Last updated: 2018-03-05 10:00
tl;dr: Google searches for “gun control” after the shooting in Parkland, Florida seem to be similar to search trends after other shootings, although it could be different this time.
A civilized, modern society should be able to function nearly entirely without guns. We should be able to go about our lives without ever thinking about guns. We should be able to go the school, to go to church, to go to the grocery store, to leave our house and move about in the world without ever thinking about guns. Without ever having to worry about a gun, by accident or intention, ending your life or the lives of your loved ones.
We face a public health crisis unique to our country alone with respect to gun violence. We are inflicting pain and economic burden on ourselves at a yearly cost of $229 billion – equivalent to the size of Portgual’s economy. It is pointless, devastating, entirely unnecessary and utterly heart breaking.
I very clearly remember Columbine as an inflection point in education in my childhood, bomb threats and fire drills replacing class time like DVDs replacing VHS. Columbine happened 1,700 miles away from my middle school, but we felt it, even within the bubble of privelege that surrounded my community. I cannot even begin to fathom how today’s children feel the gun violence they see around them.
It absolutely breaks my heart that they have to go through this. It breaks my heart that the “adults” in our society can’t get our shit together. It breaks my heart that they have to fight so hard to be heard and protected, that we need them to keep fighting, that we need them to stand up and speak out when we should be comforting them, when we should have done something sooner.
In their darkest hours, we are asking too much of them. And yet we need them. How else will this ever change?
Yesterday, Nate Silver put words to our collective hope that this time something is different, that we’re getting closer to #NeverAgain.
My first thought was how do you know?. What does the “fade” look like? Clearly, I wasn’t alone as a lot of other people pointed this out in replies.
I did a quick search and found the gtrendsR package on Github. (Thank you Philippe Massicotte and Dirk Eddelbuettel.) What follows is an analysis of Google Search trends for the term “gun control” before and after other mass shooting events.
Source Code
A quick note. All of the source code for this post can be found in my Github repo for this website. I didn’t include it here in part because it’s a mess and also in part because I used caching and some unusual data types to store the Google Trends results.
Mass Shootings Database
A dataset on mass shootings in the U.S. is available thanks to the hard work of Kaggle user zusmani. They compiled a list of mass shootings in the U.S. since 1966, using public datasets and webpages including Wikipedia, Mother Jones, Stanford, and USA Today.
The data is available on Kaggle at https://www.kaggle.com/zusmani/us-mass-shootings-last-50-years/ and includes a variety of variables.
For historical comparison, I selected the 15 most deadly shootings since 2004 (earliest available dates in Google Trends). Note that total victims include casualties and injuries.
| Date | Title | Location | Victims | AR-15 (or sim.) |
Semi-Automatic |
|---|---|---|---|---|---|
| 2017-10-01 | Las Vegas Strip mass shooting | Las Vegas, NV | 585 | Yes | Yes |
| 2016-06-12 | Orlando nightclub massacre | Orlando, Florida | 102 | No | Yes |
| 2012-07-20 | Aurora theater shooting | Aurora, Colorado | 82 | Yes | Yes |
| 2007-04-16 | Virginia Tech massacre | Blacksburg, Virginia | 55 | No | Yes |
| 2017-11-05 | Texas church mass shooting | Sutherland Springs, TX | 46 | Yes | Yes |
| 2009-11-05 | Fort Hood Army Base | Fort Hood, Texas | 45 | No | Yes |
| 2015-12-02 | San Bernardino, California | San Bernardino, California | 35 | No | Yes |
| 2012-12-14 | Sandy Hook Elementary School | Newtown, Connecticut | 29 | Yes | Yes |
| 2008-02-14 | Northern Illinois University shooting | DeKalb, Illinois | 26 | No | Yes |
| 2014-05-23 | Isla Vista mass murder | Santa Barbara, California | 19 | No | Yes |
| 2014-04-02 | Fort Hood | Killeen, Texas | 19 | No | No |
| 2011-01-08 | Tucson shooting | Tucson, Arizona | 19 | No | No |
| 2009-04-03 | Binghamton shootings | Binghamton, New York | 18 | No | No |
| 2016-02-25 | Excel Industries mass shooting | Hesston, Kansas | 17 | No | Yes |
| 2016-07-07 | Dallas police shooting | Dallas, Texas | 16 | No | Yes |
Query Google Trends from R with gtrendsR
Then I queried Google Trends via gtrendsR for the search terms "gun control" and a reference term for
- The last 7 days (capturing Parkland, FL shooting)
- One week before and 3 weeks after each of the above events
- All search interest since 2004.
