Exploration into two decades of school shootings in the United States

A look into the strange American cultural phenomenon. This report is based on the Washington Post data set and data analysis on the school-shootings-analysis repository.

Introduction

This reports aims to leap into the trends and patterns behind school shootings occurring in the United States of America between April 1999 and January 2023. The main foci of this report will be the physical patterns of the events - including time periods assessment as well as the geospatial distribution of the phenomenon. Additionally, the report will review the cultural response to the happenings.

About the data set

The supporting data set for the analysis is the Washington Post school shooting data. This source was hand made by the News outlet by “attempting to identify every act of gunfire at a primary or secondary school during school hours”. These include primarily events of accidental discharges, unrelated shootings occurring on school properties and what started it all - school shootings.

The Washington Post used this data to approximate the number of students who have been impacted directly or indirectly by school shootings in their article “More than 349,000 students have experienced gun violence at school since Columbine”. The main conclusions from the Post’s research are

A final note pertaining to the source of data. The Federal government of the United States does not maintain an official source on school shooting occurrences. This has resulted in independent organizations creating data sets with separate methodologies, aims and, inevitably, biases. There exist other sources such as the K-12 School Shooting Database or the Mother Jones US Mass Shootings. The differences are not compared in this report. The Washington Post data set was selected for its transparency of methodology and accessibility.

Research questions

More specifically, the questions in the mind of the authors are the following

  1. What is behind the average of a school shooting occurring every 24 days?
    • A look into when the bulk of shootings have occurred.
  2. When should children consider skipping school?
    • A look into the trends for shootings, especially high risk days of the year and hours of the day.
  3. At which point has the shock value diminished?
    • A look into the cultural response and news trends.
  4. How effective are counter strategies?
    • A look into the strategies employed to address the issue and what the result is.

Methodology

The source code for the calculations and analysis done for the report is available at the public repository school-shootings-analysis on GitHub.

An overview

This section provides an overview of the observations of the data set to provide context for further analysis.

Set statistics

The data set entails 366 shootings over the time period of April 20th 1999 until February 2nd 2023. With a period of 8689 days, there was a shooting occurring on school grounds every 23.74 days.

There have been 603 casualties, of which 192 were deaths. On average, around two were injured and one person was killed during every two school shootings. The highest amount of deaths occurred during the Sandy Hook shooting (December 14th, 2012) with 26 deaths.

States affected

An initial step to understanding the geospatial distribution of the phenomenon is to look at the states with the most occurrences of shootings. Additionally, it is important to observe the amount of total deaths as a result of these shootings, available as an separate layer on the map below. States of interest have been marked with a bolded outline (these are California, Texas, Florida, Colorado, North Carolina and Connecticut).

Initially, it appears that the core of the issue lies in California with the highest number of occurrences over the two decades at 40. With 38 individuals harmed by the shootings there certainly is a problem in this state, yet a relatively small death toll of 7 brings California to smaller side of the spectrum for state deaths.

The states with the highest death counts are Connecticut and Texas, homes of Sandy Hook Elementary and Robb Elementary. Florida (Stoneman Douglas) and Colorado (Columbine) are second to follow. All of these states have a death toll of at least 15 and up to 32. States such as California and North Carolina may have a large amount of shootings, yet a small amount of deaths and (in the case of North Carolina) injuries.

The takeaway is that most shootings are not lethal and more commonly result in injuries, while mass shootings claim lives in extremely high numbers. There are some states which are defined by the mass shootings which occur in their schools, bringing them to the forefront of the school shooting problem.

Shootings on the rise

A concerning observation made by the authors of the data set, corroborated as well by similar data set maintainers, is that the trend of school shootings is increasing annually.

Within the year 2018, the number of school shootings has increased twofold compared to the previous rolling average, a trend which is followed in the next years. It is unclear whether this is due to an increase in recording of events - where previous “minor” events such as accidental discharges were never reported beyond a closed internal news environment, or if there is a societal and psychological reason for the drastic rise in school shootings.

It is important to note that the year 2023 does not bode well. Just one month of data (January 1st - February 2nd) already brings the 2023 shootings count to 7, exactly as many as the entire year of 2015.

The covid lockdown effect

The year 2020, infamous for the start of the pandemic and the introduction of remote schooling has observed a proportional decrease in shootings, yet the count of shootings is equal to that of the average shootings in the 2010s, a time of only in-person schooling. While the adoption of remote schooling was very successful at keeping children safe not only from the virus, but even more so from gun violence, the number of recorded shootings has not dropped to any level below previous trends.

Upon closer inspection, during the remote schooling period (March 2020 - February 2021) there were “only” four shootings, all of which were deemed accidental with just one injured casualty. March 2021 saw the reopening of many schools and with it five shootings occurred in March alone, three of which were deemed as targeted.

Time trends

To attempt to understand which periods of time pose the greatest risk to school students, a look into the trends of hours, weekdays and months are made.

Hourly

The first aggregation to understand is the pattern behind shooting frequency is by hour. This provides a glance into understanding which parts of the day are most potentially dangerous to students.

