Prevalence and Predictors of Surveillance Cameras in Law Enforcement

Abstract

Using data from more than 2,500 law enforcement agencies, the goal of this study was to identify predictors of advanced surveillance technologies. The findings suggest that the adoption of modern surveillance cameras is neither uniform nor comprehensive and that the adoption process is ongoing with agency officials implementing and discontinuing technologies over time. Most important, stakeholders both inside and outside the organization have the greatest influence on the adoption process, and cameras in vehicles and mobile devices are most prevalent in improvised communities. As cameras become smaller and less expensive, they have the potential to democratize surveillance and equalize the relationship between the police and the public during encounters. However, the democratization effect will only occur if implementation is widespread and all segments of the community have an equal voice in the process. The research findings suggest that significant progress still needs to be made in these areas.

Keywords

cameras community policing collective bargaining LEMAS organizational differentiation police accountability S-curve stakeholders surveillance technology union

Although scholars and policymakers have raised concerns about the implementation of advanced surveillance technologies, including issues related to privacy, the militarization of the police, and the implications of private verses public control (Byrne & Marx, 2011), in general, there is widespread support for the use of surveillance equipment in law enforcement, and many researchers agree that technology-driven forms of policing are growing in popularity (Ratcliffe, 2008; Willis, Mastrofski, & Weisburd, 2007). Between 1995 and 2002, more than 1.3 billion dollars was distributed to law enforcement agencies for technology-related hardware and software (Groff & McEwen, 2008). Between 2000 and 2005, the Community-Oriented Policing Services (COPs) office awarded 21 million dollars to state law enforcement agencies to purchase in-car cameras (International Association of Chiefs of Police, 2004). In a survey conducted by the Police Executive Research Forum (PERF; 2011), 70% of police executives reported using some type of advanced technologies to prevent crime and reduce violence, and almost 90% reported planning to increase their use of technologies related to predictive policing over the next 5 years. Furthermore, there is growing evidence that if implemented correctly and used in tandem with other law enforcement strategies, such as hot-spot policing or problem-orientated policing, technologies can help improve police performance (Braga, 2006; Koper, Taylor, & Woods, 2013; Piza, Caplan, Kennedy, & Gilchrist, 2014; Ratcliffe, Taniguchi, & Taylor, 2009).

In light of these facts, the question remains, “Why do all law enforcement agencies not use advanced surveillance technologies?” The purpose of this study is to identify which agencies are utilizing modern surveillance equipment. More specifically, the goal is to understand the organizational, community, and political factors associated with the adoption of in-car and mobile cameras. Scholarship that evaluates the adoption of advanced surveillance equipment is important because new developments in technology have the potential to influence organizational change and restructure police work. Surveillance technologies have the potential not only to reduce crime and violence but also to improve institutional transparency and increase police accountability. Studying the adoption of emerging technologies in law enforcement is important because it elucidates key policy implications for police executives and the communities that they represent.

Background and Literature Review

The adoption of advanced surveillance technology by police agencies is a function of complex interactions between the characteristics of the technology, the cultures within the organizations, and other factors in the larger social-structural environment. Technology shapes the way people interact with the world around them; however, people and the larger social-structural environment also influence the structure and function of technology. To understand the adoption of technology in law enforcement, we need to move beyond a purely rational approach and take more of an interpretive approach. From this perspective, the adoption of specific devices results from the interplay among organizational needs, completing technologies, stakeholders’ conflicting goals, and chance events (Ackroyd, Soothill, Harper, Hughes, & Shapiro, 1992; Robey & Sahay, 1996). Stakeholders construct the meaning of the technology, and this construction represents opportunities for organizational change (Barley, 1986).

In general, technologically focused innovation in law enforcement is driven by seven basic imperatives: (a) increasing organizational effectiveness, (b) increasing efficiency, (c) creating a safer work environment for officers (see PERF, 2012), (d) adopting cost-effective replacements for traditional law enforcement methods, (e) fulfilling the demands for greater information sharing among government organizations, (f) meeting the requests for more accountability and transparency from both within and outside the agency (Ackroyd et al., 1992; Chan, 2001), and (g) strengthening the professional status of the policing occupation and enhancing organizational legitimacy (Chan, 2001; Ericson & Haggerty, 1997; Manning, 1992). Interestingly, the last three imperatives are associated with the “entrepreneurial revolution” in law enforcement (Ackroyd et al., 1992), and the last two are associated with the new professionalism model of modern policing (see Stone & Travis, 2011).

The factors that influence the course of technological change in policing can be organized into four areas: (a) the technology itself, including the hardware, software, and human factors and ergonomics (HF&E) of the system; (b) characteristics of the department, including institutional rules and organizational structures as well as attitudes about technology held by members of the agency; (c) environmental factors in the larger social-structural context, such as crime rates and characteristics of the community; and (d) political factors, which include stakeholders inside and outside the organization and the conflict that results from the introduction of the new technology. Each of these is described in greater detail below.

Technology

The design of the hardware and software and the ways in which they are managed can influence the adoption of the technology and the degree to which it is implemented in the organization. Braga and Weisburd (2007) argue that “police most easily adopt innovations that require the least radical departure from their hierarchical paramilitary organizational structures, continue incident-driven and reactive strategies, and maintain police sovereignty over crime issues” (p. 17). In other words, technologies that support the current organizational structure and do not alter the balance of power are more likely to be adopted than those that support radical changes to the policing profession. Furthermore, technology that is ineffective, poorly managed, or quickly outdated is also less likely to be adopted (Schuck & Rosenbaum, 2008). In a study of Canadian police organizations, Ericson and Haggerty (1997) argue that technology that is more difficult to circumvent will have a greater impact on the organization.

