N =776). An analysis of the final sample showed some differences with the sampling frame derived from the2012NDLEA.Forinstance,thefinalsampleund errepresents agencies from the Northeast and Midwest and overrepresents agencies from the W est. Additionally, the sample underrepresents police departments but overrepresents county/sheriff ’s offices, tribal agencies, and state/ highway agencies. A higher proportion of the s ample is composed of agencies with at least 100 full-time sworn officers. To adjust these propo rtions so that they resemble proportions from the 2012 NDLEA, we used a procedure known as raking ratio estimation. Raking assigns a weight value to each survey respondent so that marginal totals of the a djusted weights on specified characteristics are in line with the corresponding totals for the population. A major advantage of raking is its ability to produce respondent wei ghts that are based on multiple control totals (Kalton 1983, Battaglia et al.2004 ). Twenty-seven agencies were dropped from ana lysis because they answered only a few ques- tions at the beginning of the survey. An assessment of these agencies ’key characteristics as they relate to size, type, and region did not indicate any systematic bias and thus there was no reason to believe that dropping these cases altered the results. Missing data on individual items through- out the survey were minimal. On average, 2.8% and 4.2% of respondents had missing data on items related to policing strategies and policing activities, respectively. 3Multiple imputation was used to estimate a set of plausible values for missing data and to replace missing values with the combined results (Little and Rubin 2002). A series of five imputations was used to predict missing values; the resulting estimates reflected statistically valid inferences with adjusted standard errors that account for the uncertainty that derives from missing values (Allison 2002). Measures Dependent variables This study has six dependent variables, one for each of the six types of technology: crime mapping, social media, data mining, car cameras, LPRs, and BWCs. Respondents were asked to indicate whether they had used each technology in the past two years. Responses were coded 0 if the agency had not used the technology in the past two years and 1 if it had. Independent variables The key independent variables were eight survey items that asked respondents to indicate how important various policing strategies were for supporting the agency ’s core mission on a scale of 1 (not important at all) to 5 (highest importance). The policing strategies derived from the expert panel and include professional, community, problem-oriented, zero-tolerance, hot spot, offender tar- geting, intelligence-led, and predictive policing. 734 J. A. HENDRIX ET AL. Control variables Our analytical models predict the odds of technology use with agency strategy, controlling for region, agency type, and agency size. Initially, our models included controls for agency annual budget and local crime rates; however, concerns related to collinearity led us to drop these items from the models. There were few instances in which these items were statistically significant. Four regions of the United States were represented by three dummy variables: Northeast, South, and Midwest as designated by the US Census. The West region was assigned as the reference category. In the NDLEA 2012, agency type was originally composed of four values: police department/city sheriff’s offices (municipal), county police/sheriff ’s offices, state police or highway patrol, and tribal police departments. Because of small cell sizes for tribal and state police/highway patrol agencies, we recoded agency type into a single dummy variable comparing municipal police departments to all others. The number of sworn officers was recoded into an interval item with seven categories for the full sample, based on the number of full-time sworn officers reported (0 –4; 5 –9; 10– 24; 25 –49; 50– 99; 100 –249; 250+). For the subsample of large agencies, the sworn officer variable was recoded to represent three levels (250 –499; 500 –999; 1000+ full-time sworn officers). Analytical approach Our primary models predict the odds of technology use in the past two years for six types of tech- nology by strategy and other characteristics using logistic regression. We also predict the odds of technology use among a subsample of large law enforcement agencies (i.e. agencies with 250 or more full-time sworn officers) ( n= 302). Results Table 2 provides descriptive statistics for key variables used in the analysis. As shown, car cameras are the most commonly used technology; about 70% of the sample reported use in the past two years. Similarly, about 68% of respondents reported that they had used social media for public communi- cation in the past two years. About a third of the sample had used BWCs or GIS technology, whereas smaller proportions of the sample had used LPRs or data mining tools. Turning