Content uploaded by Daniel S Lawrence
Author content
All content in this area was uploaded by Daniel S Lawrence on Sep 06, 2023
Content may be subject to copyright.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
Lessons learned implementing gunshot detection technology: Results of a
process evaluation in three major cities
Daniel S. Lawrence*, Nancy G. La Vigne, Margaret Goff, and Paige S. Thompson
Urban Institute, Justice Policy Center, Washington, DC 20037
*Corresponding author:
Daniel S. Lawrence
Justice Policy Center
Urban Institute
Tel: (202) 261-5309
E-mail address: dlawrence2@urban.org
Daniel S. Lawrence, PhD in Criminology, Law, and Justice, is a Senior Research Associate in
the Justice Policy Center at the Urban Institute. His research interests include police technology,
police legitimacy and procedural justice, police screening and hiring practices, and community
policing. Dr. Lawrence received his MA and PhD from the University of Illinois at Chicago. He
was a Co-Principal Investigator on this project.
Nancy La Vigne, PhD is vice president for justice policy at the Urban Institute. She publishes
research on prisoner reentry, criminal justice technologies, crime prevention, policing, and the
spatial analysis of crime and criminal behavior. Her work appears in scholarly journals and
practitioner publications and has made her a sought-after spokesperson on related subjects. She
was the Principal Investigator on this project.
Margaret Goff, MS is a Research Analyst in the Justice Policy Center at the Urban Institute.
Her research interests include incarcerated women, incarcerated parents and their children, and
sexual assault and domestic violence. Ms. Goff received her MS in Justice Studies from Arizona
State University.
Paige S. Thompson, MA is a Research Analyst in the Justice Policy Center at the Urban
Institute. Her research interests include community policing, policing technology, human
trafficking, and the intersection of mental health and the justice system. She received her MA in
Criminology, Law, & Society with a concentration in Policy and Practice from George Mason
University.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
Lessons learned implementing gunshot detection technology: Results of a
process evaluation in three major cities
Abstract
This paper describes the experiences and lessons learned from the implementation
of gunshot detection technology (GDT) in three US cities. Data were derived from
stakeholder interviews, community focus groups, and review of firearm-related
criminal case files. Findings indicate that stakeholders view GDT to generate
valuable investigative information, that officers are compliant with GDT response
and protocols, and that residents accept GDT despite low levels of trust and
confidence in the police. This paper concludes with recommendations for future
GDT implementations.
Keywords: police, gunshot detection technology, shotspotter, firearm
investigations, shootings
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
1
Introduction
Gunshot detection technology (GDT) is a relatively new tool for law enforcement agencies (LEAs)
to use in measuring, responding to, and investigating firearm shootings. The system uses a network
of outdoor acoustic sensors that are installed in selected (typically high crime) areas. When four
or more sensors detect gunfire, the system sends a signal to a processing algorithm that detects and
discriminates gunshots from other types of noises, such as fireworks or weather events, while
simultaneously computing spatial coordinates within 25 meters of the location where the shot was
fired (Aguiliar, 2015; ShotSpotter, 2018). GDT has been rapidly deployed by LEAs in the US,
with 16.2% of large agencies (250 or more sworn employees) reporting deployment of the
technology from 2012 to 2014 and another 4.0% expressing intent to implement the technology
between 2014 to 2016 (Strom et al., 2016).
Introduced with the promise of producing a new and more comprehensive metric of gun violence,
past studies have found GDT to increase accuracy of data on where and when gunfire is occurring,
improve the speed of response times to scenes of potentially violent gun crime, and aid in
investigations by increasing the likelihood of identifying suspects, witnesses, and evidence from
gunfire incidents that may go unreported without the technology’s detection (Carr and Doleac,
2016; Choi et al., 2014; Irvin-Erickson et al., 2017; Mazerolle et al., 1998; 2000; Watkins et al.,
2002); although, studies have also found GDT to increase officer workloads (Mazerolle et al.,
1998; Ratcliffe et al., 2018). Yet despite GDT’s increased implementation across the country, no
research to date has conducted an evaluation of how the technology is implemented and used by
police. This information is critical, as past research indicates that the success of policing
technologies lies more in the way they are implemented than in the accuracy of the technology
itself (Koper et al., 2015).
The current study documents the implementation processes and challenges associated with
ShotSpotter technology, a GDT manufactured by ShotSpotter, Inc., and describes the opportunities
for improvement in the technology’s use and impact. We explore several research questions
focused on the contextual and operational issues associated with implementing GDT. These
include: Why is GDT implemented and what policies are created for its use? How are GDT-
generated data integrated into LEA systems and combined with other data and resources? How is
the technology used in investigations? What is the agency/organizational culture surrounding GDT
and its use? How are community members informed of the technology and what are their views of
its purpose and effectiveness? These questions are answered through qualitative analyses of
findings from interviews with criminal justice stakeholders, focus groups with community
members, and quantitative case file reviews of firearms-related crimes in Richmond, California;
Milwaukee, Wisconsin; and Denver, Colorado.
Gunshot Detection Technology
GDT consists of a network of acoustic sensors strategically placed throughout an area with high
levels of firearm crimes. Sensors are installed in high locations (e.g. on top of buildings or light
poles) with unobstructed paths to other nearby sensors to improve triangulations of identified
gunshots. The technology has the capability of detecting whether single or semi-automatic shots
are fired, which aids police in the investigation of these crimes (Watkins et al., 2002). The sensors
themselves are nondescript and discrete so that the public cannot identify or tamper with them in
any way.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
2
In the case of ShotSpotter, Inc.’s more recent Flex system GDT, alerts on the detection, location,
and identification of gunshots are sent to human operators at ShotSpotter headquarters within a
few seconds after the shot is recognized by the system (Aguilar, 2015; ShotSpotter, 2018). These
individuals are trained to identify and screen out incidents that were not recorded accurately and
to confirm true gunfire incidents, thus ensuring the information that is sent to a LEA’s computer
aided dispatch (CAD) system is accurate (ShotSpotter, 2018). Under the Flex system, which
includes yearly subscription costs based on the size of the coverage area, ShotSpotter, Inc.
maintains and repairs sensors, as necessary, and routinely provides updates to its software. Prior
to the Flex system, which began rolling out in 2012, LEAs purchased the sensors and software
from ShotSpotter, Inc. and owned the equipment (hereafter referred to as the legacy system). As
such, the LEA was responsible for its upkeep and managed whether dispatchers were trained to
listen to and recognize gunshots from the audio recordings prior to assigning officers to respond.
The field of research on the impact of GDT is still new and expanding but research indicates that
GDT systems are improving in terms of their accuracy of documenting the time and location of
the gunshots. Research shows that the technology generates both few false negatives (failure to
detect gunfire) and few false positives (detecting a noise as gunfire that is not; Chacon-Rodriguez
et al., 2011; Selby et al., 2011). An early study conducted by Mazerolle et al. (2000) found that
ShotSpotter detected nearly 8 in 10 test shots that were fired and was able to accurately locate
those test shots within a margin of error of 41 feet. A 2006 study in Charleston, South Carolina
found that ShotSpotter detected 99.6% of the 234 gunshots detected across 23 different firing
locations and that it located 90.9% of those shots within a 40-foot margin of error (Goode, 2012).
