Mass Shootings and the Trump Effect (Part II)

Casualties (killed, wounded) from mass shootings are not continuously distributed; this suggests an alternative approach — given the high variance (shown in Figure 1 below for a subsample of the data) — I estimate a negative binomial regression (quasi-maximum likelihood).

Figure 1: Mass shooting fatalities (dark red), and wounded (light red). Orange denotes Trump administration. Source: Mother Jones accessed 8/5/2019, author’s calculations.

Estimate over 1982M08-2019M08 (through 8/5/2019 for August):

casualtiest = -2.60 – 0.64 bant + 1.54 trumpt + 0.014 time

Adj-R2 = 0.048, SER = 29.95, NOBS = 445. Bold denotes significance at 10% msl, standard errors corrected for overdispersion (GLM option in EViews).

In words, each year under a Trump administration, 18 154 percent more people are injured or killed in mass shooting events than would otherwise occur. As discussed in this post, a year during the Trump administration is associated with 8.64 over 5 more events/year than would otherwise occur. [correction to interpretation 8/7 8:15am Pacific; h/t Rick Stryker]



58 thoughts on “Mass Shootings and the Trump Effect (Part II)

  1. dilbert dogbert

    As trump has been spewing hate speech for longer than he has been in the oval office, maybe the orange overlay should be extended? Certainly that speech has been widely publicized in major news outlets and therefore has had wide impact on fragile personalities.

    1. Moses Herzog

      @ dilbert dogbert
      Believe it or not this same thought also entered my head, while I was reading Menzie’s post just a few minutes ago. Which is actually kinda weird considering how many of these posts on this topic Menzie has done. Unless it was a “passing thought” and I didn’t take note of it?? But it’s a very fair point, it’s kinda like the reverse of economic lag on policy decisions. Dark rhetoric can appeal to people’s emotions very quickly—and donald trump was already using race, gender, fear of economic exclusion, and other issues at his rallies long before November 2016, so arguably the effect of his rhetoric could go back to pre-inauguration months. I do think Menzie is making the right choice of January 2017 as his demarcation point—because these things should always be measured on the conservative side in my opinion—but still in the back of our minds we should also be aware of how the campaign rhetoric could have very well impacted the pre-inauguration rate of violent acts.

      Everyone has very dark thoughts inside them (that is my personal opinion anyway, and I’m not afraid to say that). Sometimes when I see movie characters commit horrendous acts, even murder….. I can find myself “reveling in” those acts inside the small moment, in a weird, almost vicarious way. It is precisely because I know I will probably never do these things to people I have thought about or envisioned doing them to, that makes me enjoy watching a film character do them. We’re not supposed to say this, but I think the very vast majority of people have these same thoughts. That is why people such as donald trump are so extremely extremely dangerous. They remind us of these segments of ourselves that might have these inclinations. Which is why people drawn into trump’s world or “dark sanctum” (who are not very moralistic people to begin with) become even more unscrupulous.

  2. Rick Stryker


    I think you are misinterpreting these regressions. In the previous poisson regression, for example, you assert that a 0.7 regression coefficient on the Trump dummy means that there are 0.7 more mass shooting events on average per month while Trump is president. Not so if you are using the standard Poisson regression model in which the conditional mean is specified as E(y|x) =exp(x’B), since dE(y|x)/dx(i) = exp(x’B)B(i), which implies that the effect of a change in the x(i) regressor on E(y|x) is the percent change in the conditional expectation of y given by (100B(i))dx(i). That’s true for small changes in x(i). So, the Poisson regression coefficient means that the number of mass shootings is on the order of 70% higher in months when Trump is president, conditional on the rest of the data. (Used derivative, so haven’t adjusted for discrete change in the dummy from 0 to1 but the point should be clear.) Similar comment for the interpretation of the negative binomial model in this post.

    But so what? If you look at the details behind these mass shootings they have nothing to do with Trump. The regressions just codify what is obvious from looking at the data, which is that there have been more mass shootings in the last few years than previously. But you”ll need more than these trivial regressions to convince anyone that Trump has anything to do with them.

    1. Menzie Chinn Post author

      Rick Stryker: Thank you very much for the correction; I’ve amended the text accordingly in both posts. I understand the point that no causal relation is “proved”. If you read the text carefully (as you have), then you see I did not write that Trump “caused” the higher fatalities and casualties — merely that the Trump administration is “associated” with elevated numbers of mass shootings and mass shooting casualties. Unlikely to be a coincidence (that is what statistical inference is for, yes?), but could be common factor (people tending to want to shoot en masse other people became more prevalent *and* voted in Trump), or reverse causality (more people wanting to shoot up people en masse caused the election of Trump), or the other — Trump’s election and rhetoric caused elevated shooting events and associated casualties.

