Gun Violence Data and Methodology
What is the Gun Violence Archive? Is it valid? What is the FHR, how is it calculated, and what does it mean?
I obtained firearm fatality and injury data from the Gun Violence Archive (GVA). The GVA an open-source database that aggregates daily US-based firearm violence incidents reported by over 7500 news media, government, law enforcement, and commercial sources. Every incident in the GVA is a source-verified shooting event that includes data about the number of victims (separated intro columns for injuries and fatalities), event date, city, state, and street address.
The decision to use the GVA to obtain city-level data was made after careful consideration of the strengths and limitations of various firearm injury databases. For instance, official firearm homicide and injury statistics compiled by the Centers for Disease Control and Prevention (CDC), Federal Bureau of Investigation (FBI), law enforcement, and large hospitals have been shown to be accurate and consistent across official sources, especially for cities with populations of 100,000 or greater (Loftin et al., 2015). However, data from less-populous cities and counties is often under-reported, so the CDC and FBI estimate missing data, usually at the state or country level, which forecloses analyses including communities that don’t report, which tend to be smaller cities (Pridemore, 2005; Conner et al. 2019). Because the GVA relies upon a different methodology for data collection it is less likely to systematically under-report at the community level. Furthermore, Donnelly et al. (2018), Conner et al. (2019) and Herring and Kersten (2020) both found high agreement across open sources (including the GVA) and the CDC’s official National Violent Death Reporting System (NVDRS) or Wide-ranging OnLine Data for Epidemiologic Research (WONDER), coupled with small and unsystematic differences across sources.
GVA data has also been used in previous peer-reviewed academic research to study mass shootings (Booty et al., 2019); community gun violence (Johnson et al., 2021); and trends during the Covid-19 pandemic (Ssentongo et al., 2021; James & Menzies, 2022). News media also frequently rely upon Gun Violence Archive data in reporting on mass shootings and officer-involved firearm fatalities. Considering the objectives of my study and the desired unit of analysis, coupled with researching demonstrating high validity of GVA data compared to official databases, I determined the GVA to be a valid source for obtaining annual, city-level firearm fatality and injury count data.
Working from my list of cities with populations ≥ 50,000, plus principal cities of Micropolitan Statistical Areas derived from the US Census, annual city-level firearm incident lists for the years 2015-2021 were retrieved through the GVA database. This totaled approximately 318,891 incidents reported across my 1328 cities. With this, plus population data from the US Census Bureau, I calculated annual rates per 100,000 for firearm homicides and injuries for each city for the years 2015-2021. In my analyses, I report Fatal Firearm Injury (FFI) counts, Non-fatal Firearm Injury (NFI) counts, Firearm Homicide Rates (FHRs) and Firearm Injury Rates (FIRs). FHRs and FIRs are always reported per 100,000 while counts are absolute numbers. While any data may contain discrepancies, these are not population estimates but rather aggregated counts and calculated rates based on observed source-verified incidents from the Gun Violence Archive.
Source: Excepted from, Wade, M. M. (2023). “Not as Bad as the ‘90s”? Firearm Violence in Small, Mid-Size, and Large US Cities, 2015–2021. Homicide Studies, 10887679231163287.
[1] If zero firearm fatality or injury events were retrievable for a city during all seven years, the city was deemed to either have no firearm violence or no data about such violence to analyze and was omitted from the analysis. This occurred fewer than five times and exclusively in micropolitan cities with populations smaller than 5,000.
[2] GVA incidents are inclusive of both intentional and unintentional incidents (but not suicides). I nonetheless use the terms “firearm homicide” and “fatal firearm injury” interchangeably throughout my analysis, since 99% of non-suicide firearm fatalities are the result of intentional violence (Gramlich, 2022).