Organizational Factors

Research suggests that some police organizations are better at adapting, changing, and innovating than others. Drawing from organizational theories, scholars hypothesize that characteristics of the organization and administrative structure predict innovation (King, 1998). Langworthy (1986) identified four key structures of police agencies that are related to organizational complexity and control: (a) spatial, the degree to which organizations are geographically distributed; (b) hierarchical, the vertical nature of the command structure; (c) occupational, the degree to which organizations employ specialized staff; and (d) functional, the extent to which agencies created unique units or institutional structures to carry out different organizational tasks. Expanding on Langworthy’s work, Maguire (1997) added the concepts of formalization, administrative density, and task scope. Formalization is defined as the degree to which organizations formalize rules and procedures, administrative density refers to the size of the administrative division of the agency, and task scope identifies the number of different functions performed by departmental personnel (Maguire, 1997).

Although there is limited research on the application of Langworthy’s and Maguire’s organizational characteristics to the adoption of advanced video surveillance technologies, researchers have evaluated their impact on the adoption of community policing and innovations in policing more broadly. For example, Wilson (2005) found that formalization, funding, and police turnover were associated with successful implementation of community policing. In another study, King (1998) evaluated organizational characteristics on different types of innovation and found that organizational size, vertical concentration, and occupational heterogeneity were associated with technical innovations. In a study of homeland security innovations, Randol (2012) found that size, functional differentiation, and budget were associated with terrorism preparedness.

Police departments are not monolithic, static entities; rather, they evolve and adapt in response to changes in the larger social-structural environment. There are two basic models of technology adoption—the s-curve (Rogers, 2003) and the viral or hockey stick. The s-curve is used to describe technology adoption where there is a triggering event followed by quick adoption rate with high expectations for the technology. Next comes a decline in adoption rates, while some discontinue using the technology. The decline is generally attributable to the technology failing to meet expectations. Finally, we see a leveling off—an equilibrium in the adoption rate where the technology meets the expectations of its users. In contrast, the viral or hockey stick represents technology adoption that is slow in the beginning and then takes off. Regardless of the adoption model, most police organizations are already somewhere on the curve, and their position on the curve should influence the adoption of other technologies. For example, Mullen (1996) concluded that departments with more computers have a greater technological adoption rate than those with fewer computers.

Community Factors

While researchers acknowledge the importance of social contexts in understanding the adoption of technology, less is known about which environmental factors influence executive decision making. In the boarder literature on companies, researchers have identified competition as the most important environmental factor related to the adoption of advanced technologies (Chau & Tam, 1997). However, the concept of competition has limited applicability to government bureaucracies such as law enforcement agencies. In theory, organizations that are under pressure to reduce crime or solve problems associated with the agency, such as police misconduct or officer safety, will be more likely to adopt advanced surveillance technologies. Furthermore, drawing from social disorganization theory, in communities where there is a perceived need for greater supervision, such as neighborhoods with significantly more youths, police executives, and residents may see advanced surveillance technologies as a cost-effective solution to increasing supervision. Inequality may be another important factor. According to the research, increased residential security in terms of reduced access and increased surveillance occurs for two reasons—fear and prestige (Sanchez, Lang, & Dhavale, 2005). Residential fortification occurs in both affluent and disadvantaged neighborhoods in an effort to control the physical environment and prevent undesirable elements from affecting community members. Drawing from this literature, we would expect to see greater support for the utilization of advanced surveillance technologies in communities that have high economic inequality.

Political Factors

According to Manning (1996), technological changes often “destabilize the power balance between organizational segments by altering communication patterns, roles relationships, the division of labor, established formats for organizational communication, and taken-for-granted routines” (p. 54). For example, in-car dash video cameras can reduce the autonomy of street-level police officers and change the balance of power between officer and citizen, officer and supervisor, and between the organization and the public that it serves. Individual officers and organizations that represent officers, such as the Fraternal Order of Police, are likely to resist the adoption and widespread implementation of technologies that limit officers’ discretion or alter relationships inside the organization or between the agency and the public. For example, there are several cases of union resistance to the implementation of management information systems designed to increase oversight of law enforcement organizations (Walker, 2001).

Another important factor related to the adoption of technology by law enforcement is the political climate. The political climate is the collective or aggregate opinion of the populace regarding a specific issue or problem. For example, the documents leaked by Edward Snowden concerning National Security Agency (NSA) programs where millions of phone records, emails, and web chats were collected have drastically altered the political climate regarding the surveillance of American citizens. Conventional wisdom suggests that more liberal political ideologies are associated with greater concerns about issues related to personal privacy. However, more recent research suggests that conservatives are less supportive of national surveillance data collection systems, such as those run by the NSA, than liberals (Pew Research Center, 2013).

Hypotheses

Starting with organizational characteristics, it is hypothesized that larger, more formalized agencies with higher levels of spatial, hierarchical, and organizational differentiation will utilize a greater number of mobile cameras and cameras in vehicles than smaller, less formalized organizations with less spatial, hierarchal, and organizational differentiation. Moving to community factors, higher levels of crime and greater officer safety concerns will be related to more camera utilization. More poverty, unemployment, inequality, and youths as well as fewer homeowners will be related to more surveillance technology. Because the literature on residential fortification suggests that increased surveillance occurs both in neighborhoods with high economic affluence and in neighborhoods with high economic deprivation, there should be a significant interaction between poverty and inequality. Finally, turning to the political dimension, organizations with collective bargaining rights are hypothesized to have fewer in-car and mobile cameras than agencies with no collective bargaining rights. Departments that operate in more politically conservative communities are hypothesized to operate more cameras than departments that operate in more liberal political environments.

Method

Data

The data for this study came from four sources: (a) the 2000, 2003, and 2007 Law Enforcement Management and Administrative Statistics (LEMAS) surveys (U.S. Department of Justice, 2006, 2008, 2011); (b) the 1999, 2000, 2002, 2003, 2006, and 2007 FBI uniform crime reports (UCR; Federal Bureau of Investigation, 2001, 2002, 2004, 2005, 2008, 2009); (c) the 2005 to 2009 and 2006 to 2010 5-year estimates from the American Communities Survey run by the U.S. Census Bureau (U.S. Census Bureau, 2014a; 2014b); and (d) presidential election results for 2000, 2004, and 2008 from Atlas of U.S. Presidential Elections (Leip, n.d.).