      1. baffling

        “But so what? ”
        that would be an appropriate answer for somebody who washes the blood off their hands as they grin, knowing they have continued to promote a world where more gun violence and bloodshed can occur. just like moscow mitch, dick stryker enjoys obstructing any changes to the violence he projects into the world. scum of the world unite.

        1. Rick Stryker


          Obviously the “so what” was directed at the regression, not at the mass murders.

          1. baffling

            dick stryker, people have told me that your “so what” response was directed towards the murder of immigrants from mexico. such commentary is pretty sad and pathetic. i would have though your were a better man than that. many people are upset at your position.

      2. Rick Stryker


        I don’t find the regressions persuasive in showing that the recent acceleration in mass shootings is somehow associated with Trump, for a couple of reasons. The first is that if you look into the details of the mass shootings, it’s very hard to see how they could be associated with Trump or anyone else. Mass shootings tend to fall into distinct categories: psychos who fantasize about killing large numbers of people and then sometimes do it; workplace violence; and situations that get out of hand. These cases all seem to have their individual stories, although psycho cases are correlated in the sense that the psycho mass shooters tend to study and imitate each other. The el paso mass shooting is a bit different in that there was a political motivation behind it but it was the same in the sense that the el paso murderer was imitating and influenced by the Christ Church shooter. The Ohio shooter was strongly left wing in his political views but the shooting does not appear to be politically motivated. Looks like the more standard psycho shooting.

        The second reason is that you haven’t controlled for enough variables in the regressions that might explain mass shootings. You just have 2 dummies explaining mass shootings. As you well know, failure to include other relevant explanatory variables that are correlated with the dummies will lead to biased and inconsistent estimates of the coefficients associated with the dummy variables.

        1. Menzie Chinn Post author

          Rick Stryker: I welcome your suggestions of additional covariates. I welcome even more your provision of the time series, so I don’t have to do the hard work of compiling. But for now, the correlation is there, and personally (as opposed to a social scientist), I would say the null hypothesis is that Trump has *not* accelerated mass shooting casualties (ya gonna have to find a heckuva omitted variable to overturn the correlation).

          1. Rick Stryker


            It’s not a valid statistical argument to say that for now the correlation is there until critics supply some extra covariates. The correlation is not necessarily there if those excluded covariates are correlated with the dummies. A priori, I don’t believe your regression argument because I’ve looked at the stories behind the mass shootings and it’s obvious that Trump had nothing to do with them. It’s you who are making the regression argument. It’s up to you to explain why you think a regression of mass shootings on 2 dummies would likely not lead to biased and inconsistent parameter estimates. Even with just the 2 dummies, the Trump dummy is not significant at conventional confidence levels.

            Of course it’s hard to get any covariates. That’s why any regression-based argument on the determinants of mass shootings is going to be dubious.

          2. 2slugbaits

            Rick Stryker I don’t think anyone is saying that the Poisson regression is waiting for Econometrica publication. Obviously, there are some econometric issues. For example, ideally you would want a relatively constant arrival rate. It’s not clear that including a time trend entirely resolves that problem. And you could argue that at least two of the covariates are correlated with one another. And you’d want to compare the loglikelihoods of a restricted model without the Trump dummy against the unrestricted model. Like I said, it’s not like Menzie is proposing this for blue chip journal publication. But visually it’s pretty obvious that mass shootings are on the rise, and that they probably spiked in the early Obama years and definitely spiked again since Jan 2017. My take is that the mass shooting data tracks with the rise of the far right Tea Party takeover of the GOP ~2010. The Tea Party (along with Fox News crazies) actively encouraged a kind of unrestrained anti-Obama and anti-establishment GOP toxicity into political debate. The hard rightwing nut jobs in the GOP untethered themselves from the GOP establishment and whipped up the base in their safe red districts. Crazies like Steve King and Louie Gohmert and Michelle Bachman and Joe Wilson prided themselves on their over-the-top and unfiltered rants that appealed to the hardcore base that voted in the primaries. The hard right conservative base learned that impulsiveness was okay and should be applauded for its authenticity. That’s what Trump understood when other establishment types like Jeb and Reince Priebus and Romney didn’t. And Trump played on that unfiltered, anything goes, feed ’em red meat rhetoric. I have no idea whether Trump believes half the stuff he says, but he knows that it plays well with the base and that’s all he cares about. The problem is that the base takes Trump’s tweets and the Tea Party anger as acceptable. Trump and the hard right Tea Party types actively encourage any shedding of inhibitions. Impulsiveness is okay. Trump may think that his rallies are all just dinner and a show and not to be taken seriously, but his supporters do take him seriously. They don’t know that Trump is just getting the same kind of thrill that Mick Jagger gets when he struts his stuff on stage. The MAGA hatters think Trump is giving them permission to be crazy and impulsive. And when you look at the details behind the mass shootings, one common theme is a propensity for impulsiveness and a lack of self-control. Trump and the extreme anti-immigrant politicians seemed to validate the rage felt by many of the shooters.

