Home > statistics > JP Morgan suicides and the clustering illusion

JP Morgan suicides and the clustering illusion

February 20, 2014

Yesterday a couple of people sent me this article about mysterious deaths at JP Morgan. There’s no known connection between them, but maybe it speaks to some larger problem?

I don’t think so. A little back-of-the-envelope calculation tells me it’s not at all impressive, and this is nothing but media attention turned into conspiracy theory with the usual statistics errors.

Here are some numbers. We’re talking about 3 suicides over 3 weeks. According to wikipedia, JP Morgan has 255,000 employees, and also according to wikipedia, the U.S. suicide rate for men is 19.2 per 100,000 per year, and for women is 5.5. The suicide rates for Hong Kong and the UK, where two of the suicides took place, are much higher.

Let’s eyeball the overall rate at 19 since it’s male dominated and since may employees are overseas in higher-than-average suicide rate countries.

Since 3 weeks is about 1/17th of a year, we’d expect to see about 19/17 suicides per year per 100,000 employees, and seince we have 255,000 employees, that means about 19/17*2.55 = 2.85 suicides in that time. We had three.

This isn’t to say we’ve heard about all the suicides, just that we expect to see about one suicide a week considering how huge JP Morgan is. So let’s get over this, it’s normal. People commit suicide pretty regularly.

It’s very much like how we heard all about suicides at Foxconn, but then heard that the suicide rate at Foxconn is lower than the general Chinese population.

There is a common statistical problem called the clustering illusion, whereby actually random events look clustered sometimes. Here’s a 2-dimensional version of the clustering illusion:

There are little areas that look overly filled with (or strangely devoid of) dots.

There are little areas that look overly filled with (or strangely devoid of) dots.

Actually my calculation above points to something even dumber, which is that we expected 2.85 suicides and we saw 3, so it’s not even a proven cluster. Although it could be, because again we probably didn’t hear about all of them. Maybe it’s a cluster of “really obvious jump-from-a-building” suicides.

And I’m not saying JP Morgan is a nice place to work. I feel suicidal just thinking about working there myself. But I don’t want us to jump to any statistically unsupported conclusions.

Categories: statistics
  1. ionf
    February 20, 2014 at 8:02 am

    I may be speaking out of school, here, but I would suspect the sorts of things that drive one to suicide inhibit getting a job at a fortune 500, and more specifically a big bank. JPMC is big enough that it has its own suicide rate — this is the first I’ve heard of this story, but if most countries report cause of death and employer, ; it should be possible to calculate it. Just so you can more firmly ground your discussion of the clustering illusion.

    • February 20, 2014 at 10:03 am

      If we’re going there, though, we also have to take into consideration a high stress environment where people don’t get enough sleep and they are often working overseas away from friends and family.

      • ionf
        February 21, 2014 at 12:16 am

        Well, but, now we’re into guessing what makes people suicide. Statistics make me worry that my children will be suddenly stricken by malnutrition. It’s not going to happen. I’m in a subpopulation for which that’s probably very unlikely, but I only have the overall rate. That doesn’t mean I should use it. It’s painting apples orange.

  2. Mike Mintz
    February 20, 2014 at 8:30 am


    I disagree with the assumption that the JP Morgan suicides are “just regular men” and “just like any other JP Morgan employee”. Remember, with stats you can prove anything.

    if you look into the details of the deaths, they all have weirdness that is not the usual suicide pattern.

  3. Guest2
    February 20, 2014 at 8:51 am

    Contagion effects of suicide are well known. “The Sorrows of Young Werther” propelled Goethe to fame. Can you measure similar social network effects?

  4. February 20, 2014 at 9:07 am

    Mathbabe – I am a big fan of yours but you went too far with this one. As a former JPM employee for 10 years I am well aware of how stressful the job can be. But, now, as a licensed social worker who has worked with families with suicidality and major depression, I am really taken aback by your insensitivity to the tragedy that is suicide and which by definition always impacts the survivors. I don’t think I need to point out the survivors are the ones who will read these articles. I do agree with your math analysis, but math doesn’t always need to be the mediator of truth with such tragic human loss. Don’t know what else to say…..

    • February 20, 2014 at 9:12 am

      I am sorry you feel that way. I didn’t mean to offend.

  5. Josh
    February 20, 2014 at 9:56 am


    Thanks for keeping us aware of the human costs.


    I do think you make a great point that we shouldn’t jump to conclusions based on a few data points and the way stories get hyped up for no reason.

  6. FogOfWar
    February 20, 2014 at 10:11 am

    Condolences to the families and this discussion shouldn’t be taken as diminishing their loss in any way–most of us can only imagine the difficult times they are going through.

    One analytic point on the denominator: the 255k employee number includes Chase, the retail bank along with it’s massive branch network. I forget the exact breakdown, but this is by far the lion’s share of the headcount at most retail/investment bank combos. If all the people in question were actual investment bankers, it does change the analysis pretty significantly.


    • February 20, 2014 at 10:22 am

      Yes and no. If we looked we could probably find an even smaller group those guys fell in. I think a better way to decide the pertinent “bin” is to ask what the media would report on if there were another suicide – a Chase bank teller? Back office? I’m guessing yes.

  7. John Coleman
    February 20, 2014 at 10:52 am

    When I was in college, I read a news report about suicide rates of people in professions, lawyers, doctors, dentists, etc. It stated that engineers, of which I am one, had the lowest rate. I mentioned this report to a very pretty girl I was very interested in. She said that, if she was married to an engineer, she’d commit suicide.