More recently, a 2017 study by Irvin-Erickson and colleagues explored the temporal and distance
limitations of GDT, finding that the relative sensitivity of detections from GDT was strongest in
the evening and nighttime compared to daytime hours, and that the sensitivity of the detections
from GDT weakens as the distance from the center of known coverage areas increases.
The observed impacts of GDT have been significant, including a reduction in officer response
times (Mazerolle et al., 1998); the collection of crucial evidence such as the location of the source
of gunfire and the presence of suspects, witnesses, or evidence (Choi et al., 2014); and the
provision of comprehensive reporting rates of gun crime (Carr & Doleac, 2016). Data from GDT
have also been assessed to be a more accurate metric of quantity, timing, and location of gunfire
incidents in comparison to traditional 911 calls for service data, which yield estimates of reported
gunfire that are only a small percentage of the total gunfire events detected by GDT (Carr &
Doleac, 2016; Irvin-Erickson et al., 2017). Correct implementation and standardization of the
technology may reduce officer response times to gunfire incidents by approximately 7% (about 1
minute) while also increasing police effectiveness by reducing the time it takes to dispatch calls
for service (Choi et al., 2014; Mazerolle et al., 1998). These findings indicate that GDT may
provide LEAs with information that can aid in prompt response to gunfire incidents, enabling rapid
connection to medical care for victims and more immediate identification and apprehension of
suspects. Furthermore, the body of research on GDT systems suggests that the key to the
technology’s impact is the quality of implementation and the level of integration with other crime
control and prevention systems in LEAs, such as CCTV (La Vigne et al., 2011; Choi et al., 2014;
Litch & Orrison, 2011; Ratcliffe et al., 2018).
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
3
The above review of the literature provides a foundation from which to lay forth a logic model of
GDT implementation and use. Figure 1 identifies a broad range of potential uses for the system,
including responding to incidents, supporting crime analysis, facilitating improved prosecutions,
and engaging in community education and outreach. Shaded cells indicate areas outside the subject
of this article. The inputs of a GDT may be few—shooting alerts, software on computers and
smartphones, connection to other public surveillance systems, and departmental policies and
procedures—but the potential impacts are many and significant. GDT has the potential to increase
LEAs’ awareness of gunshot activity; reduce response time; and provide investigative leads such
as the approximate location of the gunshot activity, suspects, witnesses, and shell casings (Choi et
al., 2014). Using GDT, LEAs can more accurately identify the time and location of gunshots,
thereby increasing the likelihood that officers will apprehend suspects, assist victims in receiving
timely medical care, and improve the investigation and prosecution of firearms crimes. These
efforts are further improved through in-depth crime analyses of the GDT data that allow LEAs to
implement proactive, targeted patrols in areas with high rates of gun violence.
Figure 1. Logic Model of Gunshot Detection Technology Implementation and Use
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
4
The logic model raises a series of research questions around GDT deployment, policy development
and training, usage, and departmental and community engagement. The main research questions
examined in this study include:
(1) Why is GDT implemented and what policies are created for its use?
(2) How are GDT-generated data integrated into LEA systems and combined with other
data and resources?
(3) How is the technology used in investigations?
(4) What is the agency/organizational culture surrounding GDT and its use?
(5) How are community members informed of the technology and what are their views of
its purpose and effectiveness?
Within the first research question, we were interested in learning what factors were used to
determine the locations of GDT sensors, what LEA leaders hoped to accomplish with GDT, and
what were the trainings, policies, and procedures associated with GDT. The second research
question focuses on the data that GDT produces and how it is used by the LEA; for example, how
the data are analyzed and used with other resources to detect patterns and trends in time and
location of gunfire incidents. GDT can be viewed as the catalyst to a shooting investigation; as
such, the third research question focuses on how its use in-and-of itself affects investigations and
how it is used with other techniques to support investigations. The fourth research question focuses
on such things as whether executive staff communicated their expectations of the technology and
its use to patrol officers, officers’ confidence in the technology, and whether officers are held
accountable for complying with GDT trainings, policies, and protocols. Finally, the fifth research
question focuses on how LEAs inform community members about the technology, where it is
deployed, how it works, and whether community members are supportive of the LEA’s use of
GDT to reduce firearm violence. These questions are answered through qualitative analyses of
findings from interviews with criminal justice stakeholders, focus groups with community
members, and quantitative case file reviews of firearms-related crimes in Richmond, California;
Milwaukee, Wisconsin; and Denver, Colorado.
Methods
Study Sites
This research is part of a larger evaluation led by the Urban Institute (Urban) and funded by the
National Institute of Justice. The study aimed to investigate the degree to which GDT aids in the
response, investigation, and prevention of firearms violence and related crime. To help inform the
selection of impact evaluation sites and guide the development of process evaluation questions,
Urban conducted a web-based survey of LEAs that were using the ShotSpotter GDT in May 2014.
The objective of the survey was to review GDT policies, practices, and expectations for LEAs
using the ShotSpotter system. The survey results and subsequent conversations with GDT end-
users led Urban to partner with the LEAs in Richmond, CA; Milwaukee, WI; and Denver, CO as
evaluation sites. Insights from the survey revealed that these departments used GDT for a wide
array of practices likely to maximize the technology’s effectiveness as a violence prevention tool,
from incident investigation to community outreach. These departments also indicated a belief that
GDT was having a significant impact on a wide array of performance metrics, including violence
reduction, officer safety, and community engagement, making them ideal sites in which to
investigate the full range of effects of this technology.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
5
The three study sites vary considerably in terms of size, geographic location, and demographics.
Richmond has a population of roughly 107,000 people with a large (40.2%) Hispanic population
and smaller African American and white populations (21.7% and 17.6%, respectively; U.S. Census
Bureau, 2018). The city has experienced persistently high rates of crime: analyses of calls for
service data from January 2006 to October 2015 found that the average number of shooting-related
calls for service per month in Richmond was 74.13 (SD = 43.13). Richmond was an early adopter
of GDT, deploying the technology in June 2009. Six months later, in January 2010, the department
added additional sensors, expanding the original area and adding a second coverage area within
the city, which resulted in a total coverage of 5.69 square miles (as of December 2016). In July
2013, Richmond upgraded from the legacy system to ShotSpotter’s Flex system. The average
number GDT alerts per square mile each month was 32.48 (SD = 22.24) from June 2009 to October
2015. In the same area and time period, the average number of shooting-related calls for service
per square mile each month was 6.02 (SD = 2.63).
Milwaukee has a population of approximately 600,000 people with roughly equal shares of white
and African-American populations (35.9% and 38.8%, respectively), and a Hispanic population of
approximately 18.2% (U.S. Census Bureau, 2018). From January 2008 to December 2016, the
average number of shooting-related calls for service per month in Milwaukee was 848.31 (SD =
298.14). Milwaukee first deployed its GDT in February 2011, with two subsequent expansion
periods: one in which the original coverage area was expanded in April 2012 and another in August
2014 when the coverage area was again expanded to a new coverage area elsewhere in the city.