            So I take the Trump variable to be a proxy for the rise in the uncivil, take no prisoners, never compromise, burn down the house rhetoric that started with the Tea Party ~2010 and culminated with Trump’s election. The updated Goldwater motto is now “Extremism in the defense of white culture is no vice.”

            Also, if you accept that Trumpism is the problem, then you could always argue that Trumpism is transient and the mass shooter problem should go away when we return to a more civil political environment. That would save your guns rights argument. But if you believe that the problem is strictly mental health, then a guns rights argument could be hard to maintain over the long run because everyone is vulnerable to mental health issues. So the logical conclusion would be that no one should have guns because you, me, Menzie and everyone, is a potential mental health case waiting to happen. My advice would be to accept the “Donald did it” argument if you want to keep your gun rights.

  3. pgl

    Nope – not the way Team Trump is asking the data to be massaged at all. I hear that will provide the definitive analysis as soon as Bruce Hall and Princeton Steve complete their cooking of the data!

  4. pgl

    Bruce Hall is peddling this disgusting lie:

    “In the hours after a mass shooting Saturday, a right-wing conspiracy website used bogus “evidence” from a reputation management website to spread misinformation about the accused gunman. After initial reports identified Patrick Crusius as the alleged gunman behind the the El Paso Walmart massacre that left 22 people dead and more wounded, Jim Hoft, founder of the right-wing conspiracy website The Gateway Pundit, presented an inaccurate picture of Crusius to his readers. Selectively plucking information from the reputation management website MyLife, which can be edited anonymously by anyone, Hoft (pictured above) set about reporting that the gunman was really a Democrat, and that “Leftists” on the web were editing the gunman’s MyLife profile in real time.”

    That Gateway Pundit was peddling this lie was known before Bruce endorsed this garbage. Is Bruce too incompetent to realize this was a lie? Or is Bruce just your standard Trump sycophant who does not mind lying to the rest of us?

    1. Bruce Hall

      whoever, can you be sure this was not a left-wing conspiracy by an obviously left-wing site? Hmmm. Still waiting for “official” information…. But regardless, Crusius was an evil, disturbed person.

      1. pgl

        Your lie was busted. It does not matter the political affiliation of the person who caught this. It happened. And yet you still try to claim what Gateway Pundit tried to peddle is a fact. Lord – I would suggest you have shredded what little credibility you ever had but the truth is that you have been a blatant liar from day one. Hard to shred the credibility of someone who has zero credibility.

        I should apologize to Kelly Anne – she is a disgusting and serial liar but at least she is not as incredibly low as Bruce Hall.

  5. John

    If you are convinced from a regression that Trump’s rhetoric has caused additional deaths from mass shooters, would you be convinced from a regression that the rhetoric of Black Lives Matter has caused additional deaths in certain cities? If not, why?

    1. Menzie Chinn Post author

      John: I’m not convinced of a causal relation. Just sayin’ … never thought a Poisson regression would elicit such cries of “foul!”

    2. macroduck


      In the normal scheme of things, you develop a hypothesis, then test it to see if the data falsify the hypothesis.

      The hypothesis is a critical element. One can speculate that women who go from house to house asking for food or money curse those who refuse. It is entirely conceivable that, with a small enough sample over a limited time, the data would not falsify the witch hypothesis. The witch hypothesis is, however, really stupid.

      Is it stupid to think that a United States president holds sufficient sway over some peoples’ behavior that he could incite violence? I don’t think so – I was around for the Kent State shootings – but perhaps you disagree.

      Is it stupid to think that a bunch of people wearing “Don’t Shoot” on their chests cause a rise in murder? I think so, but perhaps you disagree.