A note is made on the poor quality of recording on the hour of the shooting with varying time formats used and 20 empty values.

School shootings do not occur outside of school hours (7am - 5pm), with the one exception of a recorded school shooting at 11pm which given the poor quality of time recording may be a mistake (11am).

Though the time of a shootings appears to be more or less evenly distributed over the hours of the day, there is an apparent separation of high risk time periods - the first in the morning, as lessons are starting (8am), and the second, more potent, is during lunch hours (12pm) and thereafter.

Weekdays

The same analysis is continued for weekdays aggregation.

It appears that Tuesdays were most commonly plagued with shootings, with a stronger drop towards the end of the week (Thursday, Friday).

Unsurprisingly, no shootings occurred on weekends.

Months

Finally, months of the year are observed for patterns.

The summer period observes less violence than the other parts of the year, but it is important to note that is due to the relative number of students being much lower (while summer schools are open).

A disheartening observation is the twofold increase in shootings from December to January, which is also when the most shootings occurred in the entire year (at 51).

In other parts of the year, the events are more or less evenly distributed around a trend of 40 per month.

Take-away

While some of the observations seem obvious, while others troubling, it is important to keep in mind that indiscriminate shooters will likely want to maximize damage, thus choosing more effective periods of time to strike (eg. lunch period).

The Christmas period at the end of December may be an especially difficult period of time for individuals from broken homes and those struggling with mental illnesses alone, which might be a fuse for the increased activity at the beginning of the new year.

While these observations may provide insight into when increased care should be taken for risk of a shooting occurring, a criminal with intent will likely adjust their actions to account for these measures.

Shootings, not statistics

While discussing these shooting events it is easy to abstract the observations to simple numbers, comfortably ignoring the realization that these are factual events - of pain, of anguish and of death.

To hopefully illustrate this, this section will provide information more information on individual shootings, as well as the victims and perpetrators.

The magnitude of the issue

The previously discussed graph on shooting occurrences is expanded to show casualties per year.

The four infamous mass shooting death counts are clearly visible in the years 1999, 2012, 2018 and 2022. Disregarding the mass shootings, the death count appears to be maintained as a constant ratio of all casualties.

Details on all of the individual shootings to displayed on the map below. The size of each circle indicates the relative number of casualties at the occurrence. More information about the school and shooting can be identified by selecting a specific occurrence.

Who are the shooters

This small subsection is devoted to looking at the range of the individuals who have been identified as shooters in the data set. It is worth remembering, not all of them are criminally intent. There were 45 such cases, with the shooting defined as accidental (an example of this is a loaded gun firing in a 12 year old’s back pack, injuring five - “Sal Castro Middle School Shooting: 5 Fast Facts You Need to Know”, heavy).

The image of the typical school shooter is strongly influenced by the four mass shootings - young, white teenagers who are (or very recently have been) students at the school. This notion is somewhat reinforced by the data set.

There appears to be a normal distribution of age around the mean of 15. The vast majority of shooters are male with a majority of white ethnicity.

Where are they now

Out of all the incidents, 33 are confirmed suicides. Such was the fate of the Columbine shooter, after they had taken 13 lives themselves. Another example is the shooter behind the Jacksonville Episcopal School, a former teacher who took the life of the school director and his own, the same day their work was terminated (“Head of Jacksonville Episcopal School killed in murder-suicide”, Episcopal News Service).

Others are churned through the justice system, such as the Stoneman Douglas High School shooter, who was provided a sentence four years after the shooting (“At an emotional hearing, the Parkland shooter is formally sentenced to life in prison”).

Let’s talk about guns

An important question, when discussing shootings, is about the source of the guns which fuel this violence.

The below analysis is done with the additional use of the State-Level Estimates of Household Firearm Ownership database.

Gun ownership correlation

A simple conclusion is that states with more gun-owning households is more prone to gun violence in the form of school shootings. The choropleth below displays gun ownership vs casualties from shootings to follow this train of thought.

What comes as quite a surprise is that the states with the highest percentage of households with guns have some of (if not the most) lowest consequences of shootings. For example, Montana, Wyoming and Idaho - three states with an ownership ratio above 60% (and up to 66%) - not only do not have any casualties from shootings - these states have had no recorded shootings in this data set!

Stating that there is no correlation between ownership and shootings is also another simple conclusion. Ownership ratios do not reflect the true number of firearms in the state as well as the proportion of guns to people.

From a previous section, California was established as the state with the highest number of shootings. California has a population of 21.5 million (2021 census), whereas Wyoming has a population of 0.58 million (2021 census). This is an enormous difference which is easy to overlook and indicates that interpretations should be considered carefully.

California has 13.2 million households and with firearm ownership rates of 28.32%, that tallies to around 3.7 million households with firearms. In Wyoming, 0.18 million out of 0.27 million households have firearms (66.22%). It is true that a household in Wyoming is more than twice as likely to own or have access to firearms, but there are 20 times more such households in California.