The city’s GDT covered a total of 12.68 square miles once these expansions were complete. The
department also upgraded from the legacy system to ShotSpotter’s Flex system during the April
2012 expansion. The average amount GDT alerts per square mile each month was 64.37 (SD =
22.61) from March 2011 to December 2016. In the same area and time period, the average number
of shooting-related calls for service per square mile each month was 37.39 (SD = 15.46).
Denver has a population of roughly 660,000 people with just over half (53.4%) of the population
identifying as white, a substantial Hispanic population (30.8%), and a much smaller (9.4%)
African American population compared to the other two sites (U.S. Census Bureau, 2018). From
January 2008 to June 2016, the average number of shooting-related calls for service per month in
Denver was 304.34 (SD = 81.39). Denver first deployed GDT in January 2015. The department
added two additional coverage areas in April 2016 and September 2016, resulting in GDT sensors
covering 11.54 square miles of the city. Due to the late adoption, Denver’s GDT has always
operated under the Flex system. The average number of GDT alerts per square mile each month
was 13.87 (SD = 5.06) from January 2015 to May 2016.1 In the same area and time period, the
average number of shooting-related calls for service per square mile each month was 12.10 (SD =
5.91).
Data Collection Activities
Three main research activities were conducted as part of the process evaluation within each of the
evaluation sites. They included: 1) semi-structured interviews with criminal justice stakeholders;
2) focus groups with community members; and 3) review of police case files of firearm-related
crimes. Table 1 details the final sample sizes for each of these collection efforts. Research staff
also attended live-fire tests of new acoustic sensors for a new coverage area in Denver,
observations of which are included in the evaluation below.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
6
Interviews were conducted over a period of nine months, from November 2016 through July 2017,
using a semi-structured interview protocol with open-ended and probing questions. Law
enforcement agency stakeholders included sworn staff from patrol officers to assistant chiefs.
Civilian employees of the LEA who were interviewed included crime analysts and communication
managers. Finally, staff from the city prosecutor’s office and the Bureau of Alcohol, Tobacco,
Firearms and Explosives (ATF) were interviewed as criminal justice partners with experience
using the GDT but not part of the local LEA. The protocols were tailored to align with the role of
the person being interviewed (e.g., investigator, command-level officer, line officer); covering
issues on planning GDT implementation, acquisition, installation and monitoring, policies and
procedures, training, use on the ground and investigations, and perceived value and impact of
GDT. The research team visited each site to conduct hour-long, in-person interviews with one
stakeholder at a time. One Urban researcher and one staff member from Police Foundation, who
was previously a deputy chief at a large LEA, conducted the interviews while another Urban staff
member typed notes. Participants were recruited in partnership with the research team’s point of
contact in each site, but were given the option to decline participation during the informed consent
process (although no one did decline).
Table 1: Data collections by type and site
Hour-Long Stakeholder Interviews
Milwaukee
Richmond
Denver
Total
LEA Sworn Stakeholders
10
13
9
32
LEA Civilian Stakeholders
3
2
3
8
Criminal Justice Partners
3
1
2
6
Total
16
16
14
46
Community focus groups
Milwaukee
Richmond
Denver
Total
Participants in focus group located
within an area covered by GDT
12
9
10
31
Participants in focus group located
within an area not covered by GDT
5
9
4
18
Total
17
18
14
49
Firearm-Related case file coding
Milwaukee
Richmond
Denver
Total
Pre
Post
Pre
Post
Pre
Post
Aggravated Assault
9
10
12
12
12
14
69
Homicide
3
3
3
3
2
3
17
Robbery
5
5
12
12
0
0
34
Weapon Law Violation
13
12
2
2
12
13
54
Total
30
30
29
29
26
30
174
In addition to stakeholder interviews, the research team conducted focus groups with community
members to learn residents’ views on how their local LEA responds to firearm-related crime and
to assess resident knowledge of the presence, purpose, and use of GDT. Two 90-minute focus
groups were conducted within each site, one inside the GDT coverage area and one outside the
coverage area; although neither group knew the exact coverage of GDT in their city. The
neighborhoods for the focus groups in areas not covered by GDT were selected based on having
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
7
similar crime rates and demographic characteristics as the coverage areas. Participants were
recruited in partnership with local community-based organizations and received $50 for their time
and insights. To take part in the focus group, participants had to live in the neighborhood and be
18 years or older. Participation averaged 8.2 people per focus group, and recruiting efforts attracted
groups ranging in racial/ethnic identity, age, and socioeconomic background, and number of years
living in the community. To maintain anonymity and protect the names of participants, verbal
consent was obtained prior to participation and names were not recorded in written notes.
The third category of data collection focused on the coding of 174 randomly selected case files
that pertained to firearm offenses. Half of these case files occurred before the implementation of
GDT and the other half occurred after. To ensure cases represented an array in the types of offenses
likely to be associated with firearms use, the research team randomly selected approximately 60
case files across the following offense categories: weapon violation, robbery, aggravated assault,
and homicide offenses.2 Sampling was stratified across these offense types in proportions that
approximated each site’s frequency of these offenses; however, homicide cases were slightly over
sampled to ensure a sufficient number of cases were included in the coding process.
Analyses
To analyze the data, the notes from the interviews and focus groups were cleaned (to correct typos
and misspellings) and uploaded into NVivo 10 software. Inductive and deductive coding methods
were used to develop a coding scheme, which was independently corroborated by three different
members of the research team (Fereday & Muir-Cochrane, 2006). Inductive codes were developed
based on the study’s research questions and logic model. Deductive codes were developed using
an open coding method; researchers ran a word frequency in NVivo 10 to check if any major
themes or elements of the notes had been overlooked or excluded from the coding scheme. To
ensure interrater reliability, four randomly selected interviews from across the three sites were
coded separately by two researchers. The researchers met and reconciled any differences in coding
decisions before proceeding to code the rest of the interviews. Researchers ran queries in NVivo
10 to further examine the identified themes.
The case file coding scheme was developed in partnership with the retired deputy chief from the
Police Foundation and guided by the experience and expertise of the research team. Three
researchers tested the coding scheme on a randomly selected sample of four case files. Each
researcher coded all four case files using the coding scheme, and then researchers convened to
reconcile differences by either adding or subtracting variables to the coding sheet. The same coding
sheet was used to code the case files in all three agencies, with case files in each agency assigned
randomly to each coder. After completing the case file coding, the data were combined and
analyzed using SPSS statistical software.
Results
Review and analyses of the notes and data from the interviews, focus groups, and case files
generated both quantitative and qualitative data with which to answer our five research questions,
covering such topics as how the three study sites planned for, acquired, and deployed GDT (and
GDT-generated data) in their efforts to respond to, investigate, and prevent firearms violence and
related crime, and how the community was engaged in the deployment and use of GDT.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
8
Research Question 1: Why is GDT implemented and what policies are created for its use?