  6. pgl

    Must read:

    “Where do lone wolf mentally ill mass murders get their ideas about a Hispanic invasion?”

    Paul Campos follows his question (title) with quotes from Trump and one Republican Texas Senator. He concludes:

    “The Republican party is a domestic terrorist organization, that is getting Hispanic people in this country murdered. It should be treated as such.”

  7. Bruce Hall

    I thought you’d find this interesting. It’s a compilation of terrorist attacks/mass shootings since Abraham Lincoln was shot through Oct. 2018 and goes into details about the type of attack and the political or religious affiliation of the attacker.

    What seems to be the case is that there are “clusters” of “popular” attack modes: shootings, bombings, attacks with vehicles, attacks on planes, etc. Since correlation is not necessarily causation, I can’t say that with certainty that these clusters are because of copy-cat attackers, perhaps egged on by media publicity, but that’s plausible.

    While it is a debating (and statistical?) exercise to blame the politician in power for the acts of the perpetrators, I’m not sure that the numerous shootings, bombings, vehicular attacks, and islamic perpetrators during the Obama administration are any more his “fault” than the shooting cluster during Trump’s tenure is his. I’m more inclined to conclude that evil and deranged people find a way to kill and can be inspired by the “successes” of those whose stories grab the headlines.

    1. pgl

      Still defending Trump we see. After your embarrassing hyping of the lies from the Gateway Pundit, one would think you would just go bye bye. But do continue to poison this comment section with your sycophatic devotion to the Racist in Chief!

        1. pgl

          This is from the same idiot who hyped a Gateway Pundit lie even after the world knew it was a lie. Checking sources? LOL – you never do.

        1. baffling

          bruce, that op-ed was very poorly written and counted on people like you to infer a particular outcome, while providing ZERO actual evidence to support much of anything other than innuendo. really, if you are relying on this type of work to support your position, not sure what to say. once again, rather than link to an op-ed puff piece, provide a paper with a strong analytical presentation of the data and discussion of the results. this opinion piece had none of that.

    2. 2slugbaits

      Bruce Hall If a depressed and suicidal person stands on the ledge of a tall building and someone in the crowd urges that person to jump, is the person in the crowd responsible for the resulting death? That jerk in the crowd might not be entirely responsible, but it’s hard to believe that he or she is blameless. Trump is like the jerk in the crowd.

      Unfortunately the term “terrorist” is getting muddled by the media. In the old days we used to reserve terrorism to refer to organized acts intended to freeze the populace with terror in connection with a clear political objective. So Timothy McVeigh was a terrorist and the 9/11 attacks were terrorist attacks. The KKK during the Jim Crow era was terrorism plain and simple. But a lot of these mass shootings are not rationally connected to any specific political objectives even though a warped political ideology may intensify their rage and motivate their actions. It’s unclear if the El Paso shooter used terrorism as a rationally connected tool to achieve a political objective or if he was just someone on the edge of the building with the crowd below urging him to jump. My guess is the latter. The Dayton shooter doesn’t appear to have had any political objective whatsoever. His chief political complaint seems to have been with older moderate Democrats, but with one exception those killed were all under 40 and most under 30. And one was his sister and he was with her just a few moments before the shooting. Anyway, I think we need to be more careful in the way we conflate “terrorism” and “mass shooting.” We see the same confusion in the way some media equate “mass shooting” with a high murder rate. These are all different phenomena and we shouldn’t get them all jumbled together. BTW, I wouldn’t call the Lincoln assassination an act of terrorism. It was clearly a continuing act of civil war intended to decapitate the government. It was not intended to terrorize the populace into inaction, which is the raison d’etre of a terrorist act.

    3. Ithaqua

      Actually, you can check the “cluster” hypothesis in the models that Menzie (and I, for a different model in the previous thread) built by looking at how overdispersed the events are relative to a Poisson distribution; the overdispersion parameter was quite small, around 1.07 IIRC (1.0 is the Poisson assumption), so there is actually very little evidence that the Trump bump is just a big cluster of non-independent events that happened to occur over the last year or two.

    4. macroduck

      Mr. Hall,

      The critical issue to Menzie’ analysis is the change in the number of mass killings. There has been a rise during Trump’s presidency. No rise, no hypothesis. Do we have evidence of a rise in mass killings during Obama’s tenure? If not, then no, we don’t blame Obama. There is a baseline level or trend of mass killings, which we can reasonably blame on Wayne LaPierre and Oliver North, and a rise in killings which we can reasonably speculate is the result of Trump normalizing bigotry.

