Thursday, December 18, 2008

The most dangerous job on earth?

I hadn't realised what an insanely dangerous job it is being President of the USA. Four sitting presidents (just less than 10% of the total) have been murdered while in office, something I think we can fairly class as a job-related death. Since there have been Presidents of the USA since 1789, I calculate a death rate per 100,000 worker-years of 1826. This means that being President of the USA is fifteen times more dangerous than being a timber-cutter (the most dangerous normal occupation in the USA and nearly eighteen times more dangerous than being a fisherman in the UK. It is, however, safer than being a member of the Black Gangster Disciples during the period surveyed by Levitt & Venkatesh - even during the non-war periods, footsoldiers experienced a death rate of 4980/100K worker years, although L&V note that other gang research in Boston found a death rate for crack gangs there which was more or less in line with being President of the USA.

Presumably this fairly obvious piece of arithmetic has been done for about a thousand op-ed columns already, but I hadn't seen it before.

Update: Much more back-of-envelope calculation of crude actuarial tables, in comments. Is the true death rate for Popes 427 per 100,000 pope/years or as high as 2100? Is a fisherman a fisherman while he's not fishing, and how should we look at military death rates? Is it more dangerous to be a soldier in Iraq or a King of England?

Further update: OK, thanks to some sterling work from "Robotslave" we've got presidents, Prime ministers, US and Iraqi Army, French and English monarchs and Hezbollah. Anyone else wanting to contribute, dig in.

Extra update: On the other hand, I am getting sceptical about some of this data. Surely if four US Presidents had really been killed, we would expect to see at least twelve who had just been wounded. I can only think of one (Reagan). Which suggests that the truth about JFK is that he is still alive and his death was faked by Lancet investigators who hate America.

66 comments:

  1. What, you're just going to leave it at that? No comparison to rates for Head of State (African), UKian Monarch, Latin American Dictator, Etc?

    I know, I know, that would take an hour of actual work. I'm not about to do it, either.

    ReplyDelete
  2. But are there any op-ed writers (apart from the odd Nobel-winning economist) who can actually do arithmetic? Some of them actually pride themselves on not being able to.

    This is why you should be an op-ed columnist, obv.

    ReplyDelete
  3. Also, it ought to be noted that the USian death rate statistics don't include members of the military killed outside the US.

    Damn lies, and all that.

    With those fatalities included, enlistment in either the Army or Marine corps would definitely put you in the top 5 most dangerous professions in the past 5.75 years, though my envelope-back didn't award #1 to the armed services (total deaths, of course, are quite a bit higher than the figure for lumberjacks).

    ReplyDelete
  4. I did google for that actually but forgot to put it in - US forces have experienced roughly a 0.4% death rate in Iraq - 392 deaths /100k worker years. So about three times as much as a timber cutter but a quarter of the death rate of a President. That paper only goes up to March 2006 though.

    ReplyDelete
  5. On the basis that kinging is a 24 hour duty, counting the Cromwells as monarchs in all but name, and Edward IV's death was poison, that makes 10 English monarchs in the 942 years since 1066 to have met violent deaths,I make the death rate for British [1] kings as being 1062 per 100,000 years. (Excluding all of those who died on the rigours of campaign - a surprising number).


    [1] Yeah, yeah. Scotland, I know. Hedge kings sprouting from the collapse of the Anglo-Saxon kingdom of Northumbria don't count.

    ReplyDelete
  6. (Excluding all of those who died on the rigours of campaign

    I would count that as a work-related injury - I'm not aware of any US Presidents having died from overwork, but if any did I would bump the numbers up.

    ReplyDelete
  7. "I'm not aware of any US Presidents having died from overwork"

    Woodrow Wilson perhaps?

    ReplyDelete
  8. It's where you draw the line that's the issue - Bill the Bastard dying from falling off his horse is definitely work-related, but a 60-odd year old man dying of a heart attack, like Edward I, is trickier to genuinely call work-related.

    ReplyDelete
  9. I bet there's more have died from overeating than overwork. And has anybody done the stats for Popes?

    ReplyDelete
  10. MM> Would Felix Fauré count for our purposes?

