Content uploaded by Dennis Mares
Author content
All content in this area was uploaded by Dennis Mares on Aug 03, 2023
Content may be subject to copyright.
1
Improving the Police Response to Gunfire
A COST-BENEFIT ANALYSIS
OF SHOTSPOTTER IN
WINSTON-SALEM, NC
CCSVP
Center for Crime Science and Violence Prevention
www.siue.edu/ccsvp
2
INDEX
Executive Summary. 3
Introduction. 4
ShotSpotter in Winston-Salem, NC. 6
Responding to Gunfire Alerts. 8
Results of ShotSpotter Responses. 15
Crime Reductions. 18
Cost Impacts. 23
Conclusions and Recommendations. 25
Sources. 26
www.siue.edu/ccsvp
3
EXECUTIVE SUMMARY
ShotSpotter’s gunshot detection system was deployed in Winston-Salem in
August 2021. Since then, nearly 2,000 alerts received a response by Winston-
Salem Police.
Results indicate:
Improved response to gunfire
♦ The response to alerts is significantly quicker than those called in
by residents (- 5 min.).
♦ ShotSpotter calls received significantly more investigative time,
which likely indicates improved evidence recovery.
♦ Fewer than one in four ShotSpotter alerts also received a call
from residents.
ShotSpotter produces the following actionable results:
♦ Shell casings were recovered in 581 incidents (37.1%)
♦ Firearms were recovered in 47 (3%) of alerts.
♦ Sixty-seven (3.4%) gun-related arrests are connected to alerts.
Deployment of ShotSpotter is related to a reduction in violent gun crimes:
♦ Aggravated assaults are down 26% comparing before-after
results in the ShotSpotter area.
♦ Comparable area and overall city numbers indicate an increase in
aggravated assaults during the same period. Comparatively
assaults are down 38% in the ShotSpotter community.
♦ In real numbers, there are between 51-75 fewer assaults annually
in the ShotSpotter area than would be expected.
Cost-Benefits:
♦ Our estimate suggests that ShotSpotter may save the Winston-
Salem community between $5 and $8 Million annually.
♦ Average annual implementation cost is estimated between
$230,000-350,000
♦ This indicates a $15-25 return for each dollar spent.
www.siue.edu/ccsvp
4
INTRODUCTION
ShotSpotter is an Acoustic Gunshot Detection System (AGDS) which uses
multiple sensors to detect the location of gunfire. Upon positive identification of
the sound patterns of gunfire, acoustic events are reviewed by ShotSpotter
personnel for accuracy. Once a final determination is made an alert is forwarded
to the police agency. Alerts are typically forwarded to dispatchers, also officers
may receive notifications directly on their Mobile Data Terminals (MDTs) or
mobile phones. Alerts include precise location data, the number of rounds fired
and an accurate time stamp of the incident. Audio of the incidents can be
streamed as well.
Figure 1. Gunshot detection
Source: https://www.shotspotter.com/law-enforcement/gunshot-detection-technology/
ShotSpotter systems have been deployed in well over 130 cities in the US and
the company is currently the leading vendor for gunshot detection systems.
While gunshot detection has been criticized for being inaccurate and leading
police to many so-called ‘false positives’- reported gunfire that turned out to be
other loud sounds- this is not supported by recent research (Mares, 2022). In
fact, the accuracy of the system appears quite precise (Watkins et al., 2002).
www.siue.edu/ccsvp
5
ShotSpotter also utilizes human reviewers to further limit false positives.
Gunshot detection has typically been viewed positively by community residents
(Haberman et al., 2020; Vovak et al., 2021). Moreover, 2/3 of the general public
supports the use of the technology by police1. There are, however, some
limitations to these systems. For example, they are unlikely to detect indoor
gunfire but also have difficulty picking up gunshots fired from a vehicle.
Even though gunshot detection systems are primarily deployed to reduce
gun violence, academic research has found mixed results with respect to crime
reductions (
cf.
Lawrence et al., 2019; Mares & Blackburn, 2021; Mares, 2023).
Implementation differences, however, may explain the degree of success and
agency experiences. There is little doubt, however, that gunshot detection
improves the speed and precision of the response (Piza et al., 2023).
Below we will quantify some of the results ShotSpotter has brought to the
City of Winston-Salem and how it has impacted police practices and outcomes.
A cost-benefit discussion will also be provided.
1 https://morningconsult.com/2021/09/13/data-privacy-security-gunshot-detection-technology-poll/
www.siue.edu/ccsvp
6
SHOTSPOTTER IN WINSTON-SALEM, NC
In August 2021 ShotSpotter provided Winston-Salem with a 3 square mile area
of coverage for its gunfire detection system. The area covered is located just
North-East of Downtown, South-West of Smith Reynolds Airport and intersected
by Highway 522.
Figure 2. Winston-Salem Police Beats.
Between late August 2021 and December 31st, 2022, Winston-Salem Police
Department (WSPD) responded to nearly 2,000 ShotSpotter alerts. In this report
we detail the initial results of the technology and the WSPD response to these
alerts. We will conclude with a cost-benefit assessment of the technology and
implications for current policies and practices.
2 Please note that the exact coverage area of ShotSpotter is law enforcement sensitive information and therefore
not shown.
www.siue.edu/ccsvp
7
RESPONDING TO GUNFIRE ALERTS
Winston-Salem PD officers responded to a large number of ShotSpotter alerts in
the area that received coverage, but simply responding, of course, does not
mean that this is an effective or efficient use of officer time (Blackburn & Mares,
2019). Below we detail some important metrics that provide better context for
the data.
Figure 3. Violent Crime Hot Spots in Winston Salem (red is high violence, blue is low
violence).
To make comparisons more meaningful we created a comparison area that
shares some key characteristics with the area covered by ShotSpotter.
Although no area in Winston-Salem has quite the density of gun-related crimes
as the ShotSpotter area we did find an area that has a relatively higher density
of gun-related crimes, including several pockets of multi-family housing with
even extremer densities (see figure 3). This comparison area is south of
Downtown and east of Highway 52, running primarily along Interstate 40. The
comparison area is a bit larger than the ShotSpotter area, 4.5 versus 3 square
miles, yet both have fairly equitable numbers of calls for service for gunfire and
www.siue.edu/ccsvp
8
a reasonably similar number of violent crimes (see table 1 below). That does not
mean the two areas are completely comparable but given that implementation of
ShotSpotter occurred in the area of highest need, this is the closest comparison
we are able to find and better as a comparison against overall city data alone.
Both areas share a large proportion of violent and gun offenses in Winston-
Salem. Combined they account for 30% of gunfire related calls for service
(excluding ShotSpotter alerts) and 33% of violent crimes (excluding sexual
assaults) within the city limits.
INCIDENT
TYPE
SHOTSPOTTER
AREA
COMPARISON
AREA
REST OF
CITY
DFSA
883
1,012
5,015
DFSAD
238
244
702
SHOOTING*
156
95
354
TOTAL CFS
1,277
1,351
6,071
MURDER1
18
13
45
ROBBERY
76
46
364
ASSAULT
508
378
1709
TOTAL
CRIME
602
437
2,119
Table 1. Comparison of Calls for Service / Crimes 2020-2022
*Shooting includes call codes for: drive by shooting, person shot and GSW.
1. Crime data ends on Dec 13th 2022, and thus does not include the full year.
Figures 4 and 5 below show some interesting patterns when ShotSpotter was
activated. In the ShotSpotter area, for example, we see a large drop in
Discharging of a Firearm calls for service (DFSA). While those numbers were
already trending down somewhat, they really took a dive a few months into
implementation and remained extraordinarily low from there. By contrast (figure
6), the comparison area shows mostly steady patterns with a bit of an increase
in gun discharges that caused property damage (DFSAD). A rapid decline in
what are essentially shots fired calls by the public was also seen in St. Louis
www.siue.edu/ccsvp
9
(Mares & Blackburn, 2021) and should not simply be taken as evidence that gun
crimes are down. Rather, it may suggest that calls by residents are replaced by
ShotSpotter alerts, which now outweigh prior levels of calls by residents.
Figure 4. ShotSpotter Area. Monthly trends in gunfire related calls for service.
www.siue.edu/ccsvp
10
Figure 5. Comparison Area. Monthly trends in gunfire related calls for service.
To explore how ShotSpotter impacts the response to gunfire we first examine
response-time information, comparing discharging events reported by
community members to ShotSpotter alerts. In table 2 below we can see that
ShotSpotter alerts get dispatched more than 5 minutes faster than calls by
residents, which is statistically significant. While travel time to the scene of a
gunfire incident takes slightly longer for ShotSpotter alerts, this difference is only
13 seconds. This means that from the time a gun is fired to the moment police
arrive, ShotSpotter responses receive a 5-minute advantage.
Interestingly also is that ShotSpotter investigations appear to take much
longer (7+ minutes) than calls from community members. This makes some
sense as knowing the precise location of gunfire increases the chances of
finding evidence.
www.siue.edu/ccsvp
11
SHOTS FIRED
(DFSA)
SHOTSPOTTER
ALERTS
DISPATCH TIME
8.84
3.62***
TRAVEL TIME
6.75
6.87*
INVESTIGATIVE TIME
41.06
48.29**
TOTAL TIME
54.79
55.65
CASES
168
1,395
Table 2. Calls for service times in minutes in ShotSpotter area.
Statistical significance is based on Mann Whitney U tests: * p<.05, ** p<.01, *** p<.001
What the prior table does not tackle is whether the implementation of
ShotSpotter has implications for how gunfire is responded to. To do so we
compare response times before and after ShotSpotter implementation for
several areas in Winston-Salem. Here we focus on gunfire-related calls
(ShotSpotter and DFSA) for service. Table 3 below shows that the area covered
by ShotSpotter has a significantly lower dispatch time, but significantly longer
travel, investigative and consequently total time. Interestingly, dispatch times
lengthened significantly in both the comparison and remainder of Winston-
Salem, which makes the time-saving even more impressive. Where prior to
ShotSpotter, dispatch times were within a minute of each other across the three
different areas, after implementation, the ShotSpotter area dispatched calls for
service in nearly half the time, or 4 minutes faster.
www.siue.edu/ccsvp
12
BEFORE
AFTER
SHOTSPOTTER AREA
DISPATCH
5.77
4.19***
TRAVEL
5.53
6.85***
INVESTIGATIVE
42.78
47.51***
TOTAL
53.03
55.56***
CASES
715
1563
COMPARISON AREA
DISPATCH
6.77
8.277**
TRAVEL
7.73
7.64
INVESTIGATIVE
35.05
34.54
TOTAL
48.01
54.92
CASES
560
451
REST OF CITY
DISPATCH
6.07
8.92***
TRAVEL
8.59
9.22***
INVESTIGATIVE
33.20
35.40***
TOTAL
46.69
53.95***
CASES
2845
2349
Table 3. Comparison of calls for service times before/after ShotSpotter implementation.
Statistical significance is based on Mann Whitney U tests: * p<.05, ** p<.01, *** p<.001
Interestingly both the ShotSpotter area and the rest of the city experienced an
uptick in the time it took officers to get to the scene of an incident. While it is
easy to read too much in these numbers, we suspect the increased travel times
may simply be an outcome of declining officer numbers. As police agencies face
growing attrition this is not terribly surprising. Investigative times changed
significantly for both the ShotSpotter area and the rest of Winston-Salem with
the former increasing proportionally the most. We believe this is likely an
outcome of the more serious nature of incidents in the ShotSpotter area which
requires more resources, but also may be connected to the fact that the
proportion of ShotSpotter calls to DFSA calls in that community is severely out
of balance.
www.siue.edu/ccsvp
13
In sum, results of response times indicate that implementation of
ShotSpotter fundamentally altered the times it takes police to get to gunfire and
how long police investigate the crime scene. These results stand in contrast to
the comparison area, which saw the exact opposite. It is fair to read such results
as a positive development and improving the WSPDs ability to better serve the
community. While the geographic accuracy and reporting speed of ShotSpotter
is probably a large benefit to police, crucial information about shootings is often
relayed by community members. Keeping residents connected to police and
reporting gunfire is therefore extremely important.
www.siue.edu/ccsvp
14
RESULTS OF SHOTSPOTTER RESPONSES
Because a new ShotSpotter Alert can be generated after a few second pause in
gunfire it is important to clean ShotSpotter data prior to further use. We
therefore remove alerts that occurred within 30 minutes and 500ft of one
another to create a better representation of the number of unique gunfire
incidents (Huebner et al. 2021). Out of the 1,965 alerts recorded we found that
they represented 1,595 distinct incidents. A few alerts lack additional data,
mostly recent cases for which all investigative details may not have been
available, and are also excluded from analysis. According to data tracked by
WSPD, out of 1,567 unique shooting incidents with complete data, firearms
were recovered in 47 (3%) incidents, and casings were recovered in 581
incidents (37.1%) with a total of 1,577 casings. We should point out as well that
the number of casings recovered by responding to ShotSpotter alerts (3 sqm)
nearly matches the 2,101 casings found by responding to calls for service from
the community in the entire city (134 sqm)
Like other cities, few of gunfire incidents in Winston-Salem also generated
calls for service from community members (Huebner et al.,2021; Mares 2022).
In only 287 (18.3%) cases was a ShotSpotter incident accompanied by a
community call. Crime-wise, ShotSpotter alerts were connected to 41(2.6%)
aggravated assaults, 3 robberies and 7 homicides. Seventeen initial (1%)
arrests were enacted, which is higher than reported in other locations (Mares &
Blackburn, 2021), but this only counts arrests made during the initial response.
A combined total of 67 (3.4%) gun-related arrests were made in connection to
responding
and
investigating ShotSpotter alerts, suggesting that benefits of the
system may extend beyond the initial response.
Delving a bit more into the data we also look at which factors may explain
why some ShotSpotter incidents may be more likely reported by residents. To
do so a statistical model was developed that measures the likelihood of a
ShotSpotter alert also receiving a report by a community member (yes/no). We
explored several possible explanative factors including: (1) the total number of
rounds fired during the incident (2) the number of ShotSpotter alerts per
incidents, (3) whether the alert led to the scene of a violent crime and (4)
temporal variation (month, day of the week and hour of the day) as there is
www.siue.edu/ccsvp
15
quite a bit of variability in when alerts occur (see figure 6 below, for example).
Figure 6. Time chart of ShotSpotter alerts.
Findings of the statistical analysis indicate that the only statistically important
factors impacting reporting by residents is the number of rounds fired during an
incident
and
whether the incident was connected to a violent crime (assault,
homicide or robbery). Each additional round fired increases the likelihood of
receiving a community call by around 13%. This is not surprising; more rounds
mean more clarity on whether a loud sound is actually gunfire and also increase
the chances that more people heard the gunshot. Violent crimes are also far
more likely to receive calls from the community, probably because victims or
bystanders notify 911. Our analysis suggests that violent crimes are 14 times
more likely to be called in by community members than gunfire incidents without
a victim.
In more tangible terms, 78.4% of ShotSpotter alerts in which a serious
violent crime occurred receive a community call, whereas only 16.3% of gunfire
incidents without a victim generate such a call. This also means that nearly ¼ of
crimes did not receive a call from the public and that ShotSpotter alerts provided
a faster response for victims. Some may not have been uncovered altogether, or
failed to yield a crime scene. A good number of assault victims often show up at
hospitals and refuse to cooperate with police, which can leave the crime scene
unknown and hampering investigations (Mares, 2022).
www.siue.edu/ccsvp
16
CRIME REDUCTIONS
One of the key reasons police departments purchase gunshot detection is to
tackle gun-related violent crime (Mares, 2023). Winston-Salem only has had
ShotSpotter for a little over a year, which makes extremely thorough approaches
to measure the impact not quite feasible yet. We can, however, provide some
initial insights into the efficacy of ShotSpotter by comparing before and after
levels of violent crimes.
Below (figures 7-9) we compare multiple crime incidents in the ShotSpotter
area, the comparison area and remainder of the city. First let’s examine total
DFSA, or illegal discharging of a firearm3. In the ShotSpotter area a substantial
uptick in such incidents occurred once ShotSpotter became active (represented
by the black vertical line). Reported incidents increase from around 30 to about
70 per month. This is not surprising as ShotSpotter uncovers more such events.
In the comparison area, the number of such incidents appears mostly stable,
with no discernable trend. The remainder of the city also looks fairly stable with
perhaps a bit of an overall upward trend, but clearly nothing comparable to the
substantial break seen in the ShotSpotter area.
Figure 7. Discharging incidents -ShotSpotter Area.
3 In WSPD RMS system this includes ShotSpotter alerts that are sustained as such, but also other calls for
service in which a firearm was found to be discharged but in which no person was injured.
www.siue.edu/ccsvp
17
Figure 8. Discharging incidents - Comparison Area.
Figure 9. Discharging incidents -Remainder of City.
Turning to more serious violent crime committed with a firearm: homicides,
robberies and aggravated assaults (see figures 10-12 below). For the
ShotSpotter area we observe lower levels of aggravated assaults after
ShotSpotter implementation but see no changes in trends in homicides and
robberies, which is most likely an outcome of the relative rarity of these events.
By contrast the comparison area and the remainder of Winston-Salem display a
growth in assaults levels. What is more, because aggravated assaults are most
numerous, they dominate the results we see for the combination of all three
violent gun crime types.
www.siue.edu/ccsvp
18
Figure 10. Violent gun crimes -ShotSpotter Area
Figure 11. Violent gun crimes - Comparison Area
www.siue.edu/ccsvp
19
Figure 12. Violent gun crimes -Remainder of City
That said, trends lines are often subjectively interpreted, so it is important to
examine whether the trends we described can be verified in statistical analysis.
To this end we perform a statistical test (independent t-test) that compare
average crime levels before and after ShotSpotter implementation.
Crime Type
SST area
Comparison area
Rest of City
Robbery-Pre
2.32
1.42
9.58
Robbery-Post
1.94 (-16%)
1.19 (-17%)
11.19 (+17%)
Homicide-Pre
.47
.47
1.11
Homicide-Post
.56 (+19%)
.25 (-46%)
1.50 (+35%)
Assault-Pre
16.37
9.74
45.16
Assault-Post
12.13* (-26%)
11.75 (+21%)
51.94* (+15%)
Violent-Pre
19.16
11.63
55.84
Violent-Post
14.63* (-24%)
13.19 (+13%)
64.63* (+16%)
Discharge-Pre
29.21
22.11
112.84
Discharge-Post
71.19***(+143%)
22.00 (0%)
119.56 (+6%)
Table 4. Comparison of before/after crime levels.
Statistical significance is based on independent t-test: * p<.05, ** p<.01, *** p<.001
www.siue.edu/ccsvp
20
The results indicate that the ShotSpotter area experienced substantial changes
in crime incidents and those changes are consistent with those seen in the
earlier trend graphs We should caution to readers that both robbery and
homicide are relatively rare for the ShotSpotter and comparison areas and that
proportionally large swings in those number are expected, we therefore
encourage interpretation primarily focus on the more numerous categories of
aggravated assault and discharging of a firearm.
For aggravated assaults the ShotSpotter area displays a statistically
significant reduction when comparing before/after ShotSpotter implementation,
with an average reduction of around 26%. By contrast the comparison area
exhibits a non-significant increase of 21% and the remainder of the city
experienced a significant 15% increase.
If we take the results from comparison area (+21%) as the likely change
the ShotSpotter area would have experienced without the technology the overall
comparative reduction for the ShotSpotter area equals 38%4, which is a
substantial number and comparable to results found in Cincinnati (Mares, 2023).
When examining violent crime (combining robbery, homicide and aggravated
assaults) a similar conclusion may be reached, which also shows the relatively
small numerical impact that robberies and homicides have in overall levels of
violence. The ShotSpotter area displays a 24% reduction in mean monthly
violence numbers, whereas the control area and the remainder of the city
experienced a 13% and 16% uptick respectively. Again, considering the
comparison area as a benchmark for expected changes in violence levels, the
overall likely contribution of ShotSpotter would indicate a comparative 32%
reduction.
For discharging of a firearm events the situation is quite different. Here the
ShotSpotter area shows a very large and statistically significant increase, more
than doubling prior levels. Both the comparison area (0%) and the remainder of
the city (+6%) show a fairly stable picture. While these data might be viewed by
some as alarming, they are in fact not; in fact, these results are in line with prior
work (Mares & Blackburn, 2021) and a simple outcome of more events being
reported and sustained through ShotSpotter. The increase is most likely not an
actual increase in people discharging firearms, but merely a reduction in
underreporting that is common for such crimes.
4 Assuming the ShotSpotter area would also have experienced a 21% increase, it would have experienced 19.8
assaults instead of the 12.1 reported, which is a 38% reduction from expected.
www.siue.edu/ccsvp
21
What is not included above is a small but potentially important impact the
improved speed of the police response may have. Several medical studies have
pointed out that a quicker response may -in some cases- prevent death or
serious permanent disability to victims of gun violence (Goldenberg et al., 2019;
Gontarz et al., 2021). Indeed, WSPD identified two cases in which victims
received faster medical care that according to medical professionals likely saved
their lives; in one case only a ShotSpotter alert led to the victim. While in such
cases ShotSpotter may not prevent crime, it reduces the seriousness of the
victimization and the severity of harm. Not only does this have positive effects
for victims, the cost-savings of preventing lives lost through improved access to
medical care is potentially substantial. Considering that the societal cost of a
homicides are easily in the millions, but assaults cost only a fraction of this could
have substantial repercussion for the overall cost impacts of gunshot detection
systems. The problem is that making this argument can only be robustly done
with a large sample of data, likely involving multiple cities and a substantial
period of observations.
We would be amiss not to point out to some limitations in the analysis. The
amount of data is limited, which prevents more robust types of analysis. Also,
the comparison area is not fully comparable to the ShotSpotter area, which
could under- or overstate the differences. What is more, 2020 data are likely still
very much impacted by COVID-19, which may explain some of the reductions in
2021, even though these were not seen city wide or the comparison site. In
short, some caution should be given to the conclusions, and they should be
regarded as indicative, but not definitive.
www.siue.edu/ccsvp
22
COST IMPACTS
Because aggravated assaults are the driving force of violent crime trends in both
the ShotSpotter and Comparison area and because this category in particular
show significant reductions in crime after implementation of ShotSpotter we can
use this to calculate the impact of ShotSpotter on the cost of crime. To do so we
can use two established methodologies to explore: (1) estimated additive cost
and (2) Willingness-To-Pay estimates. Additive cost approaches are essentially
a simple tallying of all the costs that are incurred by crime, such as medical cost,
lost wages and criminal justice related costs. Willingness-To-Pay (WTP)
approaches rely on economic research that suggests a simple additive cost
approach may be less accurate as it does not reflect the subjective value that
people assign to crime concerns. WTP estimates instead rely on asking a large
number of people how much they are willing to pay for a specific percentage
crime reduction.
For an estimated additive cost, we use data from Rand, which estimates
the average cost of an Aggravated Assault at $128,937.405. Because the cost of
living in Winston-Salem is lower than the national average, we adjust for this6
and derive a localized cost of $104,826.11 for each aggravated assault. Using
the pre-ShotSpotter average annual number of assaults in the ShotSpotter area
(196.44) and factoring in just the raw crime reductions achieved (-26%) this
represents a reduction of 51 aggravated assaults per year, or an annual cost-
savings of $5,353,931. If we incorporate the fact that crime went up in the
comparison area and use the comparative reduction of 38% this would suggest
75 fewer assaults per year, or a cost-saving of $7,861,958
Using WTP estimates from Cohen et al. (2004) we find that serious
assaults are estimated here to cost an inflation-adjusted average of
$124,063.39. Adjusted for the cost-of-living in Winston-Salem this brings us to
an estimated cost per aggravated assault of $100,863.54. Using again, both the
raw and comparative crime reductions calculated this means a cost savings
between $5,144,041 and $7,564,766.
Of course, ShotSpotter costs money and so does the response to the
additional gunfire incidents as well as the increasing demand for evidentiary
5 https://www.rand.org/well-being/justice-policy/centers/quality-policing/cost-of-crime.html with inflation adjusted
using https://data.bls.gov/cgi-bin/cpicalc.pl?cost1=87%2C238.00&year1=200701&year2=202301
6 https://www.bestplaces.net/cost_of_living/city/north_carolina/winston-salem
www.siue.edu/ccsvp
23
processing. The current ShotSpotter contract for Winston-Salem costs the city
about $205,000 per year. In addition, the WSPD will respond to additional calls
for service (ShotSpotter alerts). While this does not incur additional personnel
costs -they are already on duty and responses are primarily in otherwise low
volume hours (see figure 6), it will incur additional wear and tear on vehicles and
increase gas usage. Using the federal mileage rate of 58.5 cents/mile (2022) we
can estimate -very roughly- the cost of the immediate response. Assuming an
average speed of 45 miles per hour, or .75 miles per minute the average travel
distance to and from a ShotSpotter alert is 2x (.75 x 6.87) = 10.3 miles, or a cost
of $6.03. multiplying this by the number of alerts per year (1,500) this would only
add just over $9,000 for a single vehicle response. However, it is more likely that
two vehicles may respond (supervisor) and in some cases additional vehicles
may be needed. To be conservative we believe it is reasonable to put vehicle
related response cost at around $25,000.
What is less clear is what the cost of evidence processing may be.
ShotSpotter delivers a substantial increase in casing recoveries. From August
2021 through the end of 2022, records show ShotSpotter alerts led to recovery
of 1,577 shell casings, compared to 2,101 casings recovered in the entire city
through conventional means. Estimating the cost of processing these rounds is
not simple, as it involves handling by personnel to file and trace the casings as
well as the cost of the machines that perform the tracing. What is more, not all
casings recovered will need tracing, some are too deformed to reliable match,
some may be part of a large number of similar rounds. Just the same if we want
to be extremely conservative and price processing for each casing at $100, the
total cost is still fairly non-consequential for the overall cost-benefit picture.
Annualizing the casings recovered would mean about 1,200 casings per year at
a cost of $120,000.
In sum, while the cost-savings, based on crime prevention can be
estimated anywhere between $5,144,041 and $7,861,958, the costs of
increased surveillance and enforcement are only in the range between
$230,000 and $350,000 per year. It should be pointed out, however, that while
the savings are shared by all in Winston-Salem, the costs are entirely on the
WSPD side.
www.siue.edu/ccsvp
24
CONCLUSIONS AND RECOMMENDATIONS
Results of our analyses show positive results from WSPDs implementation of
ShotSpotter’s acoustic gunshot detection system. The data indicate a
substantially faster response time (~5 minutes) in the ShotSpotter area and a
reduction in serious violent crimes committed with a firearm (26-38% reduction
in aggravated assaults with firearm). Anecdotally the system also improves
health outcomes for victims of gunfire, but this is more difficult to establish with
current data. Finally, the system appears to improve investigative outcomes by
returning substantially more shell casings and increasing arrests for gun-related
offenses. These results are most certainly encouraging and indicative of a
sound implementation by WSPD.
As our results indicate, the benefits of ShotSpotter certainly appear to
outweigh the cost of the city’s investment in the technology with a net annual
gain to society of about $5-8,000,000 and a cost between $230,000 and
$350,000. Even at the lower end of our estimate the implementation of the
system indicates a 15 dollar return for each dollar spent. That by all accounts
appears to be a solid investment in limited resources. We believe it is therefore
reasonable to support continued investment in the technology by WSPD.
We should acknowledge, however, that the current data are somewhat
limited for a more comprehensive evaluation, and we encourage WSPD to
continue to track its successes with the system and reevaluate the cost-benefits
on an ongoing basis. That said, results are in line with other agencies that have
shown adherence to best practices, such as Cincinnati (Mares, 2023).
In sum, ShotSpotter deployment in Winston-Salem shows strong potential
and initial success based on the current data. We encourage a more extensive
evaluation around the 3-year mark when more data has accrued, and a more
detailed analysis is feasible. We also commend WSPD for its tracking of the
data and the results connected to their response to ShotSpotter investigations.
Close tracking of these results is important because crime reductions by
themselves may not only be rooted in deterrence, but also in investigative work.
Here we would encourage the department to track how ShotSpotter generated
evidence (casings and alert data) assist in prosecution of gun offenders. This
will likely become more important over time but can provide insights beyond
those examined here.
www.siue.edu/ccsvp
25
SOURCES
Blackburn, E., & Mares, D. (2019).
The Hidden Costs of Police Technology:
Evaluating Acoustic Gunshot Detection Systems.
Police Chief Magazine.
https://www.policechiefmagazine.org/the-hidden-costs-of-police-technology/.
Cohen, M.A., Rust, R.T., Steen, S & Tidd, S.T. (2004). Willingness-To-Pay for
crime control programs.
Criminology
, 42: 89-110. https://doi.org/10.1111/j.1745-
9125.2004.tb00514.x
Goldenberg, A., Rattigan, D., Dalton, M., Gaughan, J.P., Thomson, J.S.,
Remick, K., Butts,C. & Hazelton, J.P. (2019). Use of ShotSpotter detection
technology decreases prehospital time for patients sustaining gunshot wounds.”
Journal of Trauma and Acute Care Surgery
87(6): 1253–1259.
Gontarz B.R., Siddiqui, U.T., Campbell, B., Gates, J., O'Hare J.M., Green, C.,
McQuay J. & Shapiro D.S. (2021). Firearm Acoustic Detection in Hartford,
Connecticut: Outcomes of a Trauma Center - Law Enforcement Collaboration.
Cureus.
13(10):e18789. doi: 10.7759/cureus.18789.
Haberman, C., Ruhland, E., Frank, J., Kelsay, J. & Desmond, J. (2020).
Cincinnati Police Department: Price Hill ShotSpotter Survey Evaluation Report.
Cincinnati, OH: Institute of Crime Science.
Huebner, B. M., Lentz, T.S. & Schafer, J.A. (2020). “Heard Shots – Call the
Police? An Examination of Citizen Responses to Gunfire.”
Justice Quarterly.
DOI: 10.1080/07418825.2020.1799063.
Koren, D. (2018).
ShotSpotter Pilot Assessment
. Las Vegas, NV: Las Vegas
Police Metropolitan Police Department: Las Vegas, NV.
Lawrence, D., N. La Vigne, and P. Thompson. (2019).
Evaluation of Gunshot
Detection Technology to Aid in the Reduction of Firearms Violence
. Urban
Institute: Washington D.C..
Mares, D., and E. Blackburn. (2021). “Acoustic Gunshot Detection Systems: A
Quasi-Experimental Evaluation in St. Louis, MO.”
Journal of Experimental
Criminology
17: 193–215.
www.siue.edu/ccsvp
26
Mares, D (2022).
Gunshot Detection: Reducing Gunfire through Acoustic
Technology. Problem-Oriented Guides for Police Response Guide Series No.
14.
Bureau of Justice Assistance: Washington D.C..
Mares, D. (2023). “Evaluating an Acoustic Gunshot Detection System in
Cincinnati.” In E. Groff and C. Haberman, eds. Forthcoming.
The Study of Crime
and Place: A Methods Handbook
. Philadelphia, PA: Temple University Press.
Piza, E.L., Hatten, D.N., Carter, J.G., Baughman, J.H., and Mohler, G.O. (2023).
Gunshot detection technology time savings and spatial precision: An exploratory
analysis in Kansas City.
Policing: A Journal of Policy and Practice
.
https://doi.org/10.1093/police/paac097
Vovak, H., Riddle, T., Taniguchi, T., Hoogesteyn, K. & Yang, Y. (2021).
Strategies for Policing Innovation (SPI) in Wilmington, Delaware: Targeting
Violent Crime
. Washington, DC: National Police Foundation.
Watkins, C., Mazerolle L.G., Rogan, D. & Frank, J. (2002). Technological
Approaches to Controlling Random Gunfire: Results of a Gunshot Detection
System Field Test.
Policing
25: 345–370.
www.siue.edu/ccsvp
27
This report was prepared in March 2023 by Southern Illinois University
Edwardsville’s Center for Crime Science and Violence Prevention.
CCSVP works with Criminal Justice agencies and community organizations to
reduce violence and evaluate ongoing efforts to minimize gun violence. We
believe in collaborative, evidence-based and data informed actions to serve
public interests and those working in the criminal justice field.
Southern Illinois University Edwardsville
The Center for Crime Science and Violence Prevention
2300 West Main Street, Building A, Suite M222
Belleville, IL 62226
Contact us at:
ccsvpbelleville@siue.edu
www.siue.edu/ccsvp