According to stakeholders across all three study sites, GDT was viewed as a potentially useful tool
for generating a new metric of gun violence—one that could provide more accurate data on when
and where gunfire is occurring—which in turn could lead to faster response times, increase the
likelihood of apprehending suspects and identifying witnesses, and aid in investigations through
recovery of evidence. The potential of this technology was particularly germane for these cities
because each was experiencing high levels of group or gang-related firearms violence.
Initial investment in the technology was facilitated by grant funds, with Denver benefitting from a
Justice Assistance Grant awarded through the Bureau of Justice Assistance’s federal formula grant
program to pay for its first GDT deployment. Similarly, Milwaukee’s original investment in GDT
was paid for by a grant award from the Office of Community Oriented Policing Services, US
Department of Justice. Richmond’s GDT investment was paid for through a combination of state
asset forfeiture funds and city resources. In addition, Denver’s Crime Gun Intelligence Center,
which supports the use of GDT-generated investigative data, was funded by a collaborative
agreement with the ATF, which assigned three ATF investigators to assist the Firearms Unit and
National Integrated Ballistic Information Network (NIBIN).
Once funding was secured, the next step for LEAs was to determine where to place the sensors.
Given the high violent crime rates experienced by each jurisdiction, the acquisition of GDT was
viewed by law enforcement stakeholders as a well-justified investment, with placement of sensors
focused on the neighborhoods with the largest volume of gun-related crime based on analyses of
historical crime data. As discussed later in this article, outreach to communities in which GDT was
deployed was conducted in advance of installation in all three cities. Law enforcement
stakeholders logically concluded that deploying expensive GDT sensors in areas where firearm
shootings did not occur would not make fiscal or strategic sense. However, LEAs took care to keep
the locations of coverage areas confidential in the belief that gun violence would more likely
decrease city-wide if perpetrators believed the entire city was covered by GDT, and if perpetrators
knew the coverage of GDT then they would simply move to an area that was not covered.
After sensors are installed, LEA and vendor personnel tested the GDT using live rounds in
numerous locations within the coverage area. Researchers on this project observed such testing
when the Denver Police Department expanded its coverage area during the study period. To ensure
community cooperation and understanding of the tests, Denver staff mailed and posted flyers
around test locations, posted on social media, and responded to 911 calls for shots fired. These
tests occurred late at night into the early morning hours with a large team of patrol officers, SWAT
officers, supervisors of the GDT program, and a GDT technician who was responsible for
installation and maintenance of the sensors.
In terms of policy guidance and training curricula on GDT use, our study sites represent relative
early adopters of GDT, and as such, they faced the challenge of implementing a tool with little
documentation around effective ways to utilize it. In the case of Milwaukee, much of the policies
developed for GDT were created after initial deployment of the technology, leading to variations
in the ways in which different officers responded to alerts and made use of GDT-generated data.
For example, officers expressed confusion about whether and how they should be responding to
the GDT alerts received by the software in their vehicle versus those issued by dispatch. In
addition, some, but not all personnel were aware that GDT alerts were also accessible on their
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
9
phones via mobile application or that they could sign up to receive emails about alerts. The lack
of a standardized operating procedure during early use of GDT had left officers unsure about what
do in certain cases and likely led to under-utilization of the technology.
Overall, law enforcement personnel receive different levels of GDT training based on how closely
they interacted with the technology, and across all study sites training was more intensive during
initial deployment but diminished in frequency and consistency over time. Staff who operate
within research units received detailed training on how to read GDT alerts, use the software’s
Investigator Portal, and request “Detailed Forensic Reports” from the vendor. These staff were
those who routinely work with agency data to assess crime trends and produce reports, such as
crime analysts and those who supervise the GDT for their agency. Staff who were directly
responsible for managing the GDT program reported having participated in more extensive
trainings, such as a 40-hour long course held at ShotSpotter headquarters. This curriculum
provided instruction on use of the Investigator Portal, which allows agency analysts to investigate
single events in further detail or download bulk data to analyze GDT alert trends. Crime analysts
represented in stakeholder interviews expressed the belief that this detailed level of training is
crucial for incorporating GDT data into tactical and COMPSTAT reports and maximizing the
analytical power of GDT systems.
Personnel without direct research or oversight of the GDT program, such as patrol officers,
received rudimentary training on how to log-in and access the Respond Portal from their mobile
data terminals or smartphones, as well as guidance on how to use GDT alert information to respond
to recent shootings. The Respond Portal provides simplified information about recent shooting
events so that patrol officers can view recent gunfire activity and be alerted to new gunfire activity
as soon as it is recognized by the system. New patrol officers typically go through a very short and
somewhat informal 20-minute training on how to respond to GDT alerts during their training in
the police academy. New staff in civilian positions expressed that training on the use of GDT was
not part of their orientation or onboarding process.
Across all three LEAs, GDT alerts are categorized as Priority 1 calls by each agency’s dispatch.
However, the means transmitting alerts to dispatch varied by jurisdiction. In Denver and
Richmond, the GDT alerts are integrated with the LEA’s CAD system automatically after it is
deemed a valid gunshot by operators at ShotSpotter’s headquarters. In Milwaukee, GDT alerts are
communicated to patrol officers via GDT software on their mobile data terminals as well as on
agency-issued smart phones. At the same time, alerts are also sent to a computer within the
department’s communication division, where a sergeant reviews and manually enters the alert data
into the CAD. As such, officers could know about the alerts from the software and begin
responding before the information is entered into the CAD. While some Milwaukee sergeants
advised that they prioritize the processing of GDT alerts associated with addresses where past
alerts had been previously detected, the time lapse between the patrol car alert and CAD entry
created confusion for officers, who reported uncertainty about the urgency of GDT alerts and
whether and how to prioritize them.
Regardless of the mechanism employed to convey a GDT alert, officers expressed clarity on the
protocol to follow when responding to it, which was similar across the three agencies. Consistent
with training associated with responding to any violent call for service, all three LEAs require
officers responding to GDT alerts to approach the scene cautiously, to ensure no persons with
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
10
firearms are present, and to secure the scene. Once it is deemed safe to investigate, officers are
required to exit their vehicles, search for shell casings, and conduct a canvass of the area to find
witnesses and collect evidence.
One of the purported benefits of GDT is its ability to enhance investigations by pinpointing the
exact location of a crime scene, increasing the odds of recovering evidence. In the interests of
meeting that goal, Milwaukee officers are issued a shovel and flashlight, and in Denver, where
snow can interfere with the recovery of shell casings, metal detectors are supplied. Denver also
trained its K-9 units for shell casing recovery purposes. In the case of inclement weather or low
visibility, responding officers are required to return to the scene the following day to do an
additional search of the area for shell casings.
Analyses of the change in outcomes associated with the coded case files suggest a high degree of
fidelity to patrol officer GDT response protocols. Perhaps not surprisingly, the mention of a GDT
alert in the case files increased from 10.6% prior to the technology being deployed to 48.3%
afterwards (t(172) = -5.93, p < .001).3 This increase was largest in Denver, where officers reported
GDT alerts in 63.3% of the cases in the time period following GDT deployment, from 0% in the
time period before deployment (t(54)= -6.58, p < .001).
