Philadelphia Acoustic Gunshot Detection study

The study evaluates whether introducing an acoustic gunshot detection system (AGDS)—linked to the city's CCTV cameras—would increase police awareness of gunfire incidents in Philadelphia. The key question: Does the technology uncover more gunfire events that the public fails to report, and does it translate into more confirmed shootings? The research was undertaken by Jerry Ratcliffe along with numerous colleagues in the Philadelphia Police Department. Important note: This is not an evaluation of ShotSpotter/Sound Thinking, but another system called Sentri. It works in a different manner.

Research design

Researchers and the Philadelphia Police Department deployed 17 SENTRI acoustic sensors paired with CCTV cameras across the city. These sensors automatically directed cameras toward the suspected source of a gunshot and alerted the Real Time Crime Center.

The experiment used a partially randomized, block-matched design. Camera sites were paired based on crime levels, socio-economic indicators, and land use. Half were randomly assigned to receive AGDS sensors, while the others served as controls. Operational challenges (technical issues and site restrictions) meant that full randomization was not possible, but 17 valid pairs remained.

Data on gunshot-related incidents and confirmed (founded) shootings were collected within 900 feet of each site for eight months before and after the system went live (April 2015 – August 2016). Analyses used multilevel negative-binomial regression models to estimate treatment effects while accounting for camera-specific variation

Key Findings

After the AGDS was activated, gunshot-related incidents rose by ≈ 259%, showing the system brought many more suspected shootings to police attention.
However, there was no significant increase in confirmed (founded) shootings — cases where evidence such as shell casings or victims was found.
The system therefore expanded police workload by generating more “unfounded” calls without increasing verified gun-crime detections.

Interpretation

The findings indicate that the acoustic system successfully detects more possible gunfire sounds, but many do not correspond to confirmed criminal events. This mismatch suggests that false positives and unverifiable alerts (e.g., from loud noises, vehicles, or environmental sounds) substantially inflate call volume.

While the study does not imply that all unfounded activations were false alarms—some gunfire may have occurred out of camera view—it demonstrates that the system did not enhance measurable crime outcomes during the trial period. Instead, it redirected limited police resources to additional investigations with little evidentiary payoff.

We suggest reframing the purpose of AGDS technology: rather than aiming to reveal unreported shootings, its main utility may lie in post-incident investigation and evidentiary capture. When coordinated with strategically placed cameras, it could help document offenders’ movements or corroborate other evidence in active-shooting cases.

More details

Limitations

Because the study was only partially randomized, causal inference is limited. The trial covered just eight months post-implementation, and several sensors malfunctioned or were damaged, slightly reducing coverage. The system’s early version suffered a high rate of false alarms—up to 80–90 percent in the first months of operation, according to police reports—making officer fatigue and response downgrades likely.

Implications

We show that simply adding acoustic gunshot detection does not automatically improve gun-violence outcomes. For agencies considering such technology, the key is aligning it with clear operational goals—for instance, investigative documentation rather than 911 call replacement—and ensuring officers maintain trust in system accuracy.

Future research should extend observation periods, improve sensor reliability, and test whether redesigned deployments (e.g., camera clusters around high-risk intersections) can deliver genuine public-safety benefits.

Details

The main article is available on this website as publication number 69.

Full 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. https://doi.org/10.1007/s11292-018-9339-1

Note: This summary was created by ChatGPT5.