A Data Driven Analysis of Gun Type and Mass Shooting Deaths

This is pretty long, so I tried to create some appropriate headings (ala a Scientific Paper) if you’d like to skip the parts you aren’t interested in.  I used this as an excuse to play around/refine my coding in Pandas/Python and I’m happy to share code if anyone would like it.

Intro/Motivation

This is a slightly more serious and longer post than I’ve done before, but a lot of things have brought the topic of Guns in America to my mind in a manner I wanted to write out some of my thoughts & feelings about it.  Here is a sampling of a few of the things that got me thinking about this:

1. A couple weeks ago, I went through this interactive article on FiveThirtyEight about Gun Deaths in America. (https://fivethirtyeight.com/features/gun-deaths/)
2. I’m listening to a D&D live-stream (Critical Role) where one of the characters has invented guns for this particular world, and he works through the various results and implications of his invention.
3. I just finished another read through of the Book of Mormon, which Mormons believe is a historical account written by prophets of some of the ancient civilizations in America (600 BC – 400 AD). **Spoiler Alert**: There are two instances covered within the book that result in two nations essentially exterminating one another, and the prophets keeping the record lament about the violence and death of their people.
4. In the wake of the shooting in the school in Parkland, Florida my wife and I have talked a lot about how difficult it would be to send your child to school and not have them come home (Our oldest is 5 and goes to Kindergarten now every day).
5. Having to grown up in Utah and lived for 2 years in North Carolina, some of that in very rural areas of NC, in addition to my time in Boston, my social media feed is very diverse and I’ve seen a lot of diverging arguments from both end on what should be done in the wake of the recent shootings. On one end of the spectrum, this has focused on mental health issues and background checks rather than limiting the types of guns that can be purchased, mostly from my more conservative, gun-toting friends (who’ve let me shoot their AR-15’s and taken me skeet shooting various times). The other end of the spectrum is focused on traditional forms of “gun-control” such as getting certain types of guns off the market, or limiting magazine sizes.

It is this last point I wanted to specifically focus my thoughts on and I wanted to delve deeper into, specifically the point of banning AR-15 type rifles.  This was on the table in Florida recently (https://www.cbsnews.com/news/florida-senate-passes-immediately-reverses-ar-15-ban-2018-03-03/), but didn’t get included in the final bill that passed.  Additionally, the LA Times ran an article that claimed that banning AR-15 type rifles would end mass shootings (http://www.latimes.com/politics/la-pol-ca-skelton-florida-shooting-gun-control-20180219-story.html).  I’d heard a similar sentiment from many of my friends, but heard it vehemently denied by other friends.  I couldn’t find any clear, data-driven articles that answered that topic in a way I found satisfactory to my liking, so that is the main topic of this blog post.  Ultimately, I feel the data suggest that there isn’t a clear black and white answer, which is probably closer to the truth than many news articles I often read seem to suggest.

Finally as a disclaimer, I expect some people to disagree with my thoughts, either a little or a lot, and I think that’s great.  I’d love to better understand the perspective of people who view things differently than me so feel free to reach out to me if you’d like to discuss this issue or my analyses more.

Methods/Data Source

My data primarily come from the following 2 sources in terms of identification of “Mass Shootings” and their related definitions:

1. Stanford Mass Shootings in America (MSA) database (https://library.stanford.edu/projects/mass-shootings-america). This started in 2012 after the Sandy Hook shooting. They aimed to create a historical record of Mass Shootings from 1966 to the present day.  I felt this was the most rigorous dataset of shooting records, as each record is corroborated by 3-7 sources.  However, their most recent published records were in April 2016, which was the spring before the shooting in Orlando, Florida.  For this database, the definition of a “Mass Shooting” was there were 3 or more shooting victims, not necessarily deaths.
2. Mother Jones US Mass Shooting database (https://www.motherjones.com/politics/2012/12/mass-shootings-mother-jones-full-data/). This is a non-profit who seeks to do “hard-hitting journalism” and created this database in 2012 after the Aurora, Colorado shooting. Their data was basically a google spreadsheet, so I made sure to confirm the details I was using with at least 1-3 other sources, one of which was Wikipedia typically, in full transparency.  This database was up to date as of March 10, 2018, so I added an additional 10 Mass Shootings to the MSA database, as I wanted to include Orlando, Las Vegas, and the most recent Parkland Shooting.  However, it is important to note that for this database, the definition of a “Mass Shooting” was slightly more subjective.  They excluded events that they determined to be related to “more conventional crime.”  As well, their threshold was higher than the MSA and focused on 4 or more shooting fatalities, rather than just victims.  There were only 10 additional shootings added from this database with my manual curation and double-checking, while the rest of the shooting events are in the published MSA database.