Samples

Information on the number and type of cameras came from LEMAS. LEMAS is a national survey of American law enforcement agencies conducted once every 3 to 4 years. Data collected as part of the series includes information on the size of the agency, the population of the community served, personnel data, expenditures, and facts about organizational operations and management. The LEMAS series consists of two samples—one self-representing sample of agencies with 100 or more sworn officers and one stratified random sample of agencies. To test the hypotheses with both large and small agencies, two separate data sets were created. Self-representing agencies were matched across the three waves (2000, 2003, and 2007) to create a longitudinal or panel data set (N = 437). Agencies from the stratified random samples in 2007 (n = 1,030), 2003 (n = 877), and 2000 (n = 636) were combined to create a cross-sectional data set (N = 2,543). Using the two different data sets will allow for hypothesis testing with agencies of different sizes as well as managing statistical problems associated with error structure because of non-independence.

Variables

Definitions for the variables used in this study are presented in Table 1. The descriptive statistics for the cross-sectional data set are also presented in Table 1. The descriptive statistics for the panel data set are presented in Table 2. Hypotheses were tested using two different types of advanced surveillance cameras—cameras in patrol vehicles and mobile surveillance cameras. The dependent variables were operationalized as the number of cameras that were operated on a regular basis by the agency during a 12-month period prior to the survey being conducted. Two variables were created to control for the confounding effect of organization size—the number of motor vehicles operated by the agency and the number of sworn officers per 10,000 residents.

Table 1.

Description of Variables and Descriptive Statistics for Cross-Sectional Sample (N = 2,543).

Variable	Description	Range	M (SD)
In-vehicle cameras	The number of in-car cameras operated in the prior 12 months.	0-780	6.4 (20.8)
Mobile cameras	The number of mobile cameras operated in the prior 12 months.	0-25	0.21 (0.97)
Agency size	The number of full-time sworn officers divided by the population per 10,000 residents logged.	0.91-8.6	3.1 (0.47)
Spatial differentiation	The sum of all district, precinct, substations, and neighborhood community substations logged.	0-4.3	0.28 (0.49)
Hierarchical differentiation	The difference in the base salary between the chief executive and an entry-level officer, as a proportion of the entry-level officer’s salary.	−0.50-5.14	0.92 (0.56)
Occupational differentiation	The percentage of employees in the organizations who were civilian (non-sworn).	0-80	16.7 (12.1)
Formalization	The number of written policies and procedures implemented represented as a percentage.	0-100	80.8 (17.6)
Information technology	The number of technology systems represented as a percentage.	0-100	46.7 (25.6)
Community policing	The number of community policing activities conducted in the prior year represented as a percentage.	0-100	31.4 (26.2)
Crime rates	The crime rate per 100,000 residents logged.	0-13.2	8.0 (0.96)
Assaults on officers	The assault rate per 1,000 officers logged.	0-8.1	2.0 (2.4)
Percent homeowners	The percentage of owner-occupied housing.	13.3-100	66.8 (12.8)
Youth to adult ratio	The number of youth under 18 per adults between 18 and 64.	4.9-98.2	37.4 (11.8)
Unemployment	Percent residents unemployed.	0-45.0	6.9 (3.8)
Poverty	Percent residents living in poverty.	0.3-67.5	14.5 (9.12)
Inequality	The Gini coefficient.	0.26-0.66	0.43 (0.05)
Collective bargaining	Coded as 1 = collective bargaining rights for sworn personnel and 0 = no collective bargaining rights.	0-1	0.53 (0.50)
Percent voting Republican	Percentage of votes for the Republican presidential candidate.	10.2-91.9	51.3 (12.9)
Number of vehicles	The number of motor vehicles operated by the agency logged.	0-7.33	2.7 (0.99)

Table 2.

Descriptive Statistics for Panel Sample (N = 437).

	2000		2003		2007
	Range	M (SD)	Range	M (SD)	Range	M (SD)
In-car cameras	0-481	8.7 (31.9)	0-512	22.1 (49.5)	0-549	35.4 (65.4)
Mobile cameras	0-33	0.95 (2.7)	0-256	1.7 (14.4)	0-118	2.0 (9.3)
Agency size	2.3-4.4	3.1 (0.35)	2.3-4.3	3.1 (0.35)	2.3-4.5	3.1 (0.34)
Spatial differentiation	0-4.5	1.3 (0.94)	0-4.3	1.3 (0.91)	0-4.6	1.1 (0.89)
Hierarchical differentiation	0.27-4.7	1.5 (0.64)	0.21-4.3	1.4 (0.58)	0.33-7.4	1.5 (0.72)
Occupational differentiation	0-54.6	22.3 (9.3)	0-47.3	21.9 (9.1)	0-53.3	22.6 (9.3)
Formalization	40-100	92.1 (11.3)	46.7-100	89.4 (10.4)	47.1-100	84.7 (10.8)
Information technology	7.7-100	64.0 (21.9)	0-100	60.8 (28.0)	6.3-100	77.8 (21.0)
Community policing	0-100	58.1 (24.4)	0-100	60.5 (26.9)	0-100	61.7 (25.4)
Crime rates	3.9-9.9	8.6 (0.57)	4.3-10.0	8.5 (0.52)	6.7-9.9	8.5 (0.48)
Assaults on officers	0-7.0	3.7 (2.1)	0-7.1	3.5 (2.2)	0-6.6	3.1 (2.4)
Collective bargaining	0-1	0.66 (0.47)	0-1	0.68 (0.47)	0-1	0.65 (0.48)
Number of vehicles	2.0-8.1	4.7 (1.0)	1.8-8.0	4.7 (1.1)	1.8-8.8	4.8 (1.0)

	2000		2004		2008
	Range	M (SD)	Range	M (SD)	Range	M (SD)

Percent voting Republican	8.9-79.3	45.8 (11.9)	9.3-81.6	48.4 (12.3)	6.5-78.2	42.8 (12.3)

	2005-2009
	Range	M (SD)

Percent homeowners	21.5-89.6	57.8 (11.0)
Youth to adult ratio	13.3-68.4	37.5 (8.8)
Unemployment	2.7-22.1	8.0 (2.8)
Poverty	3.2-37.0	16.5 (7.0)
	2006-2010
	Range	M (SD)

Inequality	0.34-0.60	0.46 (0.04)

Drawing from Langworthy (1986) and Maguire (1997), four organizational measures were created. Some examples of items used to create the formalization scale include written policies regarding the use of non-lethal force, off-duty employment, interacting with the media, and dealing with mentally ill persons. Organizational variables were also created to measure the degree to which informational technology was integrated into department operations as well as the agency’s commitment to community policing. Examples of information technology items include whether computers were used to analyze community problems, dispatching, fleet identification, in-field report writing, intelligence gathering, inter-agency information sharing, and resource allocation. Examples of the community policing scale include giving patrol officers responsibility for specific beats/areas, encouraging officers to engage in SARA problem solving, and maintaining a community policing plan. Because of changes to the instrument across the three waves of the LEMAS data, the measures for formalization, information technology, and community policing were calculated as percentages.

