Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Mass Shooting Deaths in the US, 1982-2022
I found a data source yesterday for mass shootings in the US since 1982. I’ll discuss the technical details below. Here is my graph, which some of you may find interesting:
Click for a larger version.
The blue bars are the actual count of total fatalities in mass shootings, for each year. The 2022 figure is annualized, based on the reported mass shootings through yesterday (May 24, 2022). It includes the terrible school shooting in Uvalde, TX.
The orange line is the three-year moving average. The use of this type of data smoothing can help with spotting trends.
The yellow line is the overall 40-year trend line, from 1982 to 2022. Over the entire period, there is a pronounced upward trend. However, the moving average (orange line) didn’t suggest much of a trend until after 2004. So I calculated different trend lines for 1982-2004 and 2004-2022.
The green line is the trend line for 1982-2004. As you can see, there is a small downward trend (not statistically significant).
The red line is the trend line for 2004-2022. As you can see, there is a notable upward trend. The average increase in mass shooting fatalities has been 2.7/year. The slope is statistically significant (p=.0157).
The overall number of fatalities remains very small, approximately comparable to deaths by bee, wasp, and hornet stings (as discussed in my post yesterday). However, it does appear that this problem is getting worse.
This analysis, of course, relies upon the accuracy of my data source, discussed below. If a number of older mass shootings were not included in the database that I used, the conclusion about the recent upward trend would be incorrect.
Published in GunsTechnical notes:
My data source is a mass shooting database by Mother Jones (here). The explanation for their methodology is here.
Obviously, we don’t have a full year of data for 2022. I decided to include 2022, and simply annualized the total fatalities reported through yesterday (May 24, the date of the Uvalde shooting). The total reported fatalities for 2022 is 33, which I annualized to 83.65.
I made one adjustment in the Mother Jones data, by excluding mass shootings with three reported fatalities. The criteria for inclusion changed in 2013. Prior to 2013, only mass shootings with four or more fatalities were included, while beginning in 2013, mass shootings with three or more fatalities were included. To make the data comparable over the entire 40-year period, I omitted the mass shootings with three fatalities (18 such shootings with 54 fatalities).
I wonder how it tracks adjusted for total population?
Great chart, Jerry. You have nicely highlighted an inflection in 2003. (DHS formed?)
Are fatalities the only consideration? There was an incident after a Bucks game in Milwaukee on May 13 in which twenty-one individuals sustained gunshot wounds. But no one died, so I guess it wasn’t a mass shooting.
Sure seems like one to me.
Or does a mass shooting have to have been perpetrated by a single person? But non-fatal casualties don’t count?
I’m not even disputing your larger point, which I assume is that the risk of dying as a result of a mass shooting is very low. Stipulated. I’m just pointing out that there’s more to the issue than the net body count.
It doesn’t make much of a difference. Here you go:
I had the same thought myself, and it wasn’t hard to make the adjustment in Excel.
This graph uses fatality rates per 100,000 population. The slope of the 2004-2022 trendline — in red — is 0.000779. Using the current population of about 334 million, that’s an annual increase of about 2.6 in fatalities in mass shootings. The same trendline, without adjusting for population, was 2.7.
Fatalities are not the only consideration. This is the data set that I found.
The data does include information on the number of people injured, but I didn’t include this in the analysis, because I think that it creates a problem — for example, a mass shooting with 4 fatalities and 3 injuries would be included, but a mass shooting with 3 fatalities and 50 injuries would not be included.
The data set only includes mass shootings involving a single shooter, with two exceptions — Columbine and Westside Middle School. These two, in 1999 and 1998 respectively, involved two shooters.
There are other limitations in the data set. The methodology page states:
Thanks for the additional information. In light of these “limitations in the data”, I guess my takeaway is that this doesn’t illuminate much of anything useful.
I’ve learned (thanks, Ricochet!) to look for what happens if a given category of gun killing—say, mass shootings of vulnerable strangers by a lone lunatics— is grouped together with other killings that might not involve guns but are similar in other ways. So the Texas event is linked to the Brooklyn event and the Buffalo event but also the Waukesha event…but maybe not to a multiple-victim gun homicide that resulted from a gang fight in Chicago, say.