to policing strategies, professional policing had the highest average score (4.79 on the 5-point scale), whereas zero-toler- ance policing had the lowest average score (3.29). The South made up the largest geographic cat- egory, whereas the least number of agencies were from the West. Most responding agencies are municipal, and 70% of agencies have fewer than 25 officers. Table 3 presents results from logistic regression models predicting technology use in the past two years for each of the six technologies. As shown, there are few statistically significant relationships between strategies and technology. Only three exceptions can be identified. First, the odds of social media use are 2.73 times higher with each one-unit increase on the community policing measure. An assessment of predicted probabilities, holding all other variables at their means, suggests that only 8% of agencies who marked community policing as ‘not important at all’have used social media in the past two years, compared to 87% of agencies who marked it as ‘highest importance ’. Alternatively, the odds of social media use are about 60% lower for every one-unit increase on the zero-tolerance policing measure. Predicted probabilities show that 96% of agencies who marked zero-tolerance policing as ‘not important at all’had used social media in the past two years, compared with only 45% of those who marked it as highest importance. Finally, the odds of LPR use are about 67% higher for each unit increase on the predictive policing measure. Agency size was a more consistent predictor of technology use than strategy, as it was statistically significant and positive in direction for four of six technologies (GIS, data mining software, social media, and LPRs). Regional effects were scarce and inconsistent. The odds of use of data mining soft- ware were lower for Northeast agencies than for Western agencies, whereas Northeast agencies were POLICING AND SOCIETY 735 Table 1.Descriptive statistics for sample, sampling frame, and weighted sample. Sample ( N = 776) M 2012 Directory ( n = 15,847) MFinal weighted sample (N = 776) M(n ) Region Northeast 0.150.20 0.20 (155) Midwest 0.210.33 0.33 (256) South 0.360.35 0.35 (272) West 0.280.12 0.12 (93) Agency type Municipal 0.480.79 0.79 (613) County/sheriff ’s offices 0.40 0.190.19 (147) Tribal 0.060.01 0.01 (8) State or highway 0.060.01 0.01 (8) Sworn officers 0–4 0.020.20 0.20 (155) 5– 9 0.040.22 0.22 (171) 10 –24 0.090.28 0.28 (217) 25 –49 0.050.13 0.13 (101) 50 –99 0.050.08 0.08 (62) 100 –249 0.320.05 0.05 (39) 250+ 0.430.04 0.04 (31) Table 2.Descriptive statistics for key variables (weighted) ( N= 749). Mean SD Used technology in past two years Car cameras 0.700.46 Social media for public communication 0.680.47 BWCs 0.330.47 GIS 0.310.46 LPR 0.200.40 Data mining tools for massive databases 0.100.30 Policing strategies Professional policing 4.790.52 Community policing 4.400.71 Problem-oriented policing 4.300.75 Intelligence-led policing 3.910.92 Hot spot policing 3.740.96 Offender targeting policing 3.810.97 Predictive policing 3.541.01 Zero-tolerance policing 3.291.03 Control variables Geographic region Northeast 0.200.40 Midwest 0.330.47 South 0.350.47 West 0.120.32 Agency type Municipal 0.790.41 County/sheriff’ s offices 0.190.40 Tribal 0.010.10 State or highway 0.010.06 Agency size (number of full-time sworn officers) 0–4 0.200.40 5– 9 0.220.41 10 –24 0.280.45 25 –49 0.130.34 50 –99 0.080.27 100 –249 0.050.22 250+ 0.040.20 736 J. A. HENDRIX ET AL. more than five times more likely to have used social media. Midwestern agencies were considerably more likely than Western agencies to have used car cameras in the past two years. 4 Table 4presents results for a subsample of agencies with 250 or more full-time sworn officers. Agencies that embrace community policing, hot spot policing, and offender targeting policing are more likely to have used GIS technology, whereas agencies who emphasise predictive policing are less likely. Municipal agencies are also several times more likely to have used GIS than other types of agencies. No statistically significant effects were identified for data mining software or BWCs. The odds of social media use are twice as high with every incremental increase in the importance of community policing and hot spot policing. Similar results are found for LPRs; community policing and hot spot policing each positively predict the odds of LPR use. The odds of car camera use are considerably higher in the Midwest and South relative to the West, although strategy appears to be unimportant. Interestingly, agency size is not statistically significant for any of the technologies. Considering it was the most reliable predictor of technology for the national sample, there appears to be a threshold after which size is no longer important for technology acquisition. The finding that size matters more for technology acquisition in smaller agencies is consistent with some past studies (e.g. Schuck 2015). Discussion Policing and