    ReplyDelete
  11. Papal Deaths Dot Com is a frankly disappointing website. The Wikipedia page has seven popes definitely murdered and a further ten (controversially including John Paul 1) possibles. Taking only the confirmed deaths and starting the clock with Gregory the Great in 540 AD would give 476 murders/100k pope-years, slightly more dangerous than being a soldier in Iraq. Include the possibles, and it's in line with being king of the UK.

    ReplyDelete
  12. US forces have experienced roughly a 0.4% death rate in Iraq

    But this distorts the per-occupation statistic, as nowhere near all US forces are serving in Iraq. Nor even anywhere close to all active-duty personnel.

    My envelope scribblings put the deaths per 100k non-reserve soldier-years closer to 80 or so.

    Now, if you want to jimmy your statistics to look at deaths in the first two or five years of employment, I'm sure the soldiering industry could retake the lead...

    ReplyDelete
  13. Fuck, I've just noticed that they didn't count the martyrdom of the first 25 popes which I certainly would. That puts the numbers up a lot - although you basically then have to start the clock with St Peter, you still get up to 1600 on confirmed and 2100 on possibles.

    ReplyDelete
  14. No website can be that disappointing if it has an image of the skeletal hand from the tomb of Alexander VII.

    ReplyDelete
  15. But this distorts the per-occupation statistic, as nowhere near all US forces are serving in Iraq

    yeah, but I'm not counting fishermen who aren't on boats (or Presidents who get murdered after leaving office, although I don't think there are many of those) either. I see what you mean though; we need separate categories for "Army (in combat zone)" and "Army (overall)".

    ReplyDelete
  16. we need separate categories for "Army (in combat zone)" and "Army (overall)"

    I'm not at all convinced that we do, though, as troop rotation has been a core managerial practice in the soldiering industry for the better part of a century now.

    I mean, you're not disregarding the 3 least dangerous years of every presidential term when you do the math for that profession, are you?

    Also: without a cite, I'm pretty sure that you are, in fact, counting fishermen who are not on boats.

    ReplyDelete
  17. My cite is the second link in the original post - it doesn't say exactly, but it's for "work-related fatalities", and my understanding of the fishing industry in the UK is that they aren't paid when they're not at sea (as a result of which they always end up having the most unholy problems with the tax and benefits authorities). So someone who makes his livelihood as a fisherman but who isn't currently fishing shouldn't, if I am right, be contributing to the denominator of this ratio. But I am now getting a bit uneasy that it's expressed in worker-years rather than hours, so you might be on to something.

    ReplyDelete
  18. My apologies, I do entirely agree with the definition of a fisherman as "one who is currently being paid to be a fisherman." And yes, some of the typical 3-month contract (for US fishermen) is in fact spent on land or in port.

    Similarly, I think it's quite clear that the definition of "soldier" should be "someone who is currently being paid to be a soldier," which includes both those who are active but not currently deployed, and those who are currently deployed to places other than Iraq.

    ReplyDelete
  19. Actually, thinking about British history, I suspect it's not so much the head that the crown sits on that rests uneasy, more his advisors that really have something to worry about. Now there's an interesting calculation.

    Sparking off a passing comment by DD, the first, now retiring, generation of North Sea riggers are finding themselves screwed over on their pensions because some bright spark decided to saturate NI contributions in the weeks they were on the rig as opposed to the weeks they were on leave. (NI levels are calculated on a weekly basis, and for the times we're discussing, both employers and employees contributions were subject to a reasonably low cap).

    There was an interesting special commissioner's case on this recently, where the verdict was 'tough break, mate, but there's nothing I can do about it.'

    ReplyDelete
  20. I think you could count William Henry Harrison's as another work-related presidential death. Wikipedia debunks the story that the bad weather at his inauguration led to his death, but does claim that his illness was "exacerbated by the drastic pressures of his changed circumstances."

    Of course, we should also observe that presidents are, as a rule, older than soldiers or lumberjacks. In considering how dangerous the job is, we should probably take into account the different life expectancies that the people involved would have regardless of occupation.

