Social Media Data for Firearms Research: Promise and Perils

• Published date: 01 November 2022
• DOI: http://doi.org/10.1177/00027162231174320
• Author: Lisa Singh,Carole Roan Gresenz
• Date: 01 November 2022
• Subject Matter: Opportunities for Gun Violence Prevention

ANNALS, AAPSS, 704, November 2022 267

DOI: 10.1177/00027162231174320

Social Media

Data for

Firearms

Research:

Promise and

Perils

By

LISA SINGH

and

CAROLE ROAN GRESENZ

1174320ANN The Annals of the American AcademySocial Media Data for Firearms Research

research-article2023

Firearms research has been hindered by a persistent

lack of high-quality, reliable, and timely data. Using

social media data in firearms research is, therefore,

appealing: these data are large in scale, continuously

provided, passively obtained, inexpensive in some

cases, and easily combined with data from traditional

sources. We consider how social media data have been

used in firearms research to date and how they can be

best used moving forward. Despite the many advan-

tages of using social media data, there are key analytical

and ethical considerations associated with their use,

including understanding the underlying population,

constructing valid and reliable measures from shared

content, the possibility of false data and misinformation

through fake accounts, and concerns having to do with

the privacy and consent of social media users.

Notwithstanding these drawbacks, we show that there

is a compelling case for the continued development of

these data for firearms research and suggest principles

for their responsible use in this context.

Keywords: social media; firearms; gun violence; rep-

resentativeness; measurement; demo-

graphic inference; reproducibility; false

information

Accurate, timely, reliable, and comprehen-

sive data on firearm behaviors such as

acquisition and storage, firearm markets, and

firearm-related outcomes are cornerstones of

empirical firearms research, yet they are sorely

Correspondence: lisa.singh@georgetown.edu

Lisa Singh is a professor in the department of computer

science and the director of the Massive Data Institute at

Georgetown University. She has authored or coau-

thored over ninety peer-reviewed publications and

book chapters related to data-centric computing, for

example, data mining, data privacy, and data science.

Carole Roan Gresenz is an economist and professor at

Georgetown University in the McCourt School of

Public Policy and School of Health. She is a coauthor on

the first and second editions of The Science of Gun

Policy: A Critical Synthesis of Research Evidence on

the Effects of Gun Policies in the United States (RAND

Corporation 2018).

268 THE ANNALS OF THE AMERICAN ACADEMY

lacking. Persistent data limitations have hindered researchers’ ability to answer

fundamental research questions; indeed, Roman and Cook (2021) reflect that a

valid and reliable data foundation is nowhere more needed than in the realm of

firearms violence.

Roman (2020) profiles firearms data that have been discontinued, restricted

by policy, not collected, difficult to access, and/or delayed in availability. Firearm

ownership represents one example of data not systematically collected—despite

the fact that many gun policies (e.g., minimum age requirements for firearm

purchase, prohibitions on gun ownership for individuals with mental illness or a

history of domestic violence, extreme risk protection orders) are designed to

regulate who may legally own, purchase, or possess firearms (Smart et al. 2023).

The absence of longitudinal state-level estimates of firearm ownership has

affected researchers’ ability to assess the effects of these gun policies. New

efforts, including a workaround that provides historic gun ownership estimates

based on statistical blending of multiple data sources (Schell et al. 2020) and a

2021 National Firearms Survey that offers insight into purchasing behavior and

new gun ownership (Miller, Zhang, and Azrael 2022), have helped to fill, but do

not completely ameliorate, the information gap.

Another example is gun tracing data from the Bureau of Alcohol, Tobacco,

Firearms and Explosives (ATF) National Tracing Center, which uses serial num-

bers from guns used in crimes to identify information such as the guns’ original

manufacturer and purchase history. These data are useful to researchers for

understanding the ways that criminals acquire guns and the movement of fire-

arms across jurisdictions and states, but their availability has been curtailed by

provisions—known as Tiahrt amendments—annually attached to ATF appropria-

tions bills (Smart et al. 2023).

Other challenges include uneven law enforcement agency participation in the

National Incident-Based Reporting System (NIBRS) that constrains the compre-

hensiveness of data available on the use of firearms in crime; cost barriers that

curb researchers’ access to state-level information on nonfatal firearm injuries

and, thus, narrow the focus of most research to fatal events such as homicides and

suicides; and insufficient tracking systems that limit understanding of changes in

the supply of firearms from the advent of “ghost” guns, or homemade weapons

made from kits (Roman 2020; Smart et al. 2023; Stephens 2019). This is a repre-

sentative, though not an exhaustive, set of examples of the broad and systemic

data issues that plague firearms research.

However, the introduction and rapid growth of social media use over the past

two decades have generated a relatively new source of data for supporting

research and represent a potential opportunity for addressing at least some of the

pervasive gaps in firearms information. Several features of social media data con-

tribute to their general value for research. First, high rates and intensity of social

media use—roughly 70 percent of Americans indicate they use some form of

social media, with the majority of users reporting daily or more visits to their

preferred platform (Pew Research Center 2021)—together produce an extraor-

dinary amount of social media data. In 2022, for example, there were an esti-

mated 6,000 tweets per second and 500 million tweets per day (InternetLiveStats