The above doesn’t take into consideration unregistered firearms, illegally purchased weapons or “ghost guns” (lacking serial numbers).

Firearm source

In the data set there are 105 records of source of the weapon used by the shooter. While there are many singular, specific descriptions such as “constructed from parts purchased by the shooter” or “ghost gun”, most can be categorized as obtained (via theft, gift or lending) from a relative at 49 instances. The majority of those were obtainment from the perpetrators father at 16 instances. In either case, these count as household sources, implying that there is a pattern in shooters using firearms they have access to in their homes.

Cultural response

It is evident through means of common media, the news cycle and word of mouth that the collective reaction to school shootings has changed over time. Today, it is not uncommon this subject area is treated with a dose of dark humour. This does not come as a surprise, as it is simply human nature to devalue a repetitive stimulus - observed in this case as the diminishing of the shock value of the deaths of children in “safe” schools.

An example of this behaviour is the frequency of Google searches for the keywords “school shooting” over the past years. While Google does not provide exact search numbers, the relative proportions allow a reasonable conclusion on the popularity of the search term.

The four largest-impact school shootings to date are

While the Columbine shooting is out of scope for Google trends, it provides additional context for the other three events.

A reasonable first impression is the sudden rise of searches around the date of each occurrence with upticks around dates of anniversaries for previous events.

Yet, a more concerning observation is the drop in relative interest for each next shooting. Understandably, Sandy Hook was a massive shock worldwide as one of the first mass school shootings of the decade following a relatively peaceful time period after the Columbine shooting (not depicted). The next occurrence, Marjory High, noticed a drop in Google search popularity, despite a larger number of casualties (although less deaths). The popularity for the search term after the Robb Elementary event, four years later, would notice that same drop twofold.

This could be explained by a more mature news cycle forming around the school shooting phenomenon, where more platforms are either willing or have the tools to provide the news, thereby removing the need to manually search for information. A news machine like the one assumed here requires the subject area to be just as evolved, thus providing the argument for its cyclical existence - the more popular something becomes, the more standardized and normalized it becomes.

The romanticization of the antagonist

The danger of the normalization of this phenomenon and the wide-spread increase of tolerance provide a feedback loop pushing mentally unstable or disturbed individuals to pursue this path as a viable means to achieve revenge on bully culture or simply a twisted sense of greatness or strength. The investigation of school shooting reasons is a subject for another report, but it would be out of place not to mention the contagion effect and an example of a perpetrator’s perspective.

The criminal behind the Douglas High School shooting (2018), was found to have recorded and shared his thought process months before the event. From CBS Miami’s newsreel [“Chilling Cell Phone Videos Made By Released"](https://youtu.be/1UJ5-rSb3o0) a few key quotes - `"When you see me on the news you will all know who I am"` and `"My name is (...) and I am going to be the next school shooter of 2018"`.

School shooting in media

Although not the focus of this report, a few examples of school shootings in media are provided to depict the prevalence of this cultural phenomenon.

Counter strategies

The American response to address the issue has been predominantly reactive, spawning a market around the concept of campus safety. An industry of “fire-fight-fire”, pushing to make metal detectors, bulletproof doors and armed security normal at schools.

The Washington Post article “Armored school doors, bulletproof whiteboards and secret snipers” does a great job of detailing the absurdities, which among others detail a subscription-based, parent-sponsored militaristic infiltration programme based on the premise that an ex-special ops unit goes undercover to identify, befriend and neutralize potential shooters before they strike.

The introduction of resource officers also begs the question as to whether that is the appropriate response. Not only have they been shown to fail to prevent a shooting, they themselves can contribute to the problem. In the data set, out of seven cases where the shooter was established to be the resource officer, one was lethal - the officer responded to a student threatening him with a baseball bat by ending his life with a pistol (“Officer kills Fresno student”, Los Angeles Times). If not a failure in de-escalation, it is certainly a failure in an appropriate response.

Final thoughts

There is no denying that school shootings are a problem in America, though it is unclear if their magnitude has only increased due to the spotlight, pushing relatively low-impact shootings to be included in the discourse, or if the issue is now out of control.

The latter appears to be more accurate with near inaction from the government, the creation of a profitable industry focused on reactive measures, the cyclical fatum induced by Pandora’s Box with shooters inspired by their predecessors and, perhaps the most damning, the commonization of the phenomenon leading to its devaluation and a drop in its importance.

It appears that the shootings have become a tolerated, lesser evil, that which is more acceptable than the alternative immediate approach of strict gun control. A decisive action which may prove to be largely unpopular and so unattractive in a short-term government cycle based on popularity. Concerningly, addressing the root cause of the shootings - which could be mental insecurity or proper violence education - is left to nonprofit organizations whose commitment may waver to the whim of the wind.

The keen reader may have noticed that the shooters have not been named. This is not a story about them. Their deeds, however heinous, do not validate their claim to fame and a twisted sense of glory. This is a story of the American system failing their youth. And this failure is now forever etched into history by data as true as the graves of the victims.