technology are increasingly intertwined, yet unanswered questions remain on how police departments select technology. One of the key questions this study sought to address is whether law enforcement agencies adopt technologies based on the strategic, organisational frame- work they align with or if technologies are selected in a manner that is independent of larger agency goals. When looking at the national sample of all agencies, our findings show little evidence that agencies adopt technology based on strategic goals, as we found no statistically significant relation- ships between technology and strategy for GIS, data mining software, car cameras, or BWCs. A negative association was identified between zero-tolerance policing and social media use; while important, these findings are not surprising given that this style of policing is rooted in tactics that emphasise non-discretionary, heavy enforcement of both crime and public disorder. Zero-tolerance is not typically associated with strategies that emphasise extensive uses of technology to target resources in time or space or to solve particular problems. We also found a positive relationship between predictive policing and the use of LPRs, which although was unanticipated, has reasonable Table 3. Logistic regression predicting technology use in the past two years ( N= 749). GIS Data mining Social media LPR Car camera BWC Strategies Professional 0.74 (0.25) 1.23 (0.32) 0.62 (0.34) 1.14 (0.37) 0.84 (0.34) 0.74 (0.27) Community 1.06 (0.35) 0.98 (0.28) 2.73* (1.17) 1.38 (0.43) 0.69 (0.26) 0.69 (0.26) Problem-oriented 1.35 (0.39) 0.51 (0.17) 0.85 (0.35) 0.58 (0.21) 1.27 (0.46) 0.75 (0.28) Zero-tolerance 1.07 (0.23) 0.79 (0.13) 0.40** (0.12) 0.77 (0.15) 1.05 (0.21) 1.00 (0.22) Hot spot 0.62 (0.17) 1.17 (0.25) 1.01 (0.29) 1.07 (0.38) 0.64 (0.20) 0.88 (0.24) Offender targeting 0.95 (0.22) 1.41 (0.30) 1.71 (0.51) 0.78 (0.21) 1.22 (0.32) 0.76 (0.18) Intelligence-led 1.40 (0.35) 1.15 (0.33) 0.50 (0.19) 1.11 (0.31) 1.31 (0.55) 1.34 (0.41) Predictive 1.02 (0.26) 1.28 (.30) 1.75 (0.57) 1.67* (0.31) 0.87 (0.36) 1.13 (0.33) Sworn officers 1.87*** (0.29) 1.98*** (0.15) 1.49** (0.20) 2.16*** (0.26) 1.03 (0.12) 0.87 (0.11) Region Midwest 0.76 (0.43) 0.36 (0.20) 0.74 (0.49) 0.57 (0.39) 7.19** (5.03) 0.60 (0.34) South 0.72 (0.48) 0.88 (0.55) 0.87 (0.58) 0.71 (0.45) 2.55 (1.40) 1.31 (0.74) Northeast 0.71 (0.46) 0.26* (0.17) 5.61* (4.07) 0.79 (0.45) 0.46 (0.28) 0.57 (0.36) Type Municipal 0.45* (0.17) 0.21*** (0.08) 0.95 (0.38) 0.94 (0.36) 0.97 (0.42) 0.96 (0.36) F 2.73*** 14.69*** 2.95*** 6.76*** 2.43** 0.94 Pseudo- R 2 0.25 0.300.220.29 0.180.09 Note: Coefficients are odds ratios, standard errors in parentheses. *<.05; **<.01; ***<.001. POLICING AND SOCIETY 737 conceptual tie. LPRs produce an extensive amount of data that could feasibly be used to predict future patterns including crime prone locations, offender travel pathways, and other issues related to public safety. A closer examination of the benefits and limitations of the use of LPR data for pre- dictive policing is needed to better understand this connection.By default, the overall lack of relationships between technology and strategy lends more support to the entropic perspective, or the notion that technology is adopted independently of an agency ’s strategic goals. These findings may suggest that other agency- or contextual-level factors identified in past research may be more important for determining the types of technology police departments implement. It may be the case that agency characteristics such as their involvement in cosmopolitan networks, institutional pressures to appear progressive or as keeping up with other agencies, or staff experience levels handling various types of technology exert a more direct effect on the agency ’s decision to seek out technologies (Skogan and Hartnett 2005, Willis et al.2007 , Leong and Chan 2014 ). Perhaps not surprisingly, the relationship between strategy and technology was contingent on agency size. For large agencies, we found stronger connections between the policing strategies and technologies implemented. In some instances, technologies have clear implications with regard to strategy implementation. For example, as predicted, agencies that emphasise community policing and hot spot policing are more likely to have used GIS. Crime mapping technology is useful for identifying where crimes cluster, a core tenet of hot spot policing, and for informing the public regarding where crimes occur, which is compatible with underlying goals of community policing. Contrary to what we hypothesised, the odds of GIS use were lower among agencies who emphasised predictive policing. The meaning of this relationship is unclear, especially given that GIS is essential for predictive analytical techniques. It may be that predictive analytic methods are seen as indepen- dent of GIS technology. Although predictive modelling is not necessarilydependent on GIS, predictive policing methods have tended to focus on geography