There are two points to note when working with data from Google Trends.
First, Google Trends returns search trends relative to the time period queried, so it can be hard to determine the scale of interest when looking at a smaller time period, such as a month. For comparison, I included the reference term "baseline" because this is relatively stable at ~10% from 2004 through present day1. Edit 2018-02-20: I replaced "difficult" with "baseline" as the comparison search term.
Second, the data returned by Google Trends is aggregated according to the time window specified. In my tinkering I noticed that hourly trends are only available for the last 7 days. Earlier data is always aggregated at least by day. Longterm trends (such as the all time trend) are aggregated by month.
The code below isn’t exactly the code I used to gather the data, but it should give you an idea for how easy it is to use gtrendsR.
library(dplyr)
library(purrr)
library(gtrendsR)
get_trends_for_event <- function(day) {
gtrends(
keyword = c("gun control", "baseline"),
geo = "US",
time = paste(glue::glue("{day - 7} {day + 21}"))
)
}
guntrends <- selected_dates %>%
mutate(gtrends = map(Date, get_trends_for_event))
all_guntrends <- gtrends(c("gun control", "baseline"),
geo = "US", time = "all")
parkland_trends <- gtrends(c("gun control", "baseline"),
geo = "US", time = "now 7-d")All time search interest in “gun control”
Google searches for “gun control” are relatively low-volume in the United States, with a clear increase around any mass shooting event. The plot below shows search interest in “gun control” since 2004, with the largest peak occurring after the Sandy Hook Elementary School mass shooting.
Searching for “gun control” after a mass shooting
To examine Google search trends around individual mass shooting events, I looked at searches for “gun control” within a one month window around the shooting.
Each subplot shows search trends as reported by Google Trends on a “scale” from 0 to 100. The scale of searching is relative to the time period queried. Events such as the Las Vegas Strip, Orlando Night Club, Aurora Theater shootings show how the events caught the attention of the public and led to questions about our gun laws. Other events, such as the Fort Hood Army Base attack seemed not to have caught public or media attention in the same way.
The above plot highlights search trends around individual mass shooting events, but it’s difficult to assess how strong the response to a particular event is relative to previous events.
The approach that I worked out is to use a baseline search term that is relatively constant across the entire period. As I discussed above, I picked "baseline" for this role, as it nicely fits the bill. For each query of Google Trends, I also requested search trends for "baseline". Then I scaled the "gun control" search trends relative to the hits on the baseline search.
What this means is that in the following plot, the search volume for "gun control" is shown in “units” of searches for "baseline". A value of 10 means there are 10 times more search for "gun control", whereas a value of 1 means that the search volumen is roughly equivalent.
Is this time different?
I certainly hope so. If we isolate other events with a similar profile in terms of public perception, it seems possible that this time will be different. But it also seems too early to tell.
In the mean time, just remember that Google searches aren’t real things. If you want to make a difference or a change, support a local politician, donate to gun violence research, and register your friends to vote. For more ideas, take a look at Everytown’s Action Plan at https://everytown.org/throwthemout/.
I’ll close with this tweet from @davepell.
But first, to the victims in Parkland, Florida and the victims of gun violence across the United States, I want you to know that I hear you, and I stand with you.
Edits
2018-02-20: I changed the comparison term from "difficult" to "baseline". Also updated the scripts to gather data from the week prior to Feb 14, 2018.
Footnotes
It turns out that it’s actually rather tough to find a “stable” search term within the same general order of magnitude of searches as “gun control”. The term
"difficult"was not selected un-ironically (initally). Later I decided there was too much variation in the search trends for"difficult"and thought that following the irony was a good idea. So I changed the term to"baseline". Anyway, try it for yourself – go to Google Trends and see if you can find another non-seasonal, nearly zero-growth search term.↩︎