MPD has cool tech to detect gunshots, but does it help?
An initial look at ShotSpotter and why some police departments don't like it
There’s enormous demand among police departments and elected officials for “crime fighting” technology. Widespread public dissatisfaction with crime, especially shootings, have fueled rapid adoption of “Acoustic Gunshot Detection Systems” (AGDS) and in Washington, DC that led to the deployment of the ShotSpotter technology back in 2006. MPD appears to spend about $517K ( just~0.1% of its budget) on the contract with ShotSpotter and in 2022 the system reported 7,017 “alerts” of gunfire. In this post we’ll explore this data, some concerns about the system and possibly some ways to make it more useful for fighting crime.
ShotSpotter uses microphones to triangulate the location of gunshots. This kind of system is generally pretty accurate but is not perfect. ShotSpotter staff confirm if the recorded audio is likely a gunshot and then relay the location to MPD.
ShotSpotter covers “roughly 18 of the city’s 68 square miles” with no reported sensors in wealthy Police District 2D. In the other 6 districts, over 80% of reported gunshots are in districts 6D and 7D (east of the river).
This is disproportional to reported gun crimes, where ~46% of gun crimes in 2022 were in police districts 6D and 7D. This may reflect lower reporting of crimes, more ShotSpotter coverage or a mix of factors.
Gunshots spiked enormously in the summer of 2020, peaking at 1,931 in June 2020 before decreasing to a monthly average of 585 in 2022. This is still 27% higher than the pre-COVID average but that is likely skewed by MPD adding more sensors in 2019.
Partial 2023 data through February 6th (the most recent release) show an equivalent of 916 gunshots per month, which would be a significant increase over 2022 if that trend continues
42% of reported gunshots occur on the weekends and 56% occur between the hours of 9PM and 3AM
The Smart Policing Initiative (DOJ), Police Chief Magazine and the Inspector General of Chicago have questioned or outright denied that ShotSpotter actually helps fight crime
Law enforcement criticisms of the system focus on a lack of actionability from the data and the additional workload it adds to police to respond to reports of gunfire where no physical evidence is found
ShotSpotter data is reported by Police District so it may be helpful to look at the district map:
The distribution of ShotSpotter data has been pretty consistent with ~80% of alerts being issued in police districts 6D and 7D. The one notable exception is in 2020 when 10% of alerts were in 1D:
It’s important to note that this distribution of alerts is more skewed than reported gun crimes. Many civil libertarian concerns about these kinds of systems claim that they reinforce police perceptions of certain areas as “high crime” and impact how they interact with residents. Another operational concern is that if officers in 6D and 7D are responding to 10X as many ShotSpotter alerts as their colleagues in other districts they may be more burnt out and/or less available to respond to other calls.
MPD has only released detailed ShotSpotter data through June 2022 but we can still see a clear trend in monthly totals:
From Summer 2015 to Spring 2019 DC averaged ~400 gunshot alerts per month
In 2019 MPD added ShotSpotter sensors and we see a gradual increase in alerts
Once COVID hits in early 2020 we see an enormous spike in alerts, peaking at 1,931 in June 2020 and remaining elevated until winter 2020/2021
Since then we’ve been in a “new normal” of ~600 alerts per month, significantly above the pre-COVID baseline. This is further evidence, along with reported crimes and MPD arrest data, that more criminals are using guns than pre-COVID.
There are clear and unsurprising trends in when gunshots occur:
42% of reported gunshots occur on the weekends
56% occur between the hours of 9PM and 3AM
Friday night into early Saturday morning (Midnight-3AM) are by far the worst times with the highest volume of alerts
In theory, AGDS systems like ShotSpotter allow police to respond to shootings faster and to a more accurate location than by relying on residents calling 911. Theoretically this makes it more likely for police to apprehend suspects and/or provide emergency medical care to victims. Unfortunately, there is mixed evidence that these kinds of systems help. The Smart Policing Initiative, a project of the Bureau of Justice Assistance in the Department of Justice, summarized the evidence like this:
The St. Louis studies were referenced by Police Chief Magazine’s scathing review:
”The current evaluation, for example, finds no crime trend differences between AGDS neighborhoods and similar areas. The only reasonable conclusion from this is that AGDS produce less actionable data and are less efficient than traditional sources of information.”
More recently the Inspector General in Chicago also declared the system ineffective:
”From quantitative analysis of ShotSpotter data and other records, OIG concludes that [Chicago Police Department] responses to ShotSpotter alerts rarely produce evidence of a gun-related crime, rarely give rise to investigatory stops, and even less frequently lead to the recovery of gun crime-related evidence during an investigatory stop.” When discussing if the value of the system was worth the costs they found “Such a value is not clearly demonstrated by presently available data.”
This post is not a full assessment of ShotSpotter’s efficacy in DC but it’s worth noting that MPD has had this technology in the neighborhoods with the most gun crime since 2006. It’s obviously possible that gun crime could be even worse without this system but it’s not self-evident that it has helped. In most critical evaluations of ShotSpotter and AGDS one key problem is that police can only respond so quickly to alerts. Because of this they often arrive after shooters (and some targets) have fled the scene. In Chicago for instance, police only found evidence of a crime 14% of the time they responded to ShotSpotter alerts. The problem isn’t the technology itself, it’s how police departments can use it. While the total cost of the system is a relatively small share of MPD’s budget, hopefully MPD is monitoring how useful the system is given the real cost of scarce police time in responding to these alerts (which disproportionately hit 6D and 7D). It’s a real conundrum because it would be irresponsible to ignore any gunfire alert but in aggregate the volume of alerts may not be helping. MPD is unlikely to discontinue a system that’s been in use for so long (and a vendor whose new DC office opening was attended by both Mayor Bowser and Chief Contee) but maybe it can be improved.
A veteran friend shared one possible idea to augment the usefulness of this technology that I will include here not as a “fact” but simply an idea that may be helpful. While police are limited by the speed of their vehicles and their visibility is greatly limited by buildings, unmanned aerial vehicles (UAVs or “drones”) have both greater speed and visibility. To my knowledge, MPD does not use UAVs (possibly due to FAA restrictions on their use in DC’s airspace). I’m sure that many readers are uncomfortable or outright opposed to any increase in MPD’s surveillance capabilities; but having the ability to more quickly respond to shootings and track suspects fleeing on foot or by vehicle could make gunshot detection more useful to MPD. From a Constitutional perspective, sending an unarmed UAV to look at the location of a reported gunshot location ought to be no more problematic than dispatching armed police officers to that location.
If MPD seriously entertained the use of UAVs (and could sort out permission to use them from the FAA) it could also possibly have synergies with License Plate Readers (LPRs). Since UAVs are much less expensive to own and operate than traditional police helicopters, it could be cost-effective to vector a UAV to find/track a stolen vehicle that is seen by a LPR. This could also mitigate the downsides of MPD’s “chase” policy and still allow police to track a suspect vehicle via UAV without engaging in a dangerous vehicular chase. No technology is ever a silver bullet but there possible operational upsides here. It’s certainly something worth piloting in a limited capacity if MPD could sort out the regulatory barriers.
Aside from this one particular idea (which again, may not be feasible), I’d encourage leaders in MPD and DC government to be willing to experiment with and evaluate technology solutions. We have significant vacancies and recruitment pipeline challenges across a number of key government agencies including MPD that are unlikely to be solved for years. Technological solutions that save time, make employees more effective or increase useful capabilities need to be on the table at least as workarounds. As we’ve seen with ShotSpotter, how technology is implemented into workflows is just as important as how “cool” the technology is and vendor effectiveness needs to be evaluated. But DC is blessed with a large number of universities, think tanks and an independent Auditor that can help on the evaluation front. As skeptical as I am of ShotSpotter’s current usefulness, it’s exactly the kind of idea that we should be trying, testing and improving.
Anecdotally it has been helpful in finding victims when no one calls 911 and it is useful in determining when exactly the shooting started. But as with all things it doesn't hurt to evaluate and see what we can do better!