Specific to canvassing the neighborhood of a shooting, aggregate data from the three sites showed
an increase from 57.6% to 70.8% in the number of cases where a canvass was conducted from the
period before GDT implementation compared to the period following (t(172) = -1.82, p < .10). In
line with the increased canvasses, the total number of people interviewed as part of the case, which
included victims, suspects, and witnesses, increased from 1.95 people to 2.32 people from before
and after GDT implementation, but this difference was not statistically significant. However,
results specific to Denver indicate that the number of people interviewed significantly increased
from an average 0.08 people during the pre-GDT period to an average of 1.67 after GDT
deployment (t(54) = -1.68, p < .10). The number of victims interviewed in Richmond significantly
increased from an average of 1.17 to 1.62 from before and after GDT deployment, respectively
(t(56) = -2.30, p < .05).
Research Question 2: How are GDT-generated data integrated into LEA systems and
combined with other data and resources?
A single GDT alert produces much more than just a CAD event for an officer to respond to. With
each alert, data are collected on the date, time, location, the number of gunshots identified, linked
CAD numbers and local street addresses, and comments entered by the GDT operators, which can
include whether the gunshots appear to have been fired from a moving vehicle, if multiple shooters
were involved, and more details about the location of the shooting (e.g., from on top of a building).
The software also stores a 4-second audio clip from each acoustic sensor that triangulated the
gunshot.
All three of the evaluation sites employ these data for short-term, tactical decision-making in the
form of reports compiled by crime analysts. Each week, crime analysts map the GDT data onto
patrol beats and share the results with district captains as well as with sergeants and other officers
either by email or during rollcalls so officers can be aware of shooting hot spots. Crime analysts
also use the GDT data to generate routine COMPSTAT reports, to support grant applications in
describing trends in firearms violence (e.g., Project Safe Neighborhoods funding application), and
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
11
occasionally in search warrants in combination with other evidence.
In addition, prosecutors reported using GDT data as evidence in court. This could include listening
to the audio files of the shootings, which sometimes include people shouting or yelling names, or
using GDT data on the number, location, timing, and type of gunshots to prove intent or refute
defendants’ claims of self-defense. Detectives could also request “Detailed Forensic Reports” from
the GDT vendor. These reports provide a much more detailed account of the shooting event and
include maps, distance measurements, narratives of the shooting, and methodologies of the
triangulation procedures the technology employs. However, prosecutors stressed that GDT-
generated evidence on its own was insufficient to successfully prosecute, noting that the value of
the technology was in producing data that could be used in concert with other evidence.
The data generated by GDT not only informs officer responses but it also creates information that
a LEA must store, manage, and integrate into its systems. As with any new technology, integration
of GDT data may identify problems with existing system infrastructures. For example,
stakeholders expressed that GDT alerts were not always automatically integrated into the CAD
system, leaving dispatchers to manually enter alert data. While this process applies more to the
legacy system of GDT (pre-2012), Milwaukee still requires sergeants in its communication
division to manually enter GDT alerts into its CAD. However, even ShotSpotter’s Flex system
does not automatically include descriptive information in the CAD logs, such as if the shots seem
to be coming from a moving vehicle. Such information has to be communicated to responding
officers from dispatchers over the radio or reviewed directly from the GDT software. This lack of
data integration can result in delayed response times, might put officers at risk, and could
compromise the accuracy of the data due to manual entry, which presents the potential for entry
errors.
Another challenge with the data generated by GDT is that the increased awareness of gunshots and
influx of evidence can be overwhelming to agencies, particularly those with limited personnel
resources. A crucial aspect to maximizing the utility of GDT is having sufficient personnel to
respond to alerts in a timely manner, process evidence, and act upon the intelligence generated by
that evidence. Law enforcement stakeholders expressed concerns that their agency did not have
enough capacity to handle the volume of work created by its GDT. In Milwaukee, the Rapid
Response Unit that was initially dedicated to responding to GDT alerts was ultimately disbanded
because the agency’s expansion of GDT coverage areas over time made it untenable to keep up
with the volume of alerts generated from the larger geographic area.
The new volume of data generated by GDT also impacts the workloads of crime analysts. GDT
program supervisors in Milwaukee reported resorting to manual tracking of individual GDT alerts
and associated case outcomes such as arrests and case clearances, which is done as time permits –
sometimes years after the incident of interest. Staff in Denver expressed similar challenges, citing
an inability to link GDT data with other types of agency data, which had proved to be a challenge
in supporting investigations and building intelligence. Indeed, across all agencies it was apparent
that without the proper internal capacity, it is difficult to use the data and evidence GDT systems
produce for proactive (preventative) as opposed to reactive (response-driven) measures. The utility
of GDT systems is maximized when departments use the data and evidence it creates to analyze
trends, follow leads, and recognize patterns. Responding to alerts in a timely manner is important
and has its own benefits (i.e. responding to victims faster), but is only one component of GDT
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
12
usage.
Research Question 3: How is GDT data used in investigations?
Stakeholder interviews documented the ways in which GDT alerts can yield valuable evidence—
from shell casings to witness identification—based on its geographic accuracy and details of
shooting events that it generates. But from an investigative standpoint, the most promising aspect
of GDT identified through this implementation evaluation is its integration with other investigative
tools, such as the ATF’s NIBIN and its firearm eTrace program. GDT identifies the location of
gunfire thus increasing the likelihood of finding shell casings. Investigators can then run casings
through the NIBIN program to identify other shooting events with matched casings.
The three study agencies had slightly different procedures for analyzing shell casings collected
from GDT alerts. Both Milwaukee and Denver had the resources and infrastructure to perform
rapid testing of the casings that produced results within 24 hours, and both had a policy that all
recovered shell casings must be submitted for analysis. Furthermore, both agencies partnered with
the ATF to have all shell casing data entered into the NIBIN, which allows searches of matched
shell casings from firearms across a national database. NIBIN enables investigators to identify
whether the same gun had been used in multiple offenses based on matched shell casings. In
Milwaukee, certified ballistics experts were able to analyze the casings on site within the
department’s Intelligence Fusion Center, while the Denver Police Department relied on its recently
developed Crime Gun Intelligence Center within its crime lab.
Richmond, however, did not have access to ballistics experts within its department and therefore
relied on the county’s crime lab for ballistic analyses. Because of this, the department policy did
not require all shell casings recovered from GDT alerts to be submitted for analysis but instead
they could be submitted at the discretion of the investigator assigned to the case. As a result of
fewer casings being analyzed and entered into the database, fewer cases were connected to other
shooting incidents. Furthermore, the county’s crime lab analyses took weeks or months due to its
backlog of work, which was not ideal for active investigations. Stakeholders mused that while
investigators wait for the ballistic reports, a suspect could have committed additional crimes or
fled the area, witnesses’ memories might fade, and other leads could be lost.