and, therefore, should be strongly related to GIS. Additionally, while not hypothesised, the finding that hot spot policing and social media use are interconnected among larger agencies is important. It is possible that agencies who emphasise hot spot policing perceive value in social media for gathering intelligence regarding the occurrences of crime in the community. Social media may also be seen as an important method of identifying short- term crime or disorder hotspots (e.g. flash mobs or party locations). We also did not predict that community or hot spot policing would be positively associated with LPR use. Considering that community and hot spot policing are associated with several technologies Table 4. Logistic regression predicting technology use in the past two years (large agencies) ( n= 302). GIS Data mining Social media LPR Car camera BWC Strategies Professional 0.77 (0.28) 1.06 (0.22) 0.76 (0.33) 0.90 (0.24) 1.42 (0.29) 0.88 (0.18) Community 3.79*** (1.50) 1.51 (0.41) 2.13* (0.78) 2.30** (0.73) 0.80 (0.24) 1.50 (0.40) Problem-oriented 0.52 (0.21) 0.88 (0.23) 0.59 (0.18) 0.73 (0.19) 0.83 (0.24) 0.87 (0.21) Zero-tolerance 0.62 (0.16) 0.84 (0.14) 0.73 (0.15) 0.83 (0.15) 0.78 (0.19) 0.77 (0.13) Hot spot 2.40** (0.76) 1.16 (0.27) 2.17** (0.66) 2.52*** (0.63) 0.78 (0.19) 1.77* (0.44) Offender targeting 2.43* (1.00) 1.09 (0.29) 0.90 (0.33) 0.48* (0.14) 1.39 (0.31) 1.04 (0.21) Intelligence-led 1.91 (0.85) 1.62 (0.49) 1.44 (0.74) 1.56 (0.52) 0.82 (0.24) 0.83 (0.21) Predictive 0.38* (0.17) 0.81 (0.17) 0.58 (0.22) 0.80 (0.17) 1.41 0.29 1.07 (0.21) Sworn officers 1.81 (0.77) 1.12 (0.25) 0.84 (0.27) 1.60 (0.42) 1.33 (0.37) 1.03 (0.26) Region Midwest 3.31 (3.06) 1.65 (0.87) 1.84 (1.41) 2.69 1.72) 9.67** (6.80) 0.79 (0.41) South 2.42 (1.51) 1.23 (0.53) 2.65 (1.38) 2.01 (0.87) 3.16** (1.33) 1.54 (0.64) Northeast 0.67 (0.55) 1.79 (1.04) 0.72 (0.56) 3.13 (1.84) 0.28* (0.14) 0.20 (0.18) Type Municipal 4.85** (2.55) 1.33 (0.42) 2.15 (1.07) 1.60 (0.59) 0.71 (0.29) 1.35 (0.49) Model F 3.43*** 1.23 2.03* 2.22** 3.62*** 1.41 Pseudo- R 2 0.49 0.09 0.35 0.250.260.12 Note: Coefficients are odds ratios, standard errors in parentheses. *<.05; **<.01; ***<.001. 738 J. A. HENDRIX ET AL. when the analysis is limited to large agencies, it may be that large agencies that emphasise these strategies are more technologically savvy, forward-thinking, and‘in the know’when it comes to tech- nology (Weisburd and Lum 2005). Some scholars have suggested that the adoption of emerging technologies and adherence to innovative policing strategies are ways that agency leaders can dis- tinguish themselves professionally by showing their willingness to adopt state-of-the-art methods in place of standing methods of practice (Byrne and Marx 2011). Implications The evidence suggests only a limited coupling between strategy and technology. Closer alignment between technology implementation choices and prioritised philosophies is mainly limited to larger agencies and analytically based technologies and policing strategies. The lack of a more wide- spread relationship between strategy and technology suggests that technology adoption is not typi- cally driven by careful consideration of the agency ’s larger strategy or future direction. This raises questions about how technology can and should be used to support broader policing planning and mission. Some seemingly foundational relationships between technology and strategy (e.g. hot spot policing and the use of GIS) were not found. It is unclear how agencies are operationalising these strategies without the use of these technologies. We found more evidence that strategy was linked to technology among large law enforcement agencies. Perhaps some characteristics of larger agencies make them more likely to pair technology with strategy. For instance, perhaps larger organisations need to have more complex structures and technologies to support key functions (Pugh and Hickson 1976). Larger agencies also tend to have staff in place whose roles are at least partially focused on identifying and selecting technologies and setting the strategic mission of the agency. In addition, larger agencies may have more staff and resources to develop clearly articulated strategic plans from which technology choices can be rooted. Whatever the case, future research in this area should be focused on disentangling the relation- ship between agency size, personnel capabilities and capacity, technology, and policing strategy. This work should be guided by past research on police innovation and change, as framing these questions in different organisational change perspectives may provide further insight into why police depart- ments adopt technology. Developing a better understanding of the barriers for small and mid-size agencies should also be prioritised. This research may lead to actionable information that small agencies can implement to better take advantage of technologies to support strategy. For example, if the main barrier for these agencies is the lack