What I wanted to look at was what types of guns were used, and how many fatalities were associated with the type of gun used in the shooting event.  The MSA defined 3 gun types according to this:

1. Handgun = Handgun, pistols, revolver
2. Rifle = Firearm designed to be fired from the shoulder, with a barrel less than 16 inches in length
3. Shotgun = Designed to be fired from the shoulder, with a barrel less than 18 inches in length

For the basis of what I was looking at, AR-15 type guns fall under the Rifle category.

Results/Discussion

The data I specifically looked at from these databases was the total fatalities from these mass shooting events.  This included victim fatalities, as well as the shooter if it was a murder-suicide or the shooter was taken out by police.  I then compared that by gun type used/discovered for the investigation of the shooting (Handgun/Rifle/Shotgun or combinations of each).  I wanted to keep the visualization as close to raw data as possible to try and get an intuition for the data, and not just run some stats to try and make a specific point.  As I mostly wanted to do a first pass to better discuss this topic with people in my life, I’m purposefully trying to keep my analysis high level here, as I don’t think I’m going to come up with a solution to this topic on my own.  Here is the visualization I came up with:

The x-axis is number of fatalities.  The y-axis is the gun type, or combination of gun type for each designated mass shooting.  Each bar represents a single shooting event, with the width of the bar corresponding to the number of fatalities for that specific event.  The color of the bar roughly corresponds to the year, with red being earlier and blue being most recent.  Note that the MSA warns against reading too much into temporal effects for mass shootings, as their data highlights that reporting on mass shooting events has greatly increased in America over time, as evidenced by the extreme nonlinearity in the colorbar scale.  I’ve also labeled a few particular shootings as references.

The first thing that stood out to me was the huge total of fatalities by shootings using handguns alone.  For whatever reason (perhaps I’m pretty ignorant) I didn’t anticipate that column to be so large.  Secondly, the categories with rifles do seem to have more events with a larger bar size (indicating more fatalities), but they certainly aren’t the only places with large bar sizes.  The Virginia Tech shooting is a notable example, where there were 33 fatalities and the shooters used only semi-automatic handguns.  There are also a decent number of shootings in the 8-15 fatalities range that didn’t seem to get the national press that many other shootings seem to have gotten, across all gun types used in the shootings.

Ultimately, my main conclusion from looking at these data is that precision is very important in describing the problem people are trying to address.  From my experience interacting with friends and individuals throughout the political spectrum, a clear and precise articulation of the particular problem they’re talking about is not often communicated as clearly as their solution.  I do not think that the data suggest that removing AR-15’s from our country would end mass shootings, as the LA Times article seems to suggest.  This may be true if your definition of a mass shooting has a high threshold of 10-15 fatalities, but that feels like an arbitrary definition to support a certain definition.  However, it may be productive to discuss whether or not laws that limit magazine sizes and gun types may be useful in decreasing the number of individuals killed in a particular event (making the bar-size smaller on the graph), though that may not change the frequency of mass shooting events.

Additionally, the part of the data that surprised me for shootings in the Handgun category highlight that a sizeable number of people die from more frequent, smaller scale shootings that wouldn’t be impacted in any way by legislation focusing on limiting AR-15s and magazine sizes.  Is there a more general solution that might address the fatalities in both of these categories?  While I didn’t fully quantify it for a few reasons, these data also seemed to suggest that semi-automatic weapons (handguns, shotguns, or rifles) are more closely related to the number of fatalities in mass shooting events, rather than gun type alone.  The Virginia Tech shooting is the primary outlier example for that observation.  One might think a possible (and I would suggest extreme) gun-centric solution would be to prevent the sale of any semi-automatic weapons.  This is the point my friends with guns often step in and point to mental health and tighter background checks as solutions that might more globally solve some of these problems without any gun-centric legislation.  This post is already pretty long, so I’m not going to further discuss these matters in this post, but I’d be interested to hear more opinions from others on those types of approaches as well.

Conclusion

Ultimately, I think that the importance of precision in defining and articulating a problem and proposed solution is important in science as well.  In navigating a lot of the confusion, frustration, mud-slinging and mis-communication between my friends on the topic of gun control, I’ve seen how similar issues arise within science and it’s communication to the public.  I’ve certainly learned that the more precise I can be in articulating the problems I’m trying to address for my thesis has helped me focus and better communicate my findings in both writing and presentations.

TL DR; Deaths from mass shootings in America are more nuanced than being caused by individuals using a certain type of gun.  It’s important to be precise in defining and discussing both problems and solutions as both a civilian and scientist.

Kyle Hansen