New Research in the journal Econ Journal Watch: “Is the United States an Outlier in Public Mass Shootings?”

Apr 9, 2019 | Featured

Dr. John Lott and our research director Professor Carl Moody have piece on what we believe has been one of the bigger academic frauds. The paper on an international comparison of mass public shootings can be downloaded here.


In 2016 Adam Lankford published a widely propagated article purporting to show that during a 47-year period the United States represented 31 percent of worldwide public mass shooters, and claiming that the outsized U.S. percentage is a result of gun prevalence. We examined the data from 1998 to 2012 and found that, although the U.S. has 4.5 percent of the world’s population, the U.S. represents less than three percent of worldwide mass shooting incidents or mass shooting deaths and less than one percent of mass shooters. What explains this incredible difference? While Lankford claims he is using the conventional definitions of public mass shootings from the FBI and NYPD, it turns out that Lankford’s 31-percent claim is an artifact of his having stripped out much from conventional definitions of ‘public mass shooter,’ notably excluding almost all incidents of terrorism outside the U.S. and most of the cases where more than one shooter is involved. We compare U.S. and non-U.S. public mass shootings using the official definitions from the FBI and NYPD that Lankford claims he is using. We also suggest a reasonable explanation for the fact that, while the U.S. has a relatively large number of lone-wolf shooters compared to the rest of the world, it does not have a relatively large number of public mass shooters.

Since we published our piece we learned a number of points.

1) In the last week, we now have learned that Lankford claims that he limited his list of mass public shootings to cases where there was only one mass public shooter. Nowhere in Adam Lankford’s original paper nor any subsequent discussions in the media or elsewhere does Lankford mention limiting attacks to only one shooter (a copy of his original paper is available here and his brand new paper is here). Despite his claims to the contrary, given this admission, his definition of mass public shootings doesn’t fit the FBI or NYPD definitions. Nowhere in their discussions do either of these two organizations limit their mass public shooting cases to just one shooter.


2) After years of refusing to release his list of mass public shootings, one nice thing of our paper being published is that Lankford has FINALLY made public what he claims was his list of cases.(errors in his data are shown here). Providing his data set shows that he doesn’t limit the cases consistently. Lankford includes the Columbine case where two people were shooting people (it would have been hard for him to exclude this as it was mentioned on the first page of his paper and everyone would presumably think that this is archetype case), though that is the only case involving to shooters that he includes for the U.S.. For example, Lankford misses the Jonesboro, Arkansas shooting in March 1998. There is no explanation for why. He also has one observation from Russia with two shooters, but he excludes 40 other foreign mass public shootings that fit the FBI definition with 2 shooters. (There were 34 cases with 2 shooters if you exclude any attacks that were part of the Mumbai attacks, though Pakistan other groups claim that this wasn’t the “sponsored” attack Lankford asserts that it was.)


So he talks about only having one shooter, but has to include Columbine because he mentioned it in his original paper. He also includes a case with two shooters from Russia. There is no discussion ever why he includes cases with one or two shooters but excludes examples with three or four shooters. There are a large number of cases with three shooters. Why shouldn’t they be included?


The NYPD dataset that he says that he followed had shootings with up to 10 shooters (more on this later), we found at least 3,081 shooters in the rest of the world from 1998 to 2012. The US then just made up 1.45% of all mass public shooters.


Lankford even misses 37 foreign mass public shootings involving just one shooter, and despite us providing an entire list of our cases, Lankford does not discuss why these cases these single shooter cases should have been excluded.


He inflates the number of US cases during the 1998-2012 period by including 10 cases that don’t meet the FBI or NYPD definitions of mass public shootings. He includes cases that involve another crime, such as a robbery or occur in a non-public place (such as a residence) or fewer than four killed in a single attack.