of qualified personnel, then joint powers or regional associations may be developed to thoughtfully implement technologies that are consist- ent with the aims of member organisations. Weak, or non-existent, relationships between strategy and technology adoption may represent a major disadvantage for agencies. The lack of relationships between strategy and technology may be due to the absence of coordinated planning by law enforcement agencies. Without proper longer range planning around strategic direction and how agencies acquire and implement technologies, the acquisition, implementation, maintenance, and evaluability of police technologies will be nega- tively affected. Rather, we argue that technology use within an agency can be maximised if individual technology decisions are made within a larger framework and in concert with other technology decisions. For instance, an agency planning to purchase BWCs should first determine if the techno- logical foundational is in place. For agencies, this means that planning is essential to maximise the utility of a technology in sup- porting core strategies. During the acquisition process, agencies would be required to properly vet technology for usefulness. Similarly, agencies should ensure fidelity to the technology implemen- tation and maintenance process for maximum returns. Finally, planning is essential for an agency to collect key metrics of technology use and to evaluate how new technologies impact agency out- comes and strategy. The ineffective evaluation of past technology acquisition limits an agency ’s POLICING AND SOCIETY 739 ability to learn from past efforts. Building an evaluation mindset into an agency’s regular practices is critical if an agency wants to create successful technology implementation process. Limitations Some limitations of this analysis must be acknowledged. First, the dependent variable is a binary measure of use in the past two years. This measure does not consider the span of adoption (e.g. LPRs on 2% of cars or on 80% of cars), nor does it consider the quality of adoption (e.g. deep inte- gration of LPRs with records management and other systems versus an LPR system that only collects data that must be manually reviewed on a case-by-case basis). Our inability to measure these poten- tially salient factors makes it more difficult to speculate on how strategy affects technology. Second, the agency surveys were filled out by a single point of contact in each department. Whereas some questions should be largely unaffected by this –questions about the use of a tech- nology, for example, should be relatively consistent from person to person within the agency – others may not be so consistent: questions about agency perspective and orientation towards poli- cing philosophy may have distinct differences between people within the same department. There is little information about how perceptions of policing strategy vary within the department. Further research is needed to learn if there are differences across ranks and if these differences manifest themselves in orientation towards policing strategy. Finally, we make no claim on the causal relationship between strategy and technology. Although we found social media and community policing to be related, this does not indicate that the orien- tation towards community policing caused the adoption of social media. It may be that certain tech- nologies are seen as representative of progressive law enforcement agencies and certain strategies are seen as similarly progressive. The causal relationship between strategy and technology would be better studied with longitudinal research. Conclusion Technology is now essential to police operations and will continue to accelerate in its use and reach within agencies. Implemented properly, technology can be a key resource in carrying out police activities and larger strategy. The results of this study, however, suggest that police technology and strategy are not strongly coupled in US law enforcement agencies. However, there were some relationships between technology and strategy (e.g. the use of social media and community policing) that make sense conceptually and practically. Nevertheless, other strong links between technology and agency strategy that appeared conceptually sound (e.g. use of GIS and hot spot policing among smaller agencies) did not manifest. In general, the relationships between technology and strategy were more consistent among larger agencies. Even among these, however, there were inconsistencies with theoretical perspectives. Taken together, these results suggest that, while technology implementation is being implemented in policing at an increasingly rapid pace, there needs to be greater emphasis on evi- dence-based, informed decision-making about new and existing technology. Furthermore, agencies must come together to develop agency-specific and also shared visions for how certain technologies can help them achieve their goals. Similar to new staff, technologies should be viewed as potential assets to agencies but to maximise their impact decisions need to be made that specify how they will be used, by whom, and what will be accomplished by their use. Notes 1. Results from a randomized controlled experiment in Mesa, Arizona, conducted by the Police Executive Research Forum (PERF) indicated no relationship between the number of scanned license plates and vehicle theft rates 740 J. A. HENDRIX ET AL. (Tayloret al.2012 ). Similarly, others find that the use of LPRs does not have an appreciable effect on reducing auto thefts (Lum et al.2010 ). 