To evaluate the impact of the larger social-structural environment on the adoption of video camera technology, seven community measures were created. Using data from the uniform crime reports, the crime rates (per 100,000 residents) and the assaults on officers rate (per 1,000 full-time sworn officers) were calculated. Because the 1-year reference period in the LEMAS surveys regarding the agency’s operation of cameras does not correspond with the calendar year and varies slightly from wave to wave, 2-year averages were used for the crime rate and assaults against officers rate (e.g., the average of 2006 and 2007 for 2007). The crime rate and the assaults on officer rate were transformed using the log because of skewness.

Using data from the American Communities Survey (ACS) 5-year estimates (2005-2009), information on youths, homeowners, unemployment, and poverty were matched to the law enforcement agencies using census place and county codes.¹ Five-year estimates were chosen because they provide the most reliable estimates and can be calculated for all areas (compared with 3-year estimates, which are less reliable and only provide information for areas with a population of 20,000 or more; U.S. Census Bureau, 2008). Information from 2005 to 2009 was chosen because it is the earliest 5-year data set produced by the ACS. Inequality was measured using the Gini coefficient and was gathered from the 2006 to 2010 ACS (U.S. Census Bureau, 2014b). The Gini coefficient is a summary measure of inequality based on household income and ranges from 0 (which indicates perfect equality) to 1 (which indicates total inequality). Finally, to measure political influence, data from the Atlas of U.S. Presidential Elections (Leip, n.d.) was used to calculate the percentage of votes cast in each county for the Republican presidential candidate in 2000 (George W. Bush), 2004 (George W. Bush), and 2008 (John McCain).

Results

Type and Number of Cameras

Representatives from the agencies in the cross-sectional sample (N = 2,543) reported operating a total of 20,512 cameras in vehicles and a total of 735 mobile surveillance cameras. The average number of cameras operated by the departments was 6.4 in-car cameras and 0.23 mobile cameras (see Table 1). Approximately 43.7% of agencies reported that they did not operate any in-car cameras, and 87.8% reported that they did not operate any mobile cameras. Consistent with the assertion that the number of cameras used by law enforcement agencies was growing, the average number of in-car cameras estimated for agencies drawn from the 2007 LEMAS sample was significantly higher than those drawn from the 2003 or 2000 samples (10.3 vs. 5.0 vs. 2.1), F(2, 2540) = 34.54, p < .001. This trend was also true for mobile cameras (0.28 vs. 0.22 vs. 0.16), F(2, 2540) = 3.20, p = .041. When looking at whether the agency operated any cameras, those agencies drawn from the 2007 sample were more likely to report operating at least one in-car camera than those drawn from the 2003 or 2000 sample (46.5% vs. 35.1% and 18.4%, respectively), χ²(2, N = 2543) = 87.32, p < .001. For mobile cameras, the results suggest a greater rate of adoption between 2003 and 2007 than between 2000 and 2003 (14.6% of agencies reported operating at least one mobile camera in 2007 vs. 10.1% in 2003 and 11.0% in 2000), χ²(2, N = 2543) = 9.69, p < .008.

Turning to the longitudinal sample (N = 437), agency representatives reported that in 2007, they operated a total of 15,456 in-car cameras and 881 mobile cameras. These figures were up from 3,784 in-car cameras and 416 mobile cameras in 2000. The agencies in the longitudinal sample reported operating an average of 35.4 in-car cameras and 8.7 mobile cameras (see Table 2). Like the cross-sectional sample, a substantial number of agencies in the longitudinal sample reported not operating any cameras. In 2007, 37.5% of agencies reported not operating any in-car cameras, and 71.4% reported not operating any mobile cameras.

Consistent with the s-curve theory of technology adoption, a small number of agencies either reduced the number of cameras that they operated or completely discontinued their camera system. Out of the 437 agencies in the longitudinal sample, 54 (or approximately 12.4%) decreased the number of in-car cameras they operated, and of those, 29 (53.7%) discontinued operating any type of in-car camera system. A larger percentage of agencies reduced their use of mobile cameras. Twenty-four percent (or approximately 105 agencies) reduced their number of mobile cameras, with approximately 87.6% of those (or 92 agencies) discontinuing their use of mobile cameras.

Multivariate Results

Prior to evaluating the hypotheses, the data were screened for problems using strategies summarized by Tabachnick and Fidell (2007).² A generalized linear model (GLM) framework was used to analyze the data. The GLM is a generalized version of linear regression that allows for the evaluation of dependent variables that have error structures other than a normal distribution via a link function. Because the dependent variables are counts and the variance is substantially larger than the mean (i.e., overdispersion), a negative binomial distribution was used.³ All models were estimated using Stata 13. The results for the cross-sectional data are presented in Table 3. The estimates are interpreted in terms of the differences in the logs of expected counts for the number of cameras, while controlling for the other variables in the model. The exp(b) estimates, referred to as the incidence rate ratios (IRR), are also presented in Table 3. Because the agencies were asked how many cameras they operated in the previous 12 months, the IRR represents the percentage change in the rate of cameras operated (per 12 months) for a one-unit change in the independent variable. For example, the in-car camera count increases by 9.6% for every one-unit increase in the crime rate. Because the interaction between poverty and inequality was not significant for mobile cameras (b = .10, SE = .13, p = .46, 95% confidence interval [CI] [−.16, .35]), it was removed, and the model was re-estimated to facilitate the interpretation of the main effects for poverty and inequality.

