Dual-domain analysis of gun violence incidents in the United States

TL;DR

This study employs advanced time and frequency domain techniques to analyze state-level gun violence trends in the US, revealing periodic patterns, behavioral shifts, and impacts of key events like COVID-19.

Contribution

It introduces a dual-domain analytical approach combining spectral density estimation and time series analysis to uncover complex gun violence dynamics.

Findings

Identification of changing periodic behaviors across states

Detection of behavioral shifts linked to key societal events

Challenging common assumptions about gun violence prevalence

Abstract

This paper applies new and recently introduced approaches to study trends in gun violence in the United States. We use techniques in both the time and frequency domain to provide a more complete understanding of gun violence dynamics. We analyze gun violence incidents on a state-by-state basis as recorded by the Gun Violence Archive. We have numerous specific phenomena of focus, including periodicity of incidents, locations in time where behavioral changes occur, and shifts in gun violence patterns since April 2020. First, we implement a recently introduced method of spectral density estimation for nonstationary time series to investigate periodicity on a state-by-state basis, including revealing where periodic behaviors change with time. We can also classify different patterns of behavioral changes among the states. We then aim to understand the most significant shifts in gun violence since numerous key events in 2020, including the COVID-19 pandemic, lockdowns, and periods of civil unrest. Our dual-domain analysis provides a more thorough understanding and challenges numerous widely held conceptions regarding the prevalence of gun violence incidents.