2. Hawaii does not have a state police agency. 3. To account for missing data for the remaining sample ( N= 749), we first performed tests to ensure that the missing data were missing at random. Logistic regression models were used to predict the odds of having a missing value on each of our dependent variables by key agency characteristics (region, size, type). Results did not indicate that specific agency characteristics were associated the odds of having a missing value on various technologies. 4. This effect may be directly due to differences in political climate and differences in funding received, and indirectly related to prevalence of alcohol-impaired driving (Jewett et al.2015 , Schuck 2015). Disclosure statement No potential conflict of interest was reported by the authors. Funding This work was supported by National Institute of Justice [2012-MU-CX-0043]. References Allison, P.D., 2002.Missing data . Thousand Oaks: Sage, 104. Barak, A., Farrar, F., and Sutherland, A., 2014. The effect of police body-worn cameras on use of force and citizens ’com- plaints against the police: a randomized controlled trial. Journal of quantitative criminology, 31, 509–535. Battaglia, M., et al.,2004 . Tips and tricks for raking survey data (a.k.a. sample balancing). Paper presented at the annual meeting of the American Association for Public Opinion Research, Phoenix, AZ. Bayerl, P.S., et al.,2013 . The role of macro context for the link between technological and organizational change. Journal of organizational change management, 26, 793 –810. Braga, A.A., et al.,1999 . Problem-oriented policing in violent crime places: a randomized controlled experiment. Criminology , 37, 541–580. Braga, A.A. and Bond, B.J., 2008. Policing crime and disorder hot spots: a randomized controlled trial. Criminology, 46, 577 –607. Braga, A.A., Papachristos, A.V., and Hureau, D.M., 2012. The effects of hot spots policing on crime: an updated systematic review and meta-analysis. Justice quarterly,1,1–31. Burger, L., 2013.Four ways social media can help police departments [online]. Available from:http://www.policeone.com/ social-media-for-cops/articles/6250205-4-ways-social-media-can-help-police-departments/ [Accessed 29 September 2016]. Byrne, J. and Marx, G., 2011. Technological innovations in crime prevention and policing. A review of the research on implementation and impact. Journal of police studies, 20, 17–40. Caplan, J.M., Kennedy, L.W., and Miller, J., 2011. Risk terrain modeling: brokering criminological theory and GIS methods for crime forecasting. Justice quarterly, 28, 360–381. Carter, J.G., 2016. Institutional pressures and isomorphism: the impact on intelligence-led policing adoption. Police quar- terly , 19, 435 –460. Chamard, S., 2002. The diffusion of crime mapping in American police departments. Paper presented at the annual meeting of the Western Society of Criminology, San Diego, CA. (Available from Justice Center, University of Alaska Anchorage.) Chamard, S., 2003.The adoption and discontinuance of computerized crime mapping by municipal police departments in New Jersey . Doctoral dissertation. UMI, Ann Arbor. Chamard, S., 2006. The history of crime mapping and its use by American police departments. Alaska Justice Forum [online], 23 (3). Available from: http://justice.uaa.alaska.edu/forum/23/3fall2006/a_crimemapping.html [Accessed 29 September 2016]. Chau, M., Xu, J., and Chen, H., 2002. Extracting meaningful entities from police narrative reports. Proceedings of the National Conference for Digital Government Research . Los Angeles, CA. Cohen, M.D., March, J.G., and Olsen, J.P., 1972. A garbage can model of organizational choice. Administrative science quar- terly , 17, 1– 25. Cyert, R.M. and March, J.G., 1963.A behavioral theory of the firm . Malden: Blackwell, 268. Danziger, J.N. and Kraemer, K.L., 1985. Computerized data-based systems and productivity among professional workers: the case of detectives. Public administration review , 45, 196–209. POLICING AND SOCIETY 741 Darroch, S. and Mazerolle, L.,2013. Intelligence-led policing: a comparative analysis of organizational factors influencing innovation uptake. Police quarterly, 16, 3–37. Darroch, S. and Mazerolle, L., 2015. Intelligence-led policing: a comparative analysis of community context influencing innovation uptake. Policing and society, 25, 1– 24. Dunworth, T., et al.,2001 .Police department information systems technology enhancement project . Washington, DC: U.S. Department of Justice, Office of Community Oriented Policing Services. Fayyad, U. and Uthurusamy, R., 2002. Evolving data into mining solutions for insights. Communications of the ACM, 45, 28– 31. Federal Bureau of Investigation. 