As mentioned earlier, Milwaukee had a robust set of technologies from which to draw to build
stronger investigations. While Milwaukee stakeholders uniformly underscored the importance of
complementary evidence-enhancing technologies to leverage GDT, they also emphasized that the
addition of GDT to the department’s toolkit had the unintended impact of straining the existing
infrastructure and the staff who manage it. Specifically, as the volume of recovered shell casings
increased due to more accurate shooting-related notifications via GDT, NIBIN staff became
overwhelmed by the evidentiary backlog. As such, while the volume of potentially significant
evidence expanded, the speed at which it was used and connected to other cases slowed. Moreover,
while in-house NIBIN and eTrace programs expedite the entry and linkage of shell casings across
shootings, and can motivate officers to exit their vehicles to search for casings and weapons, the
ability to maintain a NIBIN program absent collaboration with ATF or another agency is not
feasible for many departments. In-house ballistics testing requires specially trained and certified
personnel, expensive equipment, and a data management capacity. Relying on county programs,
as in the case with Richmond, can result in serious delays or the inability to connect cases due to
the investigator’s discretion of which casings are submitted.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
13
Despite the additional burdens and backlogs associated with GDT integration with ballistics testing
and gun tracing programs, it remains the case that GDT-generated data holds promise for
increasing the recovery of shell casings and weapons, the former of which can then be run through
NIBIN to link to other cases and the latter through eTrace to identify past owners of the weapon
whom investigators can then interview. Results from the coding of the case files somewhat
corroborates this narrative, finding a marginally significant increase in whether bullet shell casings
were found at the scene from before GDT implementation (55.3%) compared to after (68.5%;
t(172) = -1.81, p < .10). This increase was largest in Richmond, where the number of cases that
included the retrieval of shell casings increased from 24.1% prior to GDT implementation to
58.6% after implementation (t(56) = -2.80, p < .01). The change in the number of cases where shell
casings were found in Milwaukee and Denver was not statistically significant. Similarly, the
change in the number of cases resulting in an arrest and those for which a weapon was recovered
at the scene was not significant either within or across sites.
When separating the cases by the type of crime, we observed that homicide and robbery cases had
marginally significant increases in the proportion of cases where shell cases were retrieved, while
the differences for aggravated assaults and weapon law offenses also saw increases but were not
significantly different pre- and post-GDT implementation. Across the three sites, 50.0% of
homicide cases involving a firearm were noted to have retrieved shell casings in the time period
prior to GDT, and this increased to 88.9% after the implementation of GDT (t(15) = -1.82, p <
.10). An even larger change was noted for robberies, where the proportion of cases where shell
casings were retrieved increased from 11.8% to 41.2% before and after GDT (t(32) = -2.00, p <
.10).
Research Question 4: What is the agency/organizational culture surrounding GDT and its
use?
As documented in other literature on police use of technology, fidelity of use is closely tied to the
department culture of its acceptance and officer belief of its utility (Koper et al., 2015). In the case
of GDT implementation in the three study sites, some stakeholders reported that there was a
discernable gap in knowledge between command officers and patrol officers on the technology
because senior personnel were privy to the plans for GDT deployment well before officers of lower
ranks were informed about it or trained on its use. Interviews with line officers confirmed this
view, with officers reporting limited knowledge of or training on GDT outside of the units that
worked with the technology every day. Relatedly, stakeholders shared that initial training on GDT
was fairly robust but not routinely administered as the technology matured. The uneven training
of line officers prevented them from being educated about improvements in technology over time.
For example, both Richmond and Milwaukee were both early GDT adopters under the legacy
system, which did not include verification of gunfire alerts by trained ShotSpotter staff. When the
agencies transitioned from the legacy to the Flex system, which verifies gunfire alerts at
ShotSpotter headquarters, no additional training of line officers was conducted. This may explain
the mixed findings from stakeholder interviews of sworn officers with regard to their confidence
in the system. Officers in Richmond were least likely to express confidence in the accuracy of
GDT, in part due to prominent gun violence events in the past that were not detected by the
technology despite being in the coverage area or noting that the system would occasionally be
inoperable during storms. Denver personnel had notably higher levels of confidence in the system,
perhaps because they adopted the Flex system from the start, rather than the less reliable legacy
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
14
system.
Despite these uneven findings, higher-ranking personnel that were interviewed across all three
sites were much more likely to support and endorse GDT, and prosecutors expressed high
confidence in its accuracy. However, stakeholders shared that communication about GDT between
command-level staff to patrol and other units was often poor in Milwaukee, and that in Richmond
new recruits no longer receive GDT training. Insufficient training can compromise officer buy-in
because officers might not understand how GDT works and how to maximize its utility in the field
or for investigations. Inadequate training can also contribute to inconsistent adherence to
departmental policies and procedures regarding GDT across different teams and units.
Fortunately, higher ranking officers appear to have a clear understanding of the technology and
the policies and procedures that were developed to guide its use. These include accountability
mechanisms. In Denver and Milwaukee, sergeants with GDT-related oversight conduct additional
checks on their officers to confirm that canvasses following gunfire alerts are conducted in
accordance with policy. For example, even if a patrol officer cleared a GDT alert as “unfounded,”
a sergeant might return to the scene to search for shell casings and talk with residents about the
gunfire and whether an officer spoke with them. And in Milwaukee, if patrol officers do not
respond to an alert within 24 hours, a supervisor is sent out to conduct a canvass in their stead.
These accountability mechanisms may explain why—even in the absence of in-depth GDT
trainings and despite misgivings on the technology’s accuracy—officers of all ranks expressed the
belief that GDT alerts are superior to more traditional means of gunfire notification, such as 911
calls for service from community members. Many of the officers interviewed believed that calls
for service were notoriously unreliable for providing the exact location of a shooting incident
because community members may have heard a gunshot but typically do not know from what
direction it originated or from how far away.
While officer reliance on the technology may be viewed as a generally positive finding,
particularly if the technology is accurately pinpointing the location and time of gunfire incidents,
this support for the technology is not without its down side. Stakeholders who were interviewed
expressed concerns about an overreliance on GDT as a result of the officers’ belief in the
technology. For example, stakeholders in one site shared that officers had begun requesting GDT
reports for shootings that happened outside of the official coverage area but were still picked up
by acoustic sensors. Such reports are less likely to be admissible in court and are not typically
provided by the vendor in a detailed forensic report. In addition, command staff observed that in
emphasizing the importance of shell casing recovery, some officers had abandoned basic
investigative techniques such as neighborhood canvassing for witnesses.
Research Question 5: How are community members informed of the technology and
what are their views of its purpose and effectiveness?
Community outreach on the part of law enforcement was a component of all three agency’s GDT
deployment strategies. Securing support from community members on the departments’ use of
GDT—particularly from those residing in areas where the technology would be deployed—was
viewed by law enforcement stakeholders as a daunting task. Officers reported experiencing
apprehension or uneasiness when conducting community outreach for GDT implementation. One
stakeholder in Milwaukee conjectured that residents may have become desensitized to the violence
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
15
and another felt that residents lacked confidence in the police to do anything meaningful to address
it. Consistent with these views, focus group participants representing high-crime communities in
each jurisdiction reported low levels of trust and confidence in the police. They viewed police as
slow to respond to calls for service and observed that residents do not cooperate with the police
out of fear of retribution from gang members. One focus group participant reported that their
observations of police behavior were largely negative, citing unjustified police use of force against
residents.