Table 3.

Generalized Linear Model Results Using a Negative Binomial Distribution and Robust Standard Errors Estimating the Number of Cameras for the Cross-Sectional Sample (N = 2,543).

	In-car cameras			Mobile cameras
	b	Robust SE	Exp(b)	b	Robust SE	Exp(b)
Intercept	−2.429***	.528	0.088	−3.196*	1.254	0.041
Agency size	0.063	.059	1.065	0.490*	0.214	1.632
Spatial differentiation	−0.039	.065	0.962	0.392**	0.144	1.480
Hierarchical differentiation	−0.128	.067	0.880	0.007	0.149	1.007
Occupational differentiation	0.002	.002	1.002	0.002	0.009	1.002
Formalization	0.006***	.002	1.006	−0.012*	0.006	0.988
Information technology	0.004**	.001	1.004	0.011*	0.005	1.011
Community policing	0.001	.001	1.001	0.009*	0.004	1.010
Crime rates	0.092**	.035	1.096	−0.007	0.103	0.993
Assaults on officers	−0.023*	.011	0.978	−0.009	0.030	0.991
Percent homeowners	0.007**	.003	1.007	0.001	0.007	1.001
Youth to adult ratio	0.006**	.002	1.006	−0.017*	0.007	0.983
Unemployment	−0.002	.008	0.998	−0.038	0.025	0.963
Poverty	−0.075***	.021	0.927	0.046**	0.014	1.047
Inequality	−1.578*	.790	0.206	−4.907*	1.711	0.007
Poverty × Inequality interaction	0.169***	.043	1.184	NE	NE	NE
Collective bargaining	−0.446***	.058	0.640	−0.042	0.177	0.959
Percent voting Republican	0.011***	.002	1.011	0.019**	0.006	1.019
Data from 2000	−1.038***	.076	0.354	0.110	0.211	1.117
Data from 2003	−0.286***	.056	0.751	−0.029	0.192	0.972
Number of vehicles	0.837***	.048	2.309	0.456***	0.108	1.577
Fit information
Log pseudolikelihood	−6,049.3			−1,229.5
Deviance ([1/df] deviance)	2,603.4 [1.03]			927.8 [0.37]
Pearson ([1/df] Pearson)	1,653.6 [0.66]			3,405.5 [1.35]
AIC [BIC]	4.77 [−17,171.9]			0.98 [−18,847.5]

Note. NE = not estimated; AIC = Akaike information criterian; BIC = Bayesian information criterian.

p < .05. **p < .01. ***p < .001.

Table 4.

Generalized Estimation Equations (GEE) Results Using a Negative Binomial Distribution and Robust Standard Errors Estimating the Number of Cameras for the Panel Sample (N = 437).

	In-car cameras			Mobile cameras
	b	Robust SE	Exp(b)	b	Robust SE	Exp(b)
Intercept	−2.302	1.268	0.100	−4.527	2.612	0.011
Agency size	−0.232	0.218	0.793	−0.267	0.378	0.766
Spatial differentiation	−0.185*	0.077	0.831	0.178	0.138	1.194
Hierarchical differentiation	0.097	0.087	1.102	−0.016	0.143	0.984
Occupational differentiation	0.002	0.007	1.002	0.015	0.010	1.016
Formalization	−0.013**	0.004	0.987	−0.001	0.007	0.997
Information technology	0.002	0.002	1.002	0.008*	0.004	1.008
Community policing	0.005*	0.002	1.005	0.009	0.005	1.009
Crime rates	−0.034	0.106	0.966	0.628*	0.280	1.873
Assaults on officers	−0.033	0.023	0.967	−0.087	0.046	0.917
Percent homeowners	0.030***	0.007	1.031	0.008	0.014	1.008
Youth to adult ratio	−0.003	0.008	0.997	−0.039**	0.012	0.962
Unemployment	−0.007	0.030	0.993	−0.091	0.053	0.913
Poverty	0.053***	0.013	1.055	0.059	0.025	1.061
Inequality	0.717	1.669	2.049	−7.110	3.742	0.001
Collective bargaining	−0.658***	0.122	0.518	−0.616	0.237	0.540
Percent voting Republican	0.018***	0.006	1.018	−0.010	0.012	0.990
Number of vehicles	0.763***	0.085	2.145	0.692***	0.131	1.998

p < .05. **p < .01. ***p < .001.

The results support both hypotheses drawn from the political domain and suggest that unionization is associated with fewer in-car cameras, while a more conservative political climate is associated with greater in-car camera utilization. The findings also support the hypothesis regarding crime. Agencies in areas with higher crime rates operated more in-car cameras. However, counter to the hypothesis presented in the hypothesis section regarding officer safety, there was a negative relationship between in-car camera utilization and the rate of assault on officers.

The percentage of homeowners was related to greater in-car camera utilization. These results point to homeowners operating as a constituency, rather than a lack of homeowners (or renters) being a group in need of more surveillance. The findings also suggest that there is an interaction between poverty and inequality. Independently, poverty and inequality are associated with fewer in-car cameras; however, in communities with high levels of poverty and high levels of inequality, there are more cameras in vehicles. Regarding hypotheses drawn from the organizational domain, a more developed information technology infrastructure and more institutional formalization were associated with more in-car cameras. There is about a half percent increase in the incident rate for cameras for each percentage change in formalization and information technology.

