Global mass shootings: comparing the United States against developed and developing countries

ABSTRACT

This study compares mass shootings in the US against developed and developing countries (1998–2019). Findings indicate US mass shootings were more likely to involve workplaces, employment/financial problems, relationship problems, and multiple firearms. Mass shootings in all developed countries (including the US) were more likely than developing countries to involve foreign-born perpetrators, ideological motives, fame-seeking motives, schools, open-spaces, and handguns. Mass shootings in the US account for 73% of all incidents and 62% of all fatalities in developed countries. Mass shootings in developing countries were more likely to involve military and police perpetrators, rifles, and military/police locations. A discussion of findings offers insight for understanding and addressing the global mass shooting problem.

KEYWORDS:

• Mass shootings
• global homicide comparisons
• gun violence
• mass murder
• international mass violence

Acknowledgments

Thank you Christopher Zavala and Patrick Eisenhardt for your assistance with this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. This study uses the term mass shooting to reference incidents occurring in public locations and involving at least some victims chosen at random and/or for their symbolic value.

2. Lankford (2019, 2020) also questions Lott and Moody’s (2019) inclusion of incidents motivated by battles over sovereignty. Lankford, 2019 and Lott agree that battles over sovereignty should be excluded from mass shooting datasets. For instance, Lott (2015) stated, “to make a fair comparison with American shootings, I have excluded terrorist attacks that might be better classified as struggles over sovereignty.” However, Lankford (2019) suggests Lott and Moody’s (2019) data includes 500 foreign attacks involving battles over sovereignty.

3. This study does not include distinctions in incidents and fatalities between the US and developing countries because this has already been thoroughly explored in the previous debates over global mass shootings (Lankford, 2019, 2020; Lott & Moody, 2019, 2020).

4. These incidents are sometimes referred to as mass “public” shootings, especially when comparing all mass shootings (i.e., occurring anywhere and targeting anyone) with mass public shootings (Duwe, 2020; Krouse & Richardson, 2015). However, the abbreviated phrasing is used (see, also: Peterson & Densley, 2019) because the “public” component is often inherent in scholarly and general understanding of the phenomenon.

5. Although an exception is made for three organised terrorism incidents in the final charts addressing RQ3.

6. Lankford’s data includes mass shootings in the US and all other countries between 1966 and 2012, although only 1998 to 2012 were reviewed for this study. This data was used for Lankford’s (2016a, 2016c) global mass shooting studies, and was made available in the Lankford (2019) article. In this same article, Lankford (2019) provided Lott’s data for global mass shootings between 1998 and 2012. However, more data was obtained from Lott for US mass shootings in Lott (2020). Additionally, after corresponding with John Lott, he provided an extended version of other countries data from 2013 to 2017. From this point forward, these collective studies are simply referred to as the “Lankford” and “Lott” data.

7. See, Capellan and Gomez (2018) and Huff‐Corzine and Corzine (2020) for comprehensive lists of other publicly available mass shooting data reviewed for this study.

8. Careful consideration was given to addressing the conflicting incidents from the current data and TVP. This is because the debate over including these cases often occurred in the other sources used for this study. For instance, of the nine incidents included in the current study but excluded from TVP, some were included in the FBI (n = 3), Lankford (2019) (n = 2), Lott (2020) (n = 1), and NYPD (n = 1) data. Alternatively, of the ten incidents included in the TVP but excluded from the current study, some were included in the Lankford (2019) (n = 2), FBI (n = 2), and NYPD (n = 1) data. In other words, in all these well-regarded mass shooting studies, there were a handful of cases that could arguably be included/excluded. To address this concern, there was originally consideration given to including all 19 cases. However, Lott (2019) suggests studies are guilty of over-inclusion of US cases. Alternatively, consideration was given to dropping all 19 cases. Although again, Lankford (2019) suggests Lott and Moody (2019) are guilty of under-inclusion of US cases. In the end, it was decided the best approach was to include the original nine and exclude the ten from TVP. This was done because: (1) it provides a balance between the Lankford and Lott debates; (2) TVP is not infallible, for instance, dropping 2 cases from V1 in V2 of its release; and (3) the current data size (N = 101) is almost identical to TVP (N = 102) over the same period. In other words, the same severity of the problem is captured in both datasets, despite slightly different cases included.

9. Australia, Austria, Belgium, Bulgaria, Canada, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, United Kingdom, and the United States.

10. For two variables (i.e., law enforcement history and victim-specific motive), TVP had the variable information (they include source links for each variable), but they did not code these specific variables. In other words, TVP includes information on employment, which was reviewed for coding the law enforcement history variable. They also include information on victims, which was used to code the victim-specific variable.

11. This coding strategy was also used for the few remaining US cases not captured by the NYPD, FBI, and TVP.

12. In 38 incidents, the perpetrators age was missing. All missing age variables were in developing countries.

13. This study measures variables at the incident level so perpetrator variables could be included in the full multivariate model (Silva, 2021a). Measurements of co-offender incidents consider if one of the two perpetrators had any of the variable characteristics. For instance, if there was one female and one male, then the variable was coded as female. The age of co-offenders was averaged between the two perpetrators (Lankford & Silver, 2020; Silva, 2021a). To ensure the reliability of findings, sensitivity tests were run with all offenders and one randomly selected offender. The findings were the same. This is because: (1) co-offender incidents were rare (3%), and (2) they had similar characteristics (i.e., close in age, both had a fame-seeking motive, etc.).

14. Ideally, only firearms that were fired during the incident would be included; however, this information was often not reported in news articles (Greene-Colozzi & Silva, 2020). Instead, media reports often described the number and type of firearms present without differentiating which firearms were fired.

15. There were missing firearm variables in developing countries including handgun (n = 25, 18%), rifle (n = 29, 21%), shotgun (n = 29, 21%), and more than one firearm (n = 17, 13%).

16. These three incidents include the 2014 Belgium shooting, as well as the January 2015 and November 2015 shootings in France.

17. Certain variables were excluded from the logistic regression because they had zero cases in developed countries involving an examined characteristic (i.e., female, law enforcement history, multiple shooters, and military/police target).

18. However, most of the fatalities in France were the result of the two incidents that could have arguably been excluded because they were organised by terrorist groups, involved more than two perpetrators, and occurred over many days. The January 2015 coordinated attacks resulted in 17 fatalities, and the November 2015 attacks resulted in 130 fatalities. These two incidents make up 82% of the current mass shooting fatalities in France (147 / 179).

19. These findings are rooted in significant variables in Table 2 (i.e., US versus all countries), that were not significant in the developed columns in Tables 3 and 4. A comparison between all developed (including the US) and developing countries was conducted, however, the information was redundant to what was already identified.

Additional information

Funding

This research was supported (in part) by a Summer Stipend from the Research Center for the Humanities and Social Sciences at William Paterson University.

Notes on contributors

Jason R Silva

Jason R. Silva is an Assistant Professor in the Department of Sociology and Criminal Justice at William Paterson University. His research examines mass shootings, terrorism, and mass media. Silva’s recent publications have appeared in Aggression and Violent Behavior, American Journal of Criminal Justice, Justice Quarterly, and Victims & Offenders.