    ReplyDelete
  21. Oh, and if you are going to separate soldiers in combat from others (and actually, I don't think you should), I don't see the justification for lumping "Popes reigning under an anti-Christian empire" in with later popes.

    ReplyDelete
  22. It needs to be standardised against an age-equivalent population, which would make it even higher, since men the age of a POTUS get murdered less than younger men. Also I would have thought an ethnicity-specific standardisation would be in order.

    I wonder how it compares to the death rate for Iraqi civilians (if one imagines being an Iraqi civilian in itself as a job, and does a straight comparison of worker-years). Can't even be bothered looking for the figures.

    ReplyDelete
  23. I've just remembered that crab fishermen in the Pacific Northwest enjoy a much higher death rate than commercial fishermen generally, varying from year to year but usually topping 400 per 100k worker-years.

    This is slightly higher than the death rate for the subset of US soldiers working Iraq (though still far below the risk associated with being a US President).

    So. If we are going to take the profession of "soldier" and separate out those employees who happen to be working today's shift in Iraq and then use the more sensational figures thus obtained for purposes of argument, then we jolly well ought to take the figures for Pacific Northwest Crab Fishermen Who Are Actually At Sea and use those stats in preference to the ones for the fishermen profession.

    My point, growing increasing buried, is that the profession of soldiering really shouldn't have to be several times as dangerous as the runner-up before we lament the lives lost to it (particularly when total fatalities for the occupation dwarfs that of the winner by percentages), or try to use the numbers to scare off potential new hires (and a great fat success you'll have with that approach to keeping young men away from anything dangerous, I'm sure).

    ReplyDelete
  24. OK, vox populi vox dei, I am clearly in a minority here and therefore hereby admit that the apples to apple comparison is for soldiers across their whole anticipated period of enlistment. Which would basically be, as RS says, about a quarter of the in-Iraq estimate above and would put them slightly below fishermen and quite materially below timber cutters.

    ReplyDelete
  25. "anticipated?"

    We haven't been using projected numbers for any of this, have we?

    ReplyDelete
  26. No, they're all historical numbers, I just meant that we should recognise the existence of rotations as in your calculation(although I still actually think that "Soldier In Combat Zone" is a worthwhile category, I am bowing to the mob on this one).

    ReplyDelete
  27. I'll agree that "Soldier in a Combat Zone" is a worthwhile category, but it's clearly not a profession, if we're talking about US soldiers.

    It might be more useful to find statistics for soldiers whose managers do not have the option of rotating them out of the "Combat Zone" (scare quotes to indicate another semantic can of worms) such as, say, soldiers in the Iraqi Army.

    Pity, then, that the US DoD doesn't release the casualty figures they gather for the Iraqi army, police, security forces, or whatever else they're calling Iraqi soldiers lately*.

    There are estimates, of course. Fiddling a bit with the ones in this report, together with numbers gleaned from Wikipedia and other reliable sources, suggests an occupational death rate somewhere (likely substantially) north of 700 per 100k.

    OK, now let's have somebody do Hezbollah...


    * Yes, yes, the army and police are distinct components of the Iraqi Security Forces. The meager data available, however, does not allow us to untangle the various branches. I tried doing it for the police, using an Iraqi minister's figure for policemen killed from 2003-2005 (which was greater than any of the total Security Forces deaths in the report above) and came up with 19,200 deaths per 100,000 worker-years.

    ReplyDelete
  28. I'll take the easier task of a risk assessment on being Prime Minister. Spencer Percival was the only assassination, but Thomas Cromwell was executed, which I would consider a work-related death. I'm also counting Henry Pelham (died in office, health impaired by stress of job), Pitt the Younger (died of work-related liver failure), Lord Palmerston (died of a chill caught while on official business in Russia). George Canning and Lord Wilmington also died in office, but I can't find any reason to consider them work-related deaths.

    It's actually quite difficult, though, to know how many worker/years of Prime Ministering there have been. Wikipedia reckons that a sensible beginning for the office would be 1688, which means that I need to remove Cromwell from the list. So, four since 1688 = 1250 deaths/100k prime minister years. Adding in Canning and Wilmington would take it up to 1800, in line with the USA.