Of course, listing them this way doesn’t exactly provide comfort, does it?
‘Going Postal’ first appeared in print in 1993.
The Columbine shooting occurred in 1999.
Yes, Steve Scalise might just as well have been counted as a fatality in that he was only saved by a hair with the most modern surgery. But he apparently doesn’t show up in the categorization numbers at all if I understand correctly.
OK. 2004 data point. The Federal assault weapons ban expired in Sept 2004.
Is that the cause of inflection point? Does the database have information on the weapons used?
Ahh… I missed the link to the db. I answered my own question.
short answer … No.
whike the Federal assault weapon ban expired in 2004, the weapon that appears most frequently is “semiautomatic handgun“ … at least until 2016. Since then “semiautomatic rifle” appears most frequently.
“I Don’t Like Mondays” (Boomtown Rats) was a #1 single in 1979.
Now plot mass shooting fatalities per year that didn’t happen because someone with a gun put a stop to it before it started.
(For the record, the peak year of 2017 equates to about 0.27 deaths per day. That same year New York alone averaged more than 200 abortions per day. What is Demagogue Joe doing about that? Also, if you adjust the maximum of your y-axis to 75,000 – what appears to be approximately 10% of the number of acceptable U.S. child murders in the womb in recent years – none of this is statistically relevant.)
Exactly.
Here is a 2018 study of mass shootings worldwide. The US ranks 64th by mass attacks per 100k and 65 by people killed per 100k. Switzerland, Norway, France, and Finland rank higher.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3671740
Well, THAT’s not very helpful is it.
A fantastic finding!
I think that it is helpful, but it creates a potential data problem.
The sort of “mass shooting” that we’re concerned about, though they are very rare, involves precisely the sort of situation that the Mother Jones folks seem to be trying to capture.
Yes, but it also hides the gang related violence which, if included I suspect, would dwarf these types of events. I guess it would be better to say, that we need both data sets because they are totally different issues, but I wonder if the average person actually knows how much gang violence is in the US, and how many firearms related homicides are related to other crime, and gangs in specific.
Homicides are distributed extremely narrowly in this country. I remember once reading that the city of Detroit had more homicides than the entire rest of the state of Michigan. We have whole States where the overall homicide rate is lower than Canada and most of Europe. And then we have just a few neighborhoods in Chicago that account for a couple percent of the entire nation’s homicides.
IIRC, the most dangerous place to live in the US is a set of neighborhoods in St Louis. There was a great article a while back that did a breakdown of homicides by zip code, or it might have been neighborhood. You can see that you can travel a block or two, less than a mile, and go from no murders in a year to over a dozen.
One of the common responses to the complaint that the murder rate in the United States is much higher than that of most other countries is to point out that if you remove about five metropolitan areas of the United States (I don’t remember exactly, but it’s something like Washington, D.C., New York City, Baltimore, Chicago, and Detroit), the U.S. murder rate is as low or lower than most other countries. [If you do a racial correlation on murders relative to the racial makeup of the population, you also learn that outside of one racial group, murder rates among different countries aren’t all that different. But political correctness prevents further study of that angle.]
There are murders, and then there are ‘my people’ murders, which we take extra personally – and how we feel about the killers is also affected.
Most people see children as ‘my people’, and there is often a race to ID the killer as ‘their people’. Which my people consistently lose, btw.
The ‘my people’ aspect renders stats meaningless. Bullets are like bad weather; in a nation as big and populous as ours, somebody, somewhere is going to get rained on. Though we complain about the murder rate in general we tacitly accept it.
But when the dead are hand-picked off based on an identity we also have, we see it as an existential threat no matter how negligible the numbers may be statistically.
Which is a long way of saying that freaking out and passing laws like hate crimes and such is something we should avoid.
Even when it’s our people.
It would be useful to compare this data to the number of guns and guns per capita. It likely would show (though I don’t know for a fact) little or no correlation. Not that facts matter, but if it does, that would pretty well prove the assertion that guns do not cause violence.