3) Lankford writes in his reply: “and ‘complete data’ (192) could be rewritten as ‘complete data on independent variables.’” He wants to argue that he wasn’t saying he had complete data for mass public shootings. But the entire sentence referencing “complete data” read: “Complete data were available for 171 countries, and they averaged 1.7 public mass shooters per country from 1966 to 2012.” So the very sentencing saying he had complete data was pointing to data on mass public shooters.


4) The data set that he claims is his now doesn’t come close to fitting statements in his original paper. He says the NYPD “dataset may be nearly comprehensive in its coverage of recent decades, it may be missing some older cases.” Given that the NYPD found only 16 attacks with 27 killers outside the United States over 1998 to 2012 we examined, it seems very clear that Lankford list in his original paper comes nowhere near to being complete. If you exclude cases where more than one shooter was involved in the NYPD set of foreign cases during this period, the NYPD dataset has only 15 killers. So how exactly is that “nearly comprehensive” for the dataset of 98 mass public shootings that he has now made public?


What is particularly puzzling is that while he talks about the NYPD dataset “missing some older cases,” he never mentions that the dataset might include cases that it shouldn’t include, such as one with 10 shooters. Yet, Lankford now claims for the first time that his dataset was limited to cases with just one shooter.


For our data, excluding all the cases with more than 10 killers, as the NYPD data do, we found at least 3,081 shooters outside the US, with the US thus making up less than 1.43% of all shooters. A “may be nearly comprehensive” 27 killers versus 3,081 isn’t even close. Even if his “may be nearly comprehensive” list is off by a factor of 10, they aren’t close.


5) In conclusion, Lankford doesn’t use the FBI and NYPD definitions that he continually claimed that he used. He made up his own definition and didn’t it mention it anywhere in his paper. We think that the New York Times, Washington Post, and USA Today that frequently made use of Lankford’s claims would have been a little reticent to use his results if they knew that he had made up his definition and that definition would technically exclude Columbine (though has noted he doesn’t follow this definition in that case).


Finally, it must be made clear that Lankford still hasn’t made all his data available. He hasn’t made his data available for the regressions that he ran for his original paper. We have reached out to him about this data, but obtaining it would help provide us with answers to remaining questions.


It is obvious why Lankford refused to let anyone see his list of mass public shootings for years and why he wouldn’t even let us see it until after we had published our piece. Even with his unique definition of mass public shootings, he doesn’t follow it consistently, and it has many errors in it.

More detailed information will be available later.



  1. Tom Campbell

    Lankford needs to be fired.

    His specious papers are as bad as the Lancet paper falsely linking vaccines with autism. The policy outcomes from using this incredibly bad non-information are incalculable, given anti-gun progressivist proclivities and refusal to acknowledge real information about crimes and guns used in them.

    • Doug R

      Why you chose to note a paper linking vaccines with autism in your reply is only a sign of your ignorance on that matter and has nothing to do with this report.

  2. Eric L Muldrow

    Excellent breakdown of the problem with his work. I will definitely share this.

  3. David

    And yet the the LapDog mainstream media propagated this rubbish research, no questions asked. And it is published !
    Neither media bothers to even check the data.
    Go U of Alabama ! Great hire !

  4. MIke Hughes

    Isn’t it normal practice to show the underlying data in a study? Absent any confidentiality issues (since its public data I see none), the dataset should be provided by any study worth its salt.

    Personally I am curious because how many ‘mass shootings’ where not classified as such because of direct citizen intervention (concealed carry holder). Further, what was the one shot effectiveness irrespective of accuracy.

  5. Huapakechi

    It seems langford used the computer modeling strategy that the global warming crowd uses.

  6. thinkr2

    As a physicist, I would be excoriated if I published a paper claiming to have found a way to exceed the speed of light without including all data and methods so that other physicists could replicate my work.

    “Climate Science” and apparently all “Social Science” are seemingly exempt from this standard scientific practice.

    Furthermore, hard sciences almost always look for R-squared correlations of 80% or greater, but social science seems happy with values of 50%, which would be laughable in physics or chemistry.