2011.Uniform Crime Report: crime in the United States, 2010. Offenses cleared [online]. Available from: https://www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/2010/crime-in-the-u.s.-2010/clearancetopic. pdf [Accessed 29 September 2016]. Federal Bureau of Investigation. 1996.Uniform Crime Report: crime in the United States, 1995. Crime index offenses cleared [online]. Available from: https://www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/1995/95sec3.pdf [Accessed 29 September 2016]. Giblin, M.J., 2006. Structural elaboration and institutional isomorphism: the case of crime analysis units. Policing: an inter- national journal of police strategies and management, 29, 623 –664. Goodison, S., Davis, R.C., and Jackson, B.A., 2015.Digital evidence and the U.S. criminal justice system: identifying technology and other needs to more effectively acquire and utilize digital evidence . Santa Monica, CA: RAND, 32. Harris, C., 2007. Police and soft technology: how information technology contributes to police decision making. In:J. Byrne and D. Rebovich, eds. The new technology of crime, law, and social control . Monsey, NY: Criminal Justice Press, 153 –183. Hauck, R.V., et al.,2002 . Using Coplink to analyze criminal-justice data. IEEE computer, 35, 30–37. Hsu, J. 2014.70 percent of U.S. police departments use license plate readers [online]. Available from:http://spectrum.ieee. org/cars-that-think/transportation/sensors/privacy-concerns-grow-as-us-police-departments-turn-to-license-plate- readers [Accessed 1 September 2016]. International Association of Chiefs of Police. ( 2004).Impact of video evidence on modern policing [online]. Washington, DC: U.S. Department of Justice, Office of Community Oriented Policing Services. Available from: http://www.theiacp.org [Accessed 29 September 2016]. International Association of Chiefs of Police Center for Social Media. ( 2014).2014 survey results [online]. Available from: http://www.iacpsocialmedia.org/Resources/Publications/2014SurveyResults.aspx [Accessed 29 September 2016]. Ioimo, R.E. and Aronson, J.E., 2004. Police field mobile computing: applying the theory of task-technology fit. Police quar- terly , 7, 403 –428. Jewett, A., et al.,2015 . Alcohol-impaired driving among adults –United States, 2012. Morbidity and mortality weekly report , 64, 814 –817. Kalton, G., 1983.Compensating for missing survey data . Ann Arbor: Survey Research Center, Institute for Social Research, University of Michigan, 85. Katz, C.M., 2001. The establishment of a police gang unit: an examination of organizational and environmental factors. Criminology , 39, 37–74. Katz, C.M., et al.,2015 .Evaluating the impact of officer worn body cameras in the phoenix police department . Phoenix: Arizona State University, Center for Violence Prevention and Community Safety, 43. Lawrence, P.R. and Lorsch, J.W., 1967.Organization and environment: managing differentiation and integration . Boston: Harvard University, 296. Leong, K. and Chan, S., 2014. A study of predictors and the extent of web-based crime mapping adoption by a police agency. Security journal , 27, 421–442. LexisNexis. ( 2014).Social media use in law enforcement: Crime prevention and investigative activities continue to drive usage [online]. Available from: http://www.lexisnexis.com/risk/downloads/whitepaper/2014-social-media-use-in-law- enforcement.pdf [Accessed 29 September 2016]. Little, R.J. and Rubin, D.B., 2002.Statistical analysis with missing data . New York: John Wiley and Sons, 408. Lum, C. 2010.Gadgets for gathering evidence are not evidence of better policing: technology and the mythology of progress in American law enforcement [online]. Available from:http://www.scienceprogress.org/2010/02/police-technology/ [Accessed 1 September 2016]. Lum, C. 2010.License Plate Recognition Technology (LPR) impact evaluation and community assessment [online]. Fairfax, VA: George Mason University. Available from: http://www.scienceprogress.org/2010/02/police-technology/ [Accessed 29 September 2016]. Mamalian, C.A., LaVigne, N.G., and the Staff of the Crime Mapping Research Center. 1999.The use of computerized crime mapping by law enforcement: survey results . Research Preview [online]. Washington, DC: National Institute of Justice. Available from: https://www.ncjrs.gov/pdffiles1/fs000237.pdf [Accessed 1 September 2016]. Mastrofski, S.D., Parks, R.B., and Wilson, D.B., 2003.Influences on the adoption of community policing in the United States. Manassas, VA: George Mason University. Report to the National Institute of Justice. 