Despite these negative views of police on the part of residents, law enforcement stakeholders
expressed that they were pleasantly surprised by the community’s willingness to assist the agency
in installing GDT sensors upon their outreach to them. Officers attributed this degree of
cooperation to their agency’s proactive outreach to community engagement, which they believed
demonstrated transparency and a spirit of open communication. Indeed, each department held
community meetings to explain the technology prior to its implementation, and they also described
the technology’s presence and use during community meetings following deployment. Part of this
outreach included contacting some individual homeowners to seek their permission to install
acoustic sensors on their rooftops. This typically included a signed agreement between the
homeowner and the LEA that the owner would not publicize the presence of the sensor and would
pay for the electrical costs associated with operating it. Residents were almost always compliant
with these requests.
Counter to the transparency associated with efforts to secure community buy in, law enforcement
deliberately kept the precise locations of sensors confidential from the community at large (absent
those known to homeowners with sensors installed on their roofs). Law enforcement stakeholders
reasoned that if coverage areas were not known to the public generally, the deterrence impact of
the technology would be enhanced, creating a diffusion of crime control benefits extending beyond
the geographic reach of the sensors. Focus group participants bought into this narrative, expressing
the belief that the technology applied to the entire city. In fact, we found few notable differences
in knowledge or support of GDT between focus groups of residents within or outside of coverage
areas, with the exception of Richmond, for which non-coverage area focus group participants were
less concerned about violence in their neighborhood, citing quality of life and code enforcement
issues instead.4 Overall, residents had a limited to basic understanding of how GDT works and
how the police use the information collected by sensors.
In terms of police-community relations with regard to actual gunfire incidents detected by GDT,
officers in Denver reported departmental policy of placing door hangers on the doors of homes
within the immediate vicinity of the location of the GDT alert. The door hanger advises that gunfire
had been detected and invites residents to contact the police department with any information they
may have. The degree to which this was effective was unknown by officers we interviewed.
However, at least one focus group participant reported they no longer felt the need to call the police
when they hear gunfire because they know the technology will detect it.
Discussion
This implementation evaluation has highlighted several key findings that have direct implications
for law enforcement policy and practice. Overall, the agencies profiled in this study view GDT as
a useful technology that aids in investigations. Analyses of case file reviews confirm that view,
suggesting that GDT enhances investigative efforts through increased recovery of bullet shell
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
16
casings, particularly those associated with homicides and robberies, which can be linked to other
crimes and weapons.
While training of officers on GDT use is uneven and has diminished over time, agency policies to
hold officers accountable for using the technology are robust and officer compliance is high, as
evidenced by an increase in the volume of neighborhood canvasses conducted following GDT
implementation. GDT data is also routinely integrated into crime analysis activities, guiding
deployment of officers and enhancing crime trend analyses that support both strategic and tactical
activities. Moreover, GDT data is viewed as a useful tool to employ in concert with ballistics and
firearms tracing databases.
Despite the documented benefits of GDT, it is not without its challenges. Stakeholders reported
uneven training and some degree of skepticism in the accuracy of the technology, which may
dissuade officers from using it to its full potential, although agency accountability mechanisms for
compliance with GDT response protocols appear to be strong. GDT can also be difficult to
integrate into existing, often antiquated CAD systems, and generates massive volumes of data,
creating burdens for staff in managing intel and using the information for more strategic and
proactive purposes.
Regarding police-community relations, GDT appears to be an uneven success. On the one hand,
agencies took care to communicate with community members their plans to implement GDT.
However, the locations of GDT installations—largely in high crime communities of color—
represent areas with historically fraught relations with the police. Community members expressed
low levels of trust and confidence in the police, but those negative views did not appear to influence
opinions about GDT, which were largely favorable.
GDT can also result in some unintended consequences. Law enforcement stakeholders expressed
concerns that officers were less likely to engage in traditional investigative techniques, such as
conducting field interviews and identifying witnesses, in favor of recovering shell casings. In
addition, multiple community members felt that dialing 911 when shots are fired is no longer a
necessity now that the city has GDT. This view is perpetuated by residents’ belief that GDT covers
the entire city—a belief that is intentionally perpetuated by law enforcement to yield a greater
deterrence benefit from the technology.
The findings from this evaluation point to five concrete recommendations for best practice in GDT
deployment and use.
(1) Engage in thorough and ongoing officer training. Provide training on the purpose,
value, and accuracy of GDT and on departmental policies and procedures regarding its
use, both in the police academy and through ongoing in-service training. Effective and
consistent training supports officer buy-in and increases the likelihood that officers will
comply with policies on responding to GDT alerts. Moreover, education on how the
accuracy and use of GDT has continued to improve can reduce skepticism about the
technology’s value, further ensuring that officers comply with GDT alert protocols.
(2) Develop strong accountability mechanisms. Training should be reinforced through
accountability measures to bolster officer compliance with response protocols, such as
exiting patrol cars to canvass for shell casings, conducting field interviews, and
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
17
identifying witnesses. Requiring officers to submit reports following each GDT alert
response, and requiring supervisors to conduct field checks and canvassing follow ups,
are two promising accountability measures that could be employed.
(3) Facilitate data management and integration. Consider existing data systems’ (i.e.
CAD, RMS, etc.) capacity to integrate GDT data, and update or modify systems
accordingly, ideally in advance of GDT deployment. The ability to integrate GDT data
into existing data systems has promising implications for efficiency, investigative
utility, and impact of GDT.
(4) Improve access to complementary technologies. The investigative utility and
potential impact of GDT increases considerably when the data it generates can be used
in conjunction with other policing technologies. Specifically, a robust and efficient
NIBIN system that can produce rapid, accurate results that allow investigators to link
cases within and between jurisdictions.
(5) Engage community members early and often. While GDT systems provide law
enforcement more accurate data on firearm violence, community partnership is key for
enhancing investigations and reducing gun violence. Agency outreach to community
members on planned GDT installation and ongoing GDT use should clearly convey that
GDT is not a substitute for resident engagement with and outreach to police upon
knowledge of gun violence and other public safety concerns. Agencies should carefully
consider whether the benefits of allowing people to believe that GDT covers the entire
jurisdiction outweigh the potential unintended consequence that community members
will no longer make calls for service based on the assumption officers already know
that shooting offenses have occurred.
The success of GDT in achieving its intended gun violence reduction impacts may ultimately lie
more in how it is implemented and used by LEAs than in its ability to accurately identify and
locate firearm shootings. The accuracy of GDT is relatively high and has been documented in other
studies; this manuscript sheds new light on how officers are trained and held accountable for GDT
use, how the data GDT creates are integrated in existing systems and with other investigative
methods, and how LEAs involve the community both prior to GDT implementation and
throughout its use. We find that the policies and practices surrounding GDT implementation and
use are critical in supporting the most effect use of the technology and the data it produces.