    • February 20, 2014 at 10:56 am

      Ouch! She was way off base, though, nerds are awesome and sexy. FYI.

  8. Jim Birch
    February 20, 2014 at 11:28 pm

    IMHO there’s nothing offensive about that article.

    On the contrary, if we can’t write about things truthfully and dispassionately we are probably headed for a lot more bad stuff. Step one of dealing with any problem is eliminating our false narratives. I want a world designed run by people who can actually think in a evidence-based mode, not just emote.

  9. February 21, 2014 at 1:25 am

    The suicide rate is high for marginalised people like homeless, jobless, impoverished, addicted to drug and alcohol, ill or socially detached. I am not sure whether the average statistics of 19 men out of 100,000 is applicable here.

  10. February 21, 2014 at 8:19 am

    This is like the weird global cluster of suicides at Marconi 10 years ago. It must have been because they were all highly skilled engineers that they picked such bizaar ways of killing themselves.

  11. Thomas Ball
    February 21, 2014 at 9:31 am

    Very timely post…a recent Forbes article went into detail on the clustering illusion…which opens with this amazing story, “One ordinary night in Las Vegas (if there is such a thing) when I was sitting at an ordinary blackjack table with five others, something extraordinary unfolded. As the cards came out, face up, we were all astonished to have each received a hand totaling 21.”


    The reporter cites David Hand, author of what will soon be the Bible on risk estimation, The Uncertainty Principle, which is also just out from Farrar, Strauss and Giroux…

    Worth mentioning in this is Jung’s contribution, synchronicity, which involves experiencing a set of events as meaningfully related even when they’re not likely to be causally related. Jung described this as the “temporally coincident occurrences of acausal events.” Of course acausality is a red flag to any red-blooded empiricist.

    Related to Jung’s synchronicity is Koester’s Roots of Coincidence…another stunner from the late Hungarian polymath…

    And Sting’s album, Pat Metheny’s song…

  12. February 21, 2014 at 11:33 am

    Hi Cathy, it’s always great to see people debunk spurious clusters (autism clusters, cancer clusters, etc.) so thanks for the writeup. The back of the envelope calculation gets us a nice first-order idea. In this case, I’m worried about the fact that suicides are not evenly distributed across the population. People who hold certain types of jobs, who live in cities, who are on certain medications, etc. are more likely to kill themselves. Also, I’m guessing that suicides also are clustered in time and not evenly spread over the year. So, perhaps a second-order analysis is needed. But like you said, a better estimate of the risk would only make your point stronger.

    This clustering phenomenon is related to something larger. It’s a kind of narrative fallacy where the narrative is based on tying observables to observables, and ignoring factors that are less or not observable. My favorite example is the case of small soccer players, who I think are hurt by the fact that their height is observable. Whenever they lose the ball, fans complain that they are too small/lightweight. However, plenty of tall players lose the ball too (because they are clumsier, perhaps) but no one says they lose the ball because they are too tall.

  13. Tyche
    February 22, 2014 at 4:15 am

    Excellent reasoning why one should be cautious about press reported trends based on little data.

    I would point out that it’s not true that Suicide rates in the UK and Hong Kong are much higher than the US . The UK 2011 rate for men was 18.2 per 100,000 and for 5.6 for women (source: http://www.nisra.gov.uk/archive/demography/publications/suicides/suicides_in_the_united_kingdom_2011.pdf ) , marginally down on the US rates. Hong Kong 2012 rates are 16.9 for men and 9.1 for women n(source: http://csrp.hku.hk/web/eng/statistics.asp ) though these last figures are provisional rather than official rates.

    Of course cross-country comparisons of mortality rates by cause of death are always tricky, differences in coroner reporting practices ,registration delays and how cause of death is actually recorded make these comparisons harder, but there is little official evidence that UK and Hong Kong suicide rates are overall higher than the US , rather at least in the case of the UK , the rates are very much similar to the US ones.However , I don’t think these figures make substantive differences to the point you are making.

  14. February 22, 2014 at 8:23 am

    I clicked through and read the article! Yes, this is a small subset of JPMC’s 270,000 (per the article) employees. But, I don’t think the implication is that there’s some secret relationship between the suicides, so much as that this is fall out from comp day. It’s odd that the a article doesn’t refer to bonuses at all, but that is probably an explicit choice to avoid retribution. I get that I’m reading in a little,

  15. John smith
    February 23, 2014 at 5:33 pm

    I think you should ask yourself this question. If you worked in a skyscraper in the uk, would have any chance of getting onto the roof if you wanted to commit suicide? The answer is that without some serious insider help you wouldn’t stand a chance.

  16. Buckaroo Banzai
    March 18, 2014 at 12:55 pm

    Hi Cathy,
    There’s a huge problem with your analysis. It assumes that the suicides that are widely reported are the only ones that happen in the company. In fact, they are the only ones we know about. I’m quite confident that a tiny fraction– maybe 10,000 or so– of those 270,000 employees would rate media attention if they committed suicide. When looked at in that light, this is about 30x the normal rate.

    • March 18, 2014 at 5:24 pm

      First of all, I think I mentioned this problem. Second of all, where did you get the figure 30?

      The real problem is that once it does become a media issue, any and all suicides do seem like news. That’s why it’s so hard to pin down something like this.

  1. No trackbacks yet.
Comments are closed.

Get every new post delivered to your Inbox.

Join 976 other followers

%d bloggers like this: