Do Acoustic Gunshot Detection Systems Actually Improve Policing?

AI Summary: Research question: Does adding an acoustic gunshot detection system (AGDS) to police CCTV cameras help identify more actual shootings?

Bottom line: Not in this study. A partially randomized field experiment in Philadelphia found that installing acoustic gunshot detection sensors increased reported gunshot-related incidents by 259%, but did not significantly increase the number of confirmed (founded) shootings. Instead, the technology substantially increased police workload by sending officers to many incidents where no evidence of a shooting could be found. The findings suggest police agencies should carefully consider whether the primary value of gunshot detection technology lies in incident detection or in its investigative and evidential benefits.

Acoustic gunshot detection systems have become increasingly common in American policing. The promise is straightforward: detect gunfire immediately, dispatch officers faster, and improve public safety. Vendors often argue that these systems reveal shootings that would otherwise go unreported while providing investigators with valuable evidence.

But do they actually accomplish those goals?

That question motivated our evaluation of a pilot acoustic gunshot detection system deployed by the Philadelphia Police Department. Rather than relying on vendor claims or anecdotal success stories, we conducted a field experiment to determine whether the technology produced measurable operational benefits for police. Read this AI-generated summary for more information.

The Technology

Philadelphia installed 17 acoustic gunshot detection sensors that were integrated directly with existing public CCTV cameras.

When the system detected what it believed to be gunfire, the associated CCTV camera automatically rotated toward the suspected source of the sound while simultaneously notifying the Police Department's Real Time Crime Center. Analysts could immediately examine the video feed and dispatch officers to investigate the incident.

The idea was simple. If many shootings go unreported by residents, the system should identify additional genuine shootings that would otherwise remain unknown to police.

A Real-World Experiment

To test that hypothesis, we designed a partially randomized field experiment.

Seventeen camera locations equipped with acoustic sensors were matched to comparable control locations without sensors. We examined eight months of police incident data before implementation and eight months afterward, comparing gunshot-related incidents around both treatment and control sites.

The primary outcome was not simply the number of calls generated. Instead, we examined founded shootings—incidents where responding officers found physical evidence such as shell casings, victims, witnesses, or bullet damage that confirmed a shooting had actually occurred. This distinction is important because police often receive reports of shots fired that ultimately cannot be substantiated.

What We Found

The system certainly generated more police activity. Following implementation, gunshot-related incidents increased by approximately 259 percent around the sensor locations compared with the matched control sites. On its face, this appears impressive.

However, the more important finding was that confirmed shootings did not increase significantly.

In other words, the technology produced many more police responses but did not uncover substantially more verified shootings than officers were already finding through conventional reporting mechanisms. This meant the system significantly increased officer workload without producing a corresponding increase in confirmed criminal events.

Why Didn't It Work Better?

Several explanations are possible.

Some of the additional incidents may have been genuine gunfire where officers simply could not locate evidence. Shell casings are often difficult to recover, especially in poor weather, if shots were fired from moving vehicles, or if the actual shooting occurred outside the camera's field of view or inside nearby buildings. An incident classified as "unfounded" does not necessarily mean no shots were fired.

At the same time, acoustic systems are not immune to false activations. Loud mechanical noises, construction activity, slamming dumpsters, or other impulsive sounds can sometimes resemble gunfire closely enough to trigger the system. Those false positives inevitably consume police resources. Both explanations likely contributed to the results.

An Important Operational Lesson

One of the more important conclusions from the study is that agencies may be asking the technology to accomplish the wrong objective.

If the goal is to discover large numbers of previously unknown shootings, our findings suggest that current-generation acoustic gunshot detection systems may not deliver that benefit. However, if the goal is to improve investigations by automatically directing cameras toward active shooting scenes and capturing valuable video evidence immediately after shots are fired, the technology may still provide considerable operational value.

This distinction is important because technology should be evaluated against the objective it is actually capable of achieving, rather than the objective we hope it will achieve.

Why This Research Matters

Police technology is often adopted with considerable optimism but relatively little independent evaluation.

This study demonstrates why rigorous experimentation is essential. A technology that appears highly effective because it generates more alerts may not necessarily improve policing outcomes. Measuring workload, confirmed incidents, and operational effectiveness provides a much clearer understanding of whether new systems actually help police agencies accomplish their goals.

Evidence-based policing requires that new technologies be subjected to the same scientific scrutiny as any other policing strategy. The results are not always what we expect—but they are exactly what police leaders need when deciding how to invest limited public resources.

Citation

Ratcliffe, J. H., Lattanzio, M., Kikuchi, G., & Thomas, K. (2019). A partially randomized field experiment on the effect of an acoustic gunshot detection system on police incident reports. Journal of Experimental Criminology. 15(1): 67-76.