UNIVERSITY PARK, Pa. — On average, more than 46,000 people die from gun violence in the United States each year, with tens of thousands additional injuries, according to Giffords Law Center. But data is still limited about firearm ownership and use, the causes and consequences of gun violence, and the impacts of policy and technological interventions, which can leave policymakers with an incomplete picture when making decisions around this complicated and sensitive topic.
A team of researchers from Penn State and the University of Pennsylvania have compiled the latest gun violence research from experts in a variety of fields in a new book, which highlights how statisticians and data scientists can contribute their expertise in working with incomplete or biased data and advancing new analytical methods. The book, titled “Gun Violence: The Statistical Issues,” explores the state of data in this area, opportunities and challenges with using new analytical methods including artificial intelligence, and the benefits of working across disciplinary lines to help understand and mitigate the effects of gun violence
In this Q&A, book editors James Rosenberger, professor emeritus of statistics at Penn State who has held leadership roles with the National Institute of Statistical Sciences, the Division of Mathematical Sciences for the U.S. National Science Foundation and the American Statistical Association; Lingzhou Xue, professor of statistics at Penn State; and Charles Loeffler, associate professor of criminology at the University of Pennsylvania; speak about why they edited this book and the importance of involving statisticians in this field.
Q: How did you get involved with gun violence research?
Rosenberger: When I joined the National Institute of Statistical Sciences in 2017, we were considering how statisticians and data scientists could bring their expertise to issues that face society, and gun violence really stood out. Gun violence is a complex, sensitive issue in the United States, and regulation has been polarizing. And for nearly 20 years, the U.S. government prohibited federal funds from supporting research in this area, so the data sources we have about firearm ownership and use as well as the impacts of policy are often incomplete or biased. So, in 2019, as the limits on funding this research were lifting, we decided to bring together experts, researchers, policymakers and activists to discuss this issue through a statistical lens at the inaugural Ingram Olkin Statistics Serving Society Forum. Lingzhou and several others helped plan the meeting.
Xue: After the success of our first meeting, we decided to hold another meeting in 2024, this time virtually, and we brought on Charles to help with the planning. In this meeting, we discussed research advances and data-driven insights into the complex dynamics of gun violence as well as evidence-based solutions to mitigate its devastating effects. We invited some of the presenters from that meeting to contribute to this book, which resulted in a multi-disciplinary perspective from researchers in criminology, political science, public policy, public health, sociology and statistics.
Q: How is data around gun violence incomplete? How can statistical expertise support research in this field?
Loeffler: Gun violence data comes in a variety of forms and from a variety of sources, including crime reports from police departments, statistics from the public health system, and newer open-source and crowd-sourced databases maintained by news and other non-governmental organizations. Some kinds of data are well reported. For example, gun fatalities are usually very well investigated by police. However, police-involved shootings are not always reported uniformly, and non-fatal shootings are often not well documented. But it’s important to consider all kinds of usage to truly understand the scope of gun violence. It’s possible that newer technologies like gunfire detection technology may be able to help fill in some of the gaps, and one chapter in the book explores the accuracy of this technology, how it is being used and its impact on law enforcement.
Xue: Different data sources have their own strengths and limitations, and statisticians bring a practiced perspective about working with and assessing different kinds of data and combining datasets. By carefully selecting methodologies and acknowledging the limitations and quality of the available data, we can help ensure that analysis can withstand scrutiny by all involved
Q: How are modern methodologies changing this field of research?
Rosenberger: There are many analytical challenges in gun violence research given the incompleteness of the data, especially in modeling and inferring the effects of policy. In the book, we explore several contemporary statistical and machine learning methodologies that can potentially be used to rigorously make sense of the data, including network models, synthetic control methods, point processes models and neural networks. In one chapter, the authors analyze spatial temporal data from Atlanta, Georgia, and, in another chapter, the authors explore methods to estimate the impact of gun policy on gun violence.
Q: What other topics does the book cover? What’s next for this line of work?
Loeffler: Several chapters explore the delicacy of measuring and studying police-involved shootings, racial disparities in police shootings and the increased rate of mass shootings since the COVID pandemic. Another explores how a possible new source of data — ammunition recovered by police after a wide variety of arrests — could be used to shed light on the market for illegal ammunition.
Xue: Jim and I will also be hosting a session at the annual meeting for the American Association for the Advancement of Science in February 2027 in Chicago that extends this work, bringing in perspectives from statistics and behavioral economics. Several speakers will discuss research, such as how interpersonal conflict and emotional escalation drive shootings; infrastructure that could fill data gaps in non-fatal shootings; and a framework for computer simulations that, for example, could estimate the impact of shifting gun policy.
Rosenberger: All of this research is challenging and a work in progress, but it is vital for understanding and mitigating gun violence. Our hope is that providing a multi-disciplinary perspective with a statistical framework will encourage high-quality published research that can enable more definitive conclusions about gun policy and interventions.