      By the way, if there was a rise under Obama, but not targeting groups he vilified, then the hypothesis that a president normalizing bigotry can lead to mass killings is still not supported by the data.

      Please, don’t throw that weak stuff. It’s like you’re pitching at the bat.

    1. pgl

      “Mass Shootings Can Be Contagious, Research Shows”.

      Dillon Roof was hoping for a race war. The El Paso shooter was hoping for a race war. It seems Bruce Hall is fanning the flames hoping for the same. Atta boy Brucie. Trump should give you a bone!

      1. Bruce Hall

        whoever … and all of those Islamist attacks during Obama’s administration were because “Obama’s followers” were hoping to incite a religious war? Your thought process needs a bit of cleaning up. But keep trying.

        1. Menzie Chinn Post author

          Bruce Hall: Run Poisson regression (or negative binomial) 2000M01-2019M08, you don’t get a significant coefficient on an Obama dummy. Just FYI.

        2. pgl

          Your statistical claims need some serious checking. BTW Obama tried and succeeded in taking down Islamist efforts to kill us. Trump is inciting White Supremist terrorist. YUUUGE difference. Oh wait – you are defending the White Supremist in the White House. I bet you look good in KKK clothing!

        3. noneconomist

          Bruce, I must have missed the Obama rallies where he singled out specific Republican representatives while identifying them as dangerous unpatriotic fanatics. Or the times he labeled specific opponents in Congress as fascists. Or the occasions where he railed against Christian immigrants. Or the day he felt a judge who happened to be Jewish couldn’t be trusted. Or the times he led the chants to lock up Mitt Romney or John Boehner or Mitch McConnell.
          Fill me in, please.
          Oh, BTW. I do remember the time the governor of Texas assigned the Texas National Guard to be on the lookout for a possible armed invasion of Texas by the U.S. government. Thank God Chuck Norris was on hand to teach self defense to otherwise defenseless Texans. Seems somebody is always wanting to invade Texas.
          And I fondly remember a local Tea Party parade that featured a picture of the new President as an African witch doctor. And, hard to forget a few drivers in this area proudly displaying bumper stickers proclaiming ” Don’t blame me. I voted for the American.” In 2016 I must be one of those who cast all those illegal votes–3 million or so–that deceived people into believing Trump did not win the popular vote.
          Yes, nothing racial or potentially incendiary there.

        1. pgl

          It seem you are trying to justify the murder of 22 people in El Paso the same way Kelly Anne Conway is doing it. Exploit the tragic deaths of the Dayton 9 for raw political garbage. Kelly Anne gets paid for her disgusting racist garbage. But Brucie does the exact same thing for FREE. WHAT A GUY!

          1. Menzie Chinn Post author

            pgl: Look, give it up. Bruce Hall will only be happy with complete exoneration of Mr. Trump. Trump’s call to stop the infestation and invasion by non-whites had nothing, absolutely nothing to do with anybody’s belief it’s okay to go shoot up a group of primarily hispanic individuals. And, if it did, what’s wrong with that – America is for only those of Norwegian descent.

  8. baffling

    bruce, do you think mexican immigrants feel safer walking down the streets of el paso today or last week? any reason for that change? are you interested in reducing that fear, or in the great words of dick stryker, is your feeling “so what?”

    1. pgl

      With Bruce and Kelly Anne – it is more than “so what”. It is a new opportunity to exploit deaths for partisan purposes. Kelly Anne gets paid for this hate filled garbage. Bruce is spewing it for FREE!

          1. Willie

            It wouldn’t be the first time. He’s lost more money than any human on the planet ever lost before.

  9. Moses Herzog

    I’ve been a big Susan Rice fan for quite awhile. I’m hopeful she will be part of ANY Democratic White House Cabinet in a high ranking position either in the State Department, or as a civilian official overseeing our military. She is sharp-minded, very articulate, and a great example of the very best public servants America has to offer when voters choose to be literate about what is going on in the world around them:

    1. Moses Herzog

      I just wanted to add an addendum, Susan Rice would fit in well in a high ranking National Intelligence/National Security position also. Sharp-minded people can to many things—she’s a multi-talented woman.

    2. pgl

      Not to be confused with Condi Rice. Condi – worst NSA EVER but always a suck up for Bush-Cheney. Susan – very competent.

Comments are closed.