742 J. A. HENDRIX ET AL. Mazerolle, L.G., Bellucci, C., and Gajewski, F.,1997. Crime mapping in police departments: the challenges of building a mapping system. In: D. Weisburd and T. McEwen, eds. Crime mapping and crime prevention. Monsey: Criminal Justice Press/Willow Tree Press, 131 –155. Moore, M.H. and Trojanowicz, R.C., 1988.Corporate strategies for policing (Perspectives on Policing No. 6) [online]. Washington, DC: U.S. Department of Justice. Available from: https://www.ncjrs.gov/pdffiles1/nij/114215.pdf [Accessed 29 September 2016]. Morabito, M.S., 2008. The adoption of police innovation: the role of the political environment. Policing: an international journal of police strategies and management, 31, 466 –484. Nunn, S., 2001. Police information technology: assessing the effects of computerization on urban police functions. Public administration review , 61, 221–234. Pearsall, B., 2010. Predictive policing: the future of law enforcement. NIJ journal[online], 266. Available from: http://www. nij.gov/journals/266/pages/predictive.aspx [Accessed 1 September 2016]. Pugh, D.S. and Hickson, D.J., 1976.Organizational structure in its context: the Aston Programme I . Lexington: Saxon House/ Lexington Books, 301. Randol, B.M., 2012. The organizational correlates of terrorism response preparedness in local police agencies. Criminal justice policy review , 23, 304–326. Randol, B.M., 2014. Modelling the influence of organisational structure on crime analysis technology innovations in municipal police departments. International journal of police science and management, 16, 52 –64. Roberts, D.J., and Casanova, M. 2012.Automated license plate recognition systems: policy and operational guidance for law enforcement [online]. Washington, DC: U.S. Department of Justice. Available from: https://www.aclu.org/files/ FilesPDFs/ALPR/federal/NHTSA/15948-16075DOJ-IACP%20report.pdf [Accessed 21 September 2016]. Rogers, E.M., 1962.Diffusion of innovations . New York: Free Press, 576. Roman, J.K., 2008.The DNA field experiment: cost-effectiveness analysis of the use of DNA in the investigation of high-volume crimes . Washington, DC: The Urban Institute. Available from: https://www.ncjrs.gov/pdffiles1/nij/grants/222318.pdf [Accessed 1 September 2016]. Roth, J.A., et al.,2000 . Using COPS resources. In: J.A. Roth and J.F. Ryan, eds. National evaluation of the COPS Program –Title I of the 1994 Crime Act. Washington, DC: U.S. Department of Justice, 101 –114. Schuck, A.M., 2015. Prevalence and predictors of surveillance cameras in law enforcement: the importance of stake- holders and community factors. Criminal justice policy review, 28, 1–20. Scott, W.R., 2008.Institutions and organizations: ideas and interests . Los Angeles: Sage, 360. Sherman, L.W. and Eck, J.E., 2002. Policing for crime prevention. In: L.W. Sherman, et al., eds. Evidence-based crime preven- tion . New York: Routledge, 295 –329. Sherman, L.W. and Weisburd, D.A., 1995. General deterrent effects of police patrol in crime ‘hot spots ’: a randomized con- trolled trial. Justice quarterly , 12, 625–648. Simon, H.A., 1997.Administrative behaviour . New York: Simon and Schuster, 384. Skogan, W. and Frydl, K., 2004.Fairness and effectiveness in policing: the evidence . Washington, DC, National Academies Press, 432. Skogan, W.G. and Hartnett, S.M., 2005. The diffusion of information technology in policing. Police practice and research,6, 401 –417. Sparrow, M.K., Moore, M., and Kennedy, D.M., 1990.Beyond 9-1-1: a new era for policing . New York: Basic Books, 288. Stevens, L. 2010. Social media in policing: nine steps for success. The Police Chief[online], 77 (2). Available from: http:// www.policechiefmagazine.org/magazine/index.cfm?fuseaction=display&article_id=2018&issue_id=22010 [Accessed 29 September 2016]. Taylor, B., Koper, C., and Woods, D., 2012. Combating auto theft in Arizona: a randomized experiment with license plate recognition technology. Criminal justice review, 37, 24–50. Weisburd, D. and Braga, A., 2006.Police innovation: contrasting perspectives . New York: Cambridge University Press, 388. Weisburd, D. and Lum, C., 2005. The diffusion of computerized crime mapping in policing: linking research and practice. Police practice and research , 6, 419–434. Weisburd, D., and Neyroud, P. 2011. Police science: toward a new paradigm. New Perspectives in Policing[online]. Available from: https://www.ncjrs.gov/pdffiles1/nij/228922.pdf [Accessed 29 September 2016]. Weiss, A., 1997. The communication of innovation in American policing. Policing: an international journal of police strat- egies management, 20, 292 –310. Westphal, L.J. 2004. The in-car camera: value and impact. The Police Chief[online], 71 (8). Available from: http://www. policechiefmagazine.org/magazine/index.cfm?fuseaction=display&article_id=358 [Accessed 1 September 2016]. Willis, J.J., Mastrofski, S.D., and Weisburd, D., 2007. Making sense of COMPSTAT: a theory-based analysis of organizational change in three police departments. Law and society review, 41, 147–188. Wilson, J.Q. and Kelling, G., 1982. Broken windows: the police and neighborhood safety. Atlantic Monthly, March, pp. 29– 38. POLICING AND SOCIETY 743