Can mass shootings be forecasted? Mohammad R.K. Mofrad and colleagues constructed an agent-based model of mass shootings within a Bayesian inference framework. In the model, a shooting arises from the intersection of three factors: the emergence of a motivated potential perpetrator, access to firearms, and the presence of a suitable target population. Based on demographic data from the U.S. Census, a per-capita rate of perpetrator emergence was estimated jointly with other model parameters. Access to firearms was modeled using the local density of Federal Firearms License (FFL) holders as a proxy and emerged as the dominant predictor of risk within the model.
The authors compared model predictions to mass shooting events recorded in the Mother Jones database from 1982–2024. This database includes around 160 events with three or more fatalities, excluding the perpetrator, that were not categorized as terrorism, gang-related violence, or a domestic dispute. The models assigned higher risks of shootings to areas where shootings occurred, supporting the model's ability to distinguish areas with different levels of historical risk. The relative importance of firearm availability in the model suggests that access to a firearm may be an important limiting factor in mass shooting risk.
The ten highest-risk areas by expected events per decade are Maricopa County, AZ; Harris County, TX; Dallas County, TX; Los Angeles County, CA; Kings County, NY; Clark County, NV; Orange County, CA; Queens County, NY; Cook County, IL; and Philadelphia County, PA. According to the authors, agent-based models can help identify areas at elevated risk of mass shootings, outperform standard statistical baselines, and may provide a potential framework for directing prevention resources.
PNAS Nexus
Forecasting the future of American mass shootings with Bayesian agent-based model calibration
22-Sep-2026