    ReplyDelete
  29. Here's one way to look at "The Deer Hunter" - you had twice as much chance of dying from job-related injuries as a coal miner in Pennsylvania in 1970 than as a US soldier in Vietnam in 1970. So Walken, De Niro and that lot were taking the safe option.

    ReplyDelete
  30. Actually it's not in line with the USA, as I'm adding in that guy who died of pneumonia, which brings that up to 2283. The 25 martyred Popes should definitely be included (after all, in any long run of data, you're bound to pick up some periods of global rule by an anti-Christian totalitarian empire, aren't you?), so the figure there is 2100 (NB: for popes I'm just counting violent deaths, because every Pope dies in office, by definition). And the UK is 1875 for Prime Ministers counting Canning and Wilmington and 1062 for kings (actually monarchs, thinking about it, since one of the 10 was Lady Jane Grey), so we are currently the world leaders in occupational health & safety for our head of government and of state.

    ReplyDelete
  31. Looking at the French now and my God, it's like early Popes. I am not going to count "fell off horse while chasing girl" or "dysentery caused by greedy excesses" as necessarily work-related, but even just counting deaths in battle there's dozens. Also not counting deaths which simply happened during wars with family members, or deaths of monarchs after abdication on losing a war - although these obviously occurred as a direct consequence of the occupation, the deceased was not a king at the time of death.

    ReplyDelete
  32. I think any calculation that includes "work-related liver failure" is highly suspicious. The job of Prime Minister doesn't necessarily involve drinking vast amounts, even in 1805. And William Hague's biography reckoned it was duodenal ulcers that did for him. (Arguably more work-related - stress, you know.)

    ReplyDelete
  33. OK, using the criteria of counting a French royal death as work-related if it was a death in battle, a murder or duel, or a hardship-related death while on an active military campaign, and assuming that Louis V (the Lazy) Louis X (the Stubborn) and John I (the Posthumous) were all actually murdered, there were 9 French monarchs killed during the 1010 years and 100 days (non-contiguous) of French kinging between 843 and the Third Republic. That means that French kings actually have the lowest occupational death rate so far, at only 890 per 100k king/years! And that's even in the absence of the improvements in monarch safety since the 19th century, which helped to bring the British average right down. I thought it was going to be amazingly high, but things calmed right down after the Carolingians moved off the scene.

    But!!! On the other hand, I am now challenging Anonymous's estimate for England provided above. I count only eight work-related deaths of monarchs in England/Britain since 1066 (Harold Godwinson and Richard III, died in battle; Edward II, Henry IV, Henry VI and Edward V, all murdered; Jane Grey and Charles I, executed), which would bring us down to 842. I would prefer, though, to add Edmund the Magnificent and Saint Edward the Martyr (both murdered) and stretch the clock back to Egbert King of Wessex, which would make ten deaths since 802. That's a rate of 829, same difference really.

    ReplyDelete
  34. So, I picked the following numbers out of a hat for Hezbollah: around 1000 full-time members, adjusted upward a little for volunteer hours put in during the scuffles with the IDF, and 250-500 dea... er, gloriously martyred while Hezbollahing* since 1985.

    This puts the dea... martyrdom rate in the range of 1000-2000 glorious sacrifices per 100k terrorist-years

    Being an organized terrorist, then, is close to (maybe a little less) risky as being the boss of the United States of America.

    * Most of these in direct military confrontations with Israel, rather than terrorisming. I'm a little uncertain whether it's correct to count suicide bombers in the work-related deaths, but as they only amount to a few dozen of the total at best, I've gone and thrown them in there.

    ReplyDelete
  35. I would say count suicide bombers on the same principle as we used for soldiers - if you join Hezbollah, there's a chance that you will be assigned suicide-bombing duties and that's part of the risk you take when considering a career in gloriousness.

    ReplyDelete
  36. Consider, though, the possibility that the suicide bombing corps is all-volunteer. Does this make it a distinct profession?

    Even if it doesn't, shouldn't we still consider "Hezbollah Member Deployed to Suicide Bombing Duty" to be a worthwhile category?

    ReplyDelete
  37. Sorry by the way, Anonymous - I went wordblind there for a minute and didn't see the clearly set out reason for our differing figures - "I count the Cromwells as monarchs in all but name". I actually prefer to exclude the Cromwells as I've excluded comparable figures in France and can't be bothered going back to redo the calculation, but if you include them, Anonymous is right, sorry.

    ReplyDelete
  38. I would be interested to know what the death rate per mission is for suicide bombers - I mean it's obviously very high, but the authorities do manage to stop quite a few without killing them.

    But the rate/100kwy number is just going to run into an even more serious version of the fishermen problem; the time from being told "Ahmed, you're up" to bang is going to be really short and really dangerous. We'd need details about how Hezb manages its bench.

    ReplyDelete
  39. And if you include two Cromwells in the English monarch stakes, that bumps up the numbers to 1206/100kky for the period since Egbert.

    ReplyDelete
  40. Can you have a dead King of England before there was an acknowledged king of all England?

    ReplyDelete
  41. Richard II not Henry IV surely?

    Richard I died in battle.

    What about William Rufus? The official story was hunting accident but an assasination can't be ruled out.

    My list of violently dead monarchs would be:

    William II, Richard I, Edward II Richard II, Henry VI, Edward V, Richard III (mem. to self, if anointed King do not adopt the name Richard) and Charles I.

    So that's eight since the conquest. Nine if you want to count Harold Godwinsson.

    ReplyDelete
  42. hmmm the body count for English monarchs is rising ... good spot on R1!

    Justin - I see your point but I'm not happy with a statistical rule which excludes Alfred the Great. Maybe I should check Welsh monarchs (where the "was he really a king, or just a local princeling or thug who happened to have a tame literate monk to write about him?" issue is really serious. I will naturally be taking the most partisan and controversial approach possible, basically saying that at any time, whoever controlled the territory of Gwynedd was King of Wales.)

    ReplyDelete
  43. You really need to consider Standardised Mortality Ratios here, and for the period of time you're considering the reference population is going to be time-varying. Hamas activists, for example, are dying against the backdrop of a very dying-prone population; likewise Cromwell.

    You need age-standardised mortality ratios, which you just can't get for most of the history of the prime ministership, or for US presidents. Palmerston, for example, an old man in the 19th century going to Russia - his death was not so unusual, and some of the other deaths were consistent with the experience of men of that age in that time.

    ReplyDelete
  44. Surely if four US Presidents had really been killed, we would expect to see at least twelve who had just been wounded.

    Not so, thanks to our old friend "advances in medicine".

    If they'd only had the foresight to be shot in the era of antibiotics and X-rays*, Garfield and McKinley would most likely have been merely injured in the workplace, instead of dying six days (McKinley) or 11 weeks (Garfield) after being shot.

    So you'd have two presidents killed on the job, and four wounded (Teddy Roosevelt and Reagan rounding things out).

    Furthermore, I don't think we've tried very hard to find presidential injuries not related to assassination attempts, so the numbers might have to be further adjusted by falls down stairs, sickness at state dinners, and so on (I'd count anything where the employer picked up the tab for the hospital visit).


    * in both cases, doctors failed to find a bullet lodged in in the presidential abdomen.

    ReplyDelete
  45. robotslave, I think you missed a (hilarious!) dig at the Lancet denialists there. They all work on the idea that there must be 3 iraqi wounded for every 1 killed, and any stats which don't reflect this inimitable law of nature must be false.

    Anyway, until someone can come up with decent stats on death rates in a reference population of 55+ year old men in the 18th century who were still working and travelling, it's unlikely we will be able to get a clear sense of how dangerous the PM job is in the UK. I suspect that the standardised mortality ratio for PMs (based on the figures given here) would be below 100. But sadly the ONS didn't exist back then, and so we don't know... so it's all speculation

    ReplyDelete
  46. robotslave, I think you missed a (hilarious!) dig at the Lancet denialists there

    Do you, now?

    And you, sir, seem to have missed the bit earlier where we went over the problem of what constitutes a job-related injury/death. (Hint: Standardized Mortality Ratios have little to no bearing on it).

