Definition
As there is no legal definition of a “mass shooting” in the United States, disagreements exist over how best to operationalize the concept. However, the scholarly literature commonly defines mass shootings as shootings that result in four or more deaths by gunfire, excluding the perpetrator (Booty et al. 2019; Zeoli and Paruk 2019). For the purposes of this study, this is the definition of a fatal mass shooting that is used. The use of varying definitions results in different numbers of mass shootings being captured by different databases and may have affected the results of this study. For example, a 2019 analysis of five mass shooting databases found that there was little overlap in the number of shootings found across the five sources due to differences in definitions (Booty et al. 2019). While recent work has called for an expanded definition of mass shooting to include both fatal and non-fatal injuries, this work provides important information about the relationship between DV and mass shootings with four or more fatalities by gunfire, excluding the perpetrator (Booty et al. 2019).
Data and measures
For this analysis, we reviewed GVA data on mass shootings from 2014 to 2019. The GVA began collecting information about shootings in the United States in 2014, and the database tracks the date of the incident, city, state, and address of the incident, number killed, and number injured. The GVA defines a mass shooting as, “[Four] or more shot and/or killed in a single event [incident], at the same general time and location not including the shooter” (Gun Violence Archive n.d.-a, n.d.-b). However, as our focus was on mass shootings with four or more fatalities by gunfire, not including the perpetrator, we applied our definition to the GVA data. This resulted in a sample size of 128 mass shootings across the study period, with an average of 21.5 mass shootings per year (Gun Violence Archive n.d.-a).
We indexed our data by year and mass shooting and collected the number of deaths and injuries. Two authors independently reviewed news articles on each mass shooting and categorized whether it was DV-related (i.e., at least one victim of a mass shooting was a dating partner or family member of the perpetrator); 2) history of DV (i.e., the perpetrator had a history of DV but the mass shooting was not directed toward partners or family members); or 3) non-DV-related (i.e., the victims were not partners or family members, nor was there mention of the perpetrator having a history of DV). If there was discrepancy between the two authors in how an incident was coded, the case was discussed with the PI and the researchers came to a consensus. Of the 128 mass shootings, 120 incidents (94.0%) were coded exactly the same way by both coders. In the eight remaining mass shootings (6.0%), both coders met with the PI and a consensus was easily reached in all eight cases. While the 3/22/2017 shooting could have been coded as a history of DV mass shooting because the victims of the shooting did not include family or partners of the shooter, we have chosen to code it as a DV-related mass shooting because the perpetrator specifically targeted and intended to kill his wife.
Using a similar methodology outlined in Zeoli and Paruk’s (2019) paper, we applied our definition of a mass shooting to the data in GVA and reviewed each shooting entry and the articles listed on GVA. In addition to providing articles, GVA codes shootings based on several characteristics, one of them being domestic violence. However, understanding that GVA may omit articles, or information regarding a given shooting may change as stories develop, we did a comprehensive Google search of articles relating to each shooting. Search terms used included the offender’s name, the date of the shooting, the location of the shooting, as well as the words “domestic violence” to identify any mentions of domestic violence. For the higher-profile mass shootings, there were often dozens of news articles, including many articles in national news outlets that tended to have thorough information about the offender and the victims of the mass shooting. For lower-profile shootings, we reviewed the top 5–10 news stories, which often came from local news sources, to identify if there were media mentions of either a history of DV or if the victims of the shooting were family or intimate partners of the offender.
If the news articles mentioned that the victims were current or former intimate partners or other family members, we coded that shooting as “DV-related.” An “intimate partner” is a current or former spouse, dating partner, someone whom the offender had a child in common or lived with. A “family member” is someone related to the offender (either by blood, like a sister, brother, or cousin, or through the intimate partner, such as a mother-in-law) but who does not fall under the “intimate partner” category. If at least one news article mentioned that the offender had a known history of domestic violence (which could include a current or former partner mentioning that he or she was abusive), but the victims of the mass shooting were not intimate partners or family members, those cases were coded as a “history of DV” shooting. Actions falling under the “history of DV” category include violence (physical or otherwise) or threats of violence against a current or former intimate partner or family member (as defined above). When neither DV nor a history of DV was mentioned in any news stories, we classified the shooting as “non-DV related.”
Following the methodology used in Zeoli and Paruk (2019), if any victims of shootings with multiple perpetrators were family and/or intimate partners of the perpetrator, the mass shooting was classified as DV-related. If at least one of the perpetrators for shootings with multiple perpetrators had a history of DV, it was classified as a history of DV shooting. All other shootings were classified as non-DV related. There were 17 cases where the perpetrator was unknown, and these cases were removed from our main analysis. It is possible that there was a bias in our results based on how these unknown cases were classified.
During our preliminary analysis, we assessed the data for potential outliers in the total victim, victim death, and victim non-fatal injury counts; the Pulse Nightclub shooting in 2016 and the Las Vegas shooting in 2017 were of particular concern. We identified the Las Vegas shooting as an outlier as there were 471 total victims which was greater than three standard deviations from the mean (139 total victims). However, Pulse only exceeded three standard deviations from the mean for victim deaths, so it remained in the main analysis. A secondary analysis including the Las Vegas shooting in the analysis is available as Supplemental Materials (see Supplemental Tables 1–4).
Analytic methods
We conducted descriptive analyses to summarize the percent of mass shootings that were DV-related, history of DV, or non-DV-related. We conducted one-way ANOVA to examine whether there were differences in the average number of injuries or fatalities or the CFR between DV, history of DV, and non-DV-related mass shootings. We calculated the CFRs by category to reflect the total number killed over the total number injured and killed. We then calculated 95% confidence intervals for each CFR; category CFRs were determined to be significantly different at the p = 0.05 level if the 95% confidence intervals did not overlap. We analyzed how many perpetrators died during the mass shootings and noted whether they died by suicide or were killed by police. Finally, we created a “hybrid” category that combined DV-related shootings with history of DV shootings. A two-sample t-test was then conducted to determine whether this new hybrid DV-category had significantly different average victim fatalities and injuries from the non-DV-related shootings.
Analyses were conducted using Stata version 16.1 (StataCorp 2019). Institutional Review Board approval was not required for this non-human subjects review of publicly available data.