Indeed, a GDT system is most useful to LEAs that fully incorporate the technology into standard
policing operations. One means to integrating GDT fully into law enforcement operations is
assigning a fulltime supervisor to manage the GDT program. Identifying such a staff person would
facilitate close communication with the GDT vendor for system upkeep; support the management,
analysis, and reporting of gunfire alert data; ensure a single point of contact for investigators in
their requests for detailed forensic reports from the vendor; and hold officers accountable by
monitoring compliance with canvassing procedures and recanvassing in areas where minimal
evidence was located from GDT alerts. With full integration and dedicated staff in support of GDT,
the technology holds great promise as an additional source of data for police officers and
agencies—one that can enhance shooting investigations when used in conjunction with existing
tools, data, and procedures. In fact, if GDT is implemented with fidelity and with full and ongoing
community engagement, it may even enhance community trust in law enforcement’s response to
firearm violence.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
18
Endnotes
1. The coverage area expansions in Denver primarily occurred after the research team collected
data from the agency. As a result, data associated with the expansions were not included in
our analyses.
2. While we aimed to get 180 cases total with 60 in each site, not all of the randomly selected
cases involved a firearm. We requested replacement files when this occurred; however, this
was not always possible. Denver could not identify robbery cases that may have involved a
firearm at the time of our site visit, so a decision was made to exclude these cases from the
sample. Weapon violations included cases where an individual recklessly fired their gun,
illegally carried a weapon, was a convicted of a felony and carrying a firearm, among other
types of related events.
3. The 10.6% of cases that mentioned GDT prior to its installation come from Milwaukee, where
GDT was installed and in use prior to its official deployment date that was used in these
analyses.
4. This finding is unsurprising, as Richmond’s GDT coverage area is very large relative to the
city, leaving little opportunity to identify a comparison area that represented a good match
in terms of demographics and historic violent crime rates.
Acknowledgments
This project was supported by Award No. 2015-R2-CX-K147, awarded by the National Institute
of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and
conclusions or recommendations expressed in this publication/program/exhibition are those of the
author(s) and do not necessarily reflect those of the Department of Justice. We would like to thank
staff from the Milwaukee, Denver, and Richmond Police Departments who played a significant
role in speaking with us about their use of gunshot detection technology. We also thank colleagues
Dave McClure for reviewing an earlier version of this article and Carla Vasquez-Noriega and Dean
Obermark who work on the project as analysts.
References
Aguilar, J. (2015). Gunshot detection systems in civilian law enforcement. Journal of the Audio
Engineering Society, 63(4), 280-291. doi:10.17743/jaes.2015.00020
Carr, J. B., & Doleac, J. L. (2016). The geography, incidence, and underreporting of gun
violence: New evidence using ShotSpotter data. Washington, D.C.: Brookings Institution.
Chacon-Rodriguez, A., Julian, P., Castro, L., Alvarado, P., & Hernandez, N. (2011). Evaluation
of gunshot detection algorithms. Journal of IEEE Transactions on Circuits and Systems
58(2), 363-373.
Choi, K. S., Librett, M., & Collins, T. J. (2014). An empirical evaluation: Gunshot detection
system and its effectiveness on police practices. Police Practice and Research, 15, 48–
61.
Fereday, J., & Muir-Cochrane, E. (2006). Demonstrating rigor using thematic analysis: A hybrid
approach of inductive and deductive coding and theme development. International
Journal of Qualitative Methods, 5(1): 80-92.
Accepted preprint to: Lawrence, D.S., La Vigne, N.G., Goff, M., & Thompson, P.S. (2019). Lessons Learned Implementing
Gunshot Detection Technology: Results of a Process Evaluation in Three Major Cities. Justice Evaluation Journal 1(2), 109–129.
19
Goode, E. (2012). Shots Fired, Pinpointed and Argued Over. New York Times. Retrieved from
https://www.nytimes.com/2012/05/29/us/shots-heard-pinpointed-and-argued-over.html
Irvin-Erickson, Y., La Vigne, N. G., Levine, N., Tiry, E., & Bieler, S. (2017). What does gunshot
detection technology tell us about gun violence? Applied Geography, 86, 262-273.
doi:10.1016/j.apgeog.2017.06.013
Koper, C. S., Lum, C., Willis, J. J., Woods, D. J., & Hibdon, J. (2015). Realizing the potential of
technology in policing: A Multisite study of the social, organizational, and behavioral
aspects of implementing policing technologies. Washington, DC: National Institute of
Justice, U.S. Department of Justice. Retrieved from http://cebcp.org/wp-
content/technology/ImpactTechnologyFinalReport.pdf
La Vigne, N. G., Lowry, S. S., Markman, J. A., & Dwyer, A. M. (2011). Evaluating the use of
public surveillance cameras for crime control and prevention. Washington, DC: Urban
Institute and COPS Office.
Litch, M., & Orrison, G. A. (2011). Draft technical report for SECURES demonstration in
Hampton and Newport News, Virginia. Washington, DC: National Institute of Justice,
U.S. Department of Justice. Retrieved from
https://www.ncjrs.gov/pdffiles1/nij/grants/233342.pdf
Mazerolle, L. G., Frank, J., Rogan, D., & Watkins, C. (2000). Field evaluation of the ShotSpotter
gunshot location system: Final report on the Redwood City field trial. Washington, DC:
U.S. Department of Justice.
Mazerolle, L. G., Watkins, C., Rogan, D., & Frank, J. (1998). Using gunshot detection systems
in police departments: The impact on police response times and officer workloads. Police
Quarterly, 1, 21-49.
Ratcliffe, J. H., Lattanzio, M., Kikuchi, G., & Thomas, K. (2018). A partially randomized field
experiment on the effect of an acoustic gunshot detection system on police incident
reports. Journal of Experimental Criminology, DOI: doi.org/10.1007/s11292-018-9339-1
Selby, N., Henderson, D., & Tayyabkhan, T. (2011). ShotSpotter gunshot location system
efficacy study. Fayetteville, NC: CSG Analysis. Retrieved from http://njdc.info/wp-
content/uploads/2017/10/Shot-Spotter-Gunshot-Location-System-Efficacy-Study.pdf
ShotSpotter. (2018). ShotSpotter FAQ. Newark, CA: ShotSpotter, Inc. Retrieved from
https://www.shotspotter.com/wp-content/uploads/2018/08/FAQ_Aug_2018.pdf
Strom, K. J., Hendrix, J., Taniguchi, T., Aagaard, B., Werth, R. S., & Legacy, S. (2016).
Research on the impact of technology on policing strategy in the 21st century: Final
report. Washington, DC: Office of Justice Programs. Retrieved from
https://www.ncjrs.gov/pdffiles1/nij/sl001001.pdf
U.S. Census Bureau. (2018). 2012-2016 American Community Survey 5-year estimates.
Retrieved from https://factfinder.census.gov/
Watkins, C., Mazerolle, L. G., Rogan, D., & Frank, J. (2002). Technological approaches to
controlling random gunfire. Policing: An International Journal of Police Strategies &
Management, 25(2), 345-370. doi:10.1108/13639510210429400