    ReplyDelete
  47. Oh god, tell me that whole Standardized Mortality Ratio thing isn't an extended, belabored "joke" reference to the dull innards of an old Lancet Study debate that you just can't let go of...

    ReplyDelete
  48. well I nearly missed the joke...

    I see a long discussion about the correct way to discuss person-years for the job categories. But both jobs (PM and fishermen) need to be compared to a suitable reference population, such as "men". I think. then you can compare their death rates with each other.

    ReplyDelete
  49. Ah, so it is a dead-horse Lancet joke, then.

    Thanks for that, I'm sure some here will have picked up on it in time to enjoy it.

    ReplyDelete
  50. I'd like to enter "member of the Senate of ancient Rome" as a contender in this competition...

    ReplyDelete
  51. I am not convinced that SMRs are really worth the trouble here - I mean, 10% of all people to have been elected President of the USA have died in office. That's such a huge risk factor that the baseline risk hardly matters.

    ReplyDelete
  52. Oliver Cromwell's death wasn't work-related; even if the story is true that he refused treatment with quinine-rich bark because it was associated with Jesuits, that's a religion-related death (arguably an autodarwination).

    Quite a few other, lower-level Commonwealthsmen died in the line of duty or else by show trial or revenge killing later on.

    French presidency seems to be unusually safe. Jaures was the opposition leader when he was shot; I suppose you have to count Pierre Laval's execution. Oddly, the hit-to-kill ratio is very low. So far there have been two attempts at Sarko, Chirac was the subject of a couple, and Charles de Gaulle was pretty much continuously targeted by assassins of one kind or another (I believe he holds the record).

    ReplyDelete
  53. Actually, better revise that. There was an attempt on Mitterand as well, but the only assassination was that of Carnot in 1894. Felix Faure and Georges Pompidou both died in office.

    Laval was executed as a collaborator, so deffo a workplace death. But Faure died of a stroke which occurred while he was enjoying an afternoon blowjob from his mistress. Pompidou apparently died of Waldenstrom's disease, which is a hereditary genetic defect affecting the blood.

    Then, of course, you have to classify Napoleon's death. There's also Paul Deschanel, who resigned due to a psychiatric illness which involved a variety of odd incidents, including randomly kissing women at rallies, falling off an express train while climbing out of the window, bathing in the fountains at Rambouillet and climbing trees naked in the park; supposedly his wife signed the letter of resignation.

    After resigning he recovered enough to be re-elected as senator for Eure-et-Loir.

    ReplyDelete
  54. Then, of course, you have to classify Napoleon's death

    I coded him as a monarch but decided that his death, although obviously work-related (or at least, it would have been under my criterion for alleged poisonings), didn't happen on the job. So he is in the denominator but not the numerator of the French king statistics.

    ReplyDelete
  55. is richard i's death on the job? his entire crusades habit was basically sagging off, and when he died (shot by an arrow from a castle he was besieging) it wasn't even part of a crusade, just a bit of off-piste pillaging on the way back from one

    john - as well undertaking all kingly admin w/o being accorded the respect of being king yet -- had to pay for the vast debt his brother ran up dicking round in the holy land: for a while he apparently proposed to do this by selling the country's assets to the muslims in spain

    ReplyDelete
  56. dd, baseline risk does matter. The death rate amongst people the presidents age is quite high, while amongst men the same age as fishermen it is very low. This will tend to pull down SMRs for presidents and push up SMRs for fishermen. Of course with only4 deaths, the confidence interval around your rate for presidents is very wide so statistical comparison is irrelevant...

    This concern is even more obvious when comparing hezbollah (high background rate in refugee camps) with US soldiers (mostly a low background rate). I think it's the soldiers who are getting the raw deal there, statistically.

    There is some discussion amongst epidemiologists about how to compare indirectly standardised rates, so maybe we can only eyeball the figures and leave it at that. But death rates presented without a reference population are possibly the first mistake of epidemiology, methinks.

    (Sorry if this posted twice, I don't seem to see the comment).

    ReplyDelete
  57. Would you consider Ariel Sharon's fate work-related? Further, should we actually be looking at incapacitation in office as well as actual death?

    (Thought for the day