Turning to mobile cameras, the results suggest that larger agencies and those with greater geographical differentiation operate more mobile cameras than smaller organizations and agencies with a smaller geographic footprint. Greater implementation of information technology and community policing was associated with more mobile cameras, and greater formalization was associated with fewer mobile cameras. Greater levels of poverty were associated with the utilization of more mobile cameras, while greater inequality and fewer youths relative to adults were associated with fewer mobile cameras. Similar to in-car cameras, the percentage of Republican voters was associated with the operation of a greater number of mobile surveillance cameras.⁴

Generalized estimating equations (GEE), which are considered extensions of the GLM, were used to evaluate the hypotheses using the panel data. GEEs where first introduced by Liang and Zeger (1986) and are used to accommodate the analysis of data that is clustered or non-independent. GEEs are a marginal or population-averaged method (vs. a cluster-specific or conditional method) and are most appropriate when the primary research question is the regression equation for the marginal expectations and not the intracluster correlation structure (Hardin, 2005). Because the data were collected over time, an autoregressive (AR-1) working correlation matrix was used (Diggle, Heagerty, Liang, & Zeger, 2002). The interaction between poverty and inequality was not significant in either model.

For the large agencies, spatial differentiation, formalization, community policing, the percentage of homeowners, poverty, collective barging, and the percentage of Republican voters predicted the adoption of in-car cameras. There were fewer predictors for mobile cameras. Only information technology, crime rates, and number of youths predicted the utilization rate of mobile surveillance cameras.

Discussion

Many advances in technology that occurred during the early 20th century had a significant impact on the organization of police work. For example, the introduction of two-way radios, motor vehicles, and computer-aided dispatching significantly changed how policing was structured (C. J. Harris, 2007; Manning, 1992). In contrast, more recent developments in technology, such as those related to information technology and telecommunications, have had mixed results (C. J. Harris, 2007; Manning, 2003). The purpose of this research was to assess why some agencies embrace advanced surveillance technologies while others do not. Results from both the longitudinal and cross-sectional samples suggest that the number of camera systems operated by law enforcement agencies is growing. However, one of the most striking discoveries was that a significant number of agencies did not operate surveillance systems. More than 2,200 (or 87%) of the agencies in the cross-sectional sample and approximately 310 agencies (71%) in the longitudinal sample reported that the organization did not operate mobile cameras. Another third of agencies reported not operating in-car cameras. Although these data are several years old and it is likely that the number of agencies using advanced surveillance technologies has changed, the results presented here suggest that the utilization of camera systems is limited to a select group of law enforcement agencies.

Another important finding is that the adoption of advanced surveillance equipment follows an s-curve, with some agencies reducing or discontinuing their use of the technology. This is important because as cameras get smaller and less expensive, many advocates are calling for the widespread implementation of camera technology as a means of increasing police accountability (see D. A. Harris, 2010). If agencies do not maintain the cameras or if the cameras are easy to circumvent, it is unlikely that the system will have a significant impact on police practices. More research is needed to understand why agencies stop using cameras. Most of the research in this area is grounded in theories related to the diffusion of innovation and, as such, treat continuance as an extension of acceptance behaviors. However, these theoretical models are limited. One possible avenue for research lies in expectation-confirmation theory (Bhattacherjee, 2001), where police executives are viewed as needing to upgrade or repurchase technology and their satisfaction with the system is constantly being reevaluated. Satisfaction with the system can be a function of many things, including the impact of the technology, the cost structure of maintaining the system, long-term unintended consequences of the equipment, changing community stakeholders’ perceptions, and other police executives’ experiences.

The results suggest that the model is better at explaining cameras in medium- and small-sized agencies and that political and community factors are generally better predictors than the characteristics of the organization. Linked to the new professional model of policing, the results highlight the importance of stakeholders, both inside and outside the organization, in shaping law enforcement operations. In the future, scholars should examine the role of state and federal funding on the implementation and maintenance of advanced surveillance technologies. For example, grants are often used to promote innovation in policing, and research suggests that additional government funding can have a significant impact on organizational operations (Zhao, Scheider, & Thurman, 2002). In theory, the availability of additional revenue, as well as the infrastructure in the organization to acquire and disperse such funding, should have a consequential effect on the adoption of camera technologies. Equally important, the discontinuance of external funding programs may be one of the reasons that agencies do not replace aging or non-functioning equipment.

As with the prior research on collective bargaining, the findings from this study suggest that union representatives resist changes that reduce officer discretion or alter the relationship between the officer and the administration. Walker (2008) called unions one of the most neglected areas of research on policing. These findings contribute to the growing body of knowledge on how collective bargaining may impede innovation in police organizations.

The results also highlight the importance of the political climate on police operations and the complex relationship between the police and the public. A more conservative political climate was associated with more cameras. Liberal community members may have less confidence in the police and may be more concerned about issues related to privacy. However, this situation is somewhat of a paradox because, under the new professional model of policing, fewer cameras means greater officer discretion, less transparency, and less police accountability. Community trust is an important element of the new professionalism model. If community residents distrust the police, institutional rules and organizational procedures designed to increase the publics’ perception of organizational legitimacy, such as cameras, will be less effective.

National public opinion polls suggest that Republicans are more concerned than Democrats about broad-based national surveillance programs when Americans are the subjects of surveillance (Pew Research Center, 2013). Interestingly, the estimated effect of Republican political influence on mobile cameras in large departments was negative and not significant. This finding highlights the potential importance of administrative discretion regarding who should be the subject of surveillance on public support for police-controlled technologies. Unlike fixed or closed-circuit television (CCTV) cameras, mobile cameras can be redeployed at the discretion of the administration or the officer.

In small- and medium-sized agencies, there was a significant relationship between mobile cameras and agency size and spatial differentiation. In contrast, for large law enforcements agencies, the strongest predictor of mobile camera utilization was crime. These findings support the assertion by Ackroyd et al. (1992) that the adoption of specific technologies results from the interplay between the social context and organizational needs. Smaller agencies appear to be using mobile cameras to increase surveillance coverage in departments that are geographically large, whereas officials from larger agencies appear to be using mobile cameras to address crime problems in specific neighborhoods.

Interestingly, there was a negative relationship between the percentage of youths and the number of mobile cameras for both the cross-sectional and the panel samples. Police executives and citizens in communities with a significant number of youths may be concerned about the legal ramifications of videotaping underage residents. These findings highlight the potential importance of state and local legislation and scholars should study the impact of the law and judicial orders on the adoption of advanced surveillance technologies by law enforcement organizations in future research projects.

Limitations

This study is not without limitations. Although the data come from the most recent LEMAS survey, it is still several years old, and as such, the number of agencies that have implemented advanced surveillance technologies is probably not accurate. However, by taking data from the three most recent LEMAS surveys, the information in this study is comprehensive and provides a benchmark that other researchers can use to evaluate future changes in the adoption of surveillance technologies. Even more important, the theoretical propositions tested in this study are hypothesized to be general, and thus, the results provide an accurate evaluation of the most recent knowledge regarding the adoption of advanced surveillance technology in law enforcement.

Conclusion

Over the past 30 years, there has been a substantial increase in the utilization of advanced surveillance technologies in American law enforcement. Most agencies no longer rely solely on traditional methods of surveillance; instead, they utilize a wide range of electronic equipment, including fixed and mobile cameras, in-car cameras, red light cameras, speed cameras, wireless video streaming, automatic license plate readers (ALPR), Global Position Systems (GPS), infrared imagers, and biometric technology such as facial recognition. The findings from this study suggest that the adoption of advance surveillance technologies is neither uniform nor comprehensive and that the adoption process is ongoing and evolutionary, with agency officials implementing and discontinuing technologies over time. Furthermore, stakeholders, both inside and outside the organization, play a significant role in the adoption process, and cameras are generally more prevalent in improvised communities. As cameras become smaller and less expensive, they have the potential to democratize surveillance and equalize the relationship between the officer and the citizen during encounters. These results raise important questions regarding the democratization of surveillance. These findings raise questions regarding (a) who should have input in the process when deciding to implement advanced surveillance technologies in a community; (b) who should be subjected to more surveillance, especially surveillance that is general and not targeted at a specific problem; and (c) how much discretion should officers and police executives have regarding the redeployment of surveillance equipment. These are important questions because the results from this study confirm the proposition that technology interacts with the social context and both enables and constrains organizational implementation of surveillance resources.

Going forward, scholars should focus on studying the factors related to agencies discontinuing surveillance technologies. Users of technology expect the new infrastructure to reduce crime and increase police accountability. The findings from this study suggest that in a significant number of cases, the technology did not meet the expectations of agency officials and community residents. What were the expectations for the technology, and where did it fall short? What role did cost, effectiveness, efficiency, or changing stakeholders’ perceptions play in executives’ decision making? These questions are important, and we can learn a lot from both those who adopt and those who discontinue using advanced surveillance technologies.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Amie M. Schuck, PhD in Criminal Justice, is an associate professor in the Department of Criminology, Law and Justice at the University of Illinois at Chicago. Her areas of research interest include community-police relations, youth attitudes toward the police, quantitative methodology, and public safety technologies.

References

Ackroyd

Soothill

Harper

Hughes

J. A.

Shapiro

(1992). New technology and practical police work: The social context of technical innovation. Bristol, PA: Open University Press.

Barley

S. R.

(1986). Technology as an occasion for structuring: Evidence from observations of CT scanners and the social order of radiology departments. Administrative Science Quarterly, 31(1), 78-108.

Bhattacherjee

(2001). Understanding information systems continuance: An expectation-confirmation model. MIS Quarterly, 25, 351-370.

Braga

(2006). Police crime hot spots. In Welsh

Farrington

(Eds.), Preventing crime: What works for children, offenders, victims, and places (pp. 179-193). Great Britain, Europe: Springer.

Braga

Weisburd

D. L.

(2007, November 28-29). Police innovation and crime prevention: Lessons learned from police research over the past 20 years. Paper presented at the National Institution of Justice (NIJ) Policing Research Workshop: Planning for the Future, Washington, DC. Retrieved from https://www.ncjrs.gov/

Byrne

Marx

(2011). Technological innovations in crime prevention and policing: A review of the research on implementation and impact. Journal of Police Studies, 20(3), 17-40.

Chan

J. B. L.

(2001). The technology game: How information technology is transforming police practice. Criminal Justice, 1, 139-159.

Chau

P. Y.

Tam

K. Y.

(1997). Factors affecting the adoption of open systems: An exploratory study. MIS Quarterly, 21, 1-24.

Diggle

Heagerty

Liang

K. Y.

Zeger

(2002). Analysis of longitudinal data (2nd ed.). New York, NY: Oxford University Press.

10.

Ericson

R. V.

Haggerty

K. D.

(1997). Policing the risk society. New York, NY: Oxford University Press.

11.

Federal Bureau of Investigation. (2001). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 1999 [Computer file]. Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation (2nd ICPSR ed.). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [producer and distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

12.

Federal Bureau of Investigation. (2002). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 2000 [Computer file]. Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation (ICPSR ed.). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [producer and distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

13.

Federal Bureau of Investigation. (2004). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 2002 [Computer file] (ICPSR04008-v1). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

14.

Federal Bureau of Investigation. (2005). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 2003 [Computer file]. Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation (ICPSR04124-v1). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [producer and distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

15.

Federal Bureau of Investigation. (2008). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 2006 [Computer file]. Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation (ICPSR22400-v1). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [producer and distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

16.

Federal Bureau of Investigation. (2009). Uniform crime reporting program data [United States]: Offenses known and clearances by arrest, 2007 [Computer file] (ICPSR25101-v2). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

17.

Fox

(2002). An R and S-plus companion to applied regression. Thousand Oaks, CA: SAGE.

18.

Groff

McEwen

(2008). Identifying and measure the effects of information technologies on law enforcement agencies: The making officer redeployment effective programs. Washington, DC: Office of Community Orientated Policing Programs.

19.

Hardin

J. W.

(2005). Generalized Estimating Equations (GEE). Hoboken, NJ: John Wiley.

20.

Harris

C. J.

(2007). Police and soft technology: How information technology contributes to police decision making. In Byrne

J. M.

Rebovich

D. J.

(Eds.), The new technology of crime, law and social control (pp. 153-183). Washington, DC: National Criminal Justice Reference Services.

21.

Harris

D. A.

(2010). Picture this: Body-worn video devices (heads cams) as tools for ensuring fourth amendment compliance by police. Texas Tech Law Review, 43, 357-371.

22.

International Association of Chiefs of Police. (2004). The impact of video evidence on modern policing. Washington, DC. Retrieved from http://www.cops.usdoj.gov/

23.

King

(1998). Innovativeness in American municipal police organizations (Doctoral dissertation). University of Cincinnati, OH.

24.

Koper

C. S.

Taylor

B. G.

Woods

D. J.

(2013). A randomized test of initial and residual deterrence from directed patrols and use of license plate readers at crime hot spots. Journal of Experimental Criminology, 9, 213-244.

25.

Langworthy

R. H.

(1986). The structure of police organizations. New York, NY: Praeger.

26.

Leip

(n.d.). Dave Leip’s Atlas of U.S. Presidential Elections. Available from http://uselectionatlas.org/

27.

Liang

K. Y.

Zeger

S. L.

(1986). Longitudinal data analysis using generalized linear models. Biometrika, 73, 13-22.

28.

Maguire

E. R.

(1997). Structural change in large municipal police organizations during the community policing era. Justice Quarterly, 14, 547-576.

29.

Manning

P. K.

(1992). Information technologies and the police. In Tonry

Morris

(Eds.), Modern policing: Crime and justice, a review of research (pp. 349-398). Chicago, IL: University of Chicago Press.

30.

Manning

P. K.

(1996). Information technology in the police context: The “sailor” phone. Information Systems Research, 7, 52-62.

31.

Manning

P. K.

(2003). Policing contingencies. Chicago, IL: University of Chicago Press.

32.

Mullen

(1996). The computerization of law enforcement: A diffusion of innovation study (Doctoral Dissertation). State University of New York at Albany.

33.

Pew Research Center. (2013). Few see adequate limits on NSA surveillance program. Washington, DC: Author. Available from http://www.pewresearch.org/

34.

Piza

E. L.

Caplan

J. M.

Kennedy

L. W.

Gilchrist

A. M.

(2014). The effects of merging proactive CCTV monitoring with directed police patrol: A randomized controlled trial. Journal of Experimental Criminology, 1-27. Advanced online publication doi:10.1007/s11292-014-9211-x

35.

Police Executive Research Forum. (2011, April 4). Use of technology in policing: The chief’s perspective: Critical issues in policing series. Washington, DC: Author.

36.

Police Executive Research Forum. (2012). How are innovations in technology transforming policing? Washington, DC: Author.

37.

Randol

B. M.

(2012). The organizational correlates of terrorism response preparedness in local police departments. Criminal Justice Policy Review, 23, 304-326.

38.

Ratcliffe

J. H.

(2008). Intelligence-led policing. Cullompton, UK: Willan.

39.

Ratcliffe

J. H.

Taniguchi

Taylor

R. B.

(2009). The crime reduction effects of public CCTV cameras: A multi-method spatial approach. Justice Quarterly, 26, 746-770.

40.

Robey

Sahay

(1996). Transforming work through information technology: A comparative case study of geographic information systems in county government. Information Systems Research, 7, 93-110.

41.

Rogers

E. M.

(2003). Diffusion of innovations (5th ed.). New York, NY: Free Press.

42.

Sanchez

T. W.

Lang

R. E.

Dhavale

D. M.

(2005). Security versus status? A first look at the census’s gated community data. Journal of Planning Education and Research, 24, 281-291.

43.

Schuck

A. M.

Rosenbaum

D. R.

(2008). Building trust and a police-community website: An assessment of CLEARpath planning and community engagement process (Final Report to The John D. and Catherine T. MacArthur Foundation). Chicago: Center for Research in Law and Justice, University of Illinois at Chicago. Retrieved from https://portal.chicagopolice.org/portal/page/portal/ClearPath/Internal/FinalCLEARpathReport.pdf

44.

Stone

Travis

(2011). Toward a new professionalism in policing (New Perspectives in Policing). Cambridge, MA: Harvard Kennedy School.

45.

Tabachnick

B. G.

Fidell

L. S.

(2007). Using multivariate statistics (5th ed.). Boston, MA: Pearson.

46.

U.S. Census Bureau. (2008). A compass for understanding and using American community survey data: What general data users need to know. Washington, DC: U.S. Government Printing Office. Available from http://www.census.gov/

47.

U.S. Census Bureau. (2014a). “Summary File.” 2005–2009 American Community Survey. U.S. Census Bureau’s American Community Survey Office. Available from http://ftp2.census.gov/

48.

U.S. Census Bureau. (2014b). “Summary File.” 2006–2010 American Community Survey. U.S. Census Bureau’s American Community Survey Office. Available from http://ftp2.census.gov/

49.

U.S. Department of Justice. (2006). Bureau of justice statistics. Law Enforcement Management and Administrative Statistics (LEMAS): 2003 Sample Survey of Law Enforcement Agencies (ICPSR04411-v1). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

50.

U.S. Department of Justice. (2008). Bureau of Justice Statistics. Law Enforcement Management and Administrative Statistics (LEMAS): 2000 Sample Survey of Law Enforcement Agencies (ICPSR03565-v2). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

51.

U.S. Department of Justice. (2011). Office of Justice Programs. Bureau of Justice Statistics. Law Enforcement Management and Administrative Statistics (LEMAS), 2007 [Computer file] (ICPSR31161-v1). Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor]. Retrieved from http://www.icpsr.umich.edu/icpsrweb/NACJD/

52.

Walker

(2001). Police accountability: The role of citizen oversight. Belmont, CA: Wadsworth.

53.

Walker

(2008). The neglect of police unions: Exploring one of the most important areas of American policing. Police Practice & Research: An International Journal, 9, 95-112.

54.

Willis

J. J.

Mastrofski

S. D.

Weisburd

(2007). Making sense of COMPSTAT: A theory-based analysis of organizational change in three police departments. Law & Society Review, 41, 147-188.

55.

Wilson

J. M.

(2005). Determinants of community policing: An open systems model of implementation (Report to the National Institute of Justice). Available from https://www.ncjrs.gov/

56.

Zhao

Scheider

M. C.

Thurman

(2002). Funding community policing to reduce crime: Have COPs grants made a difference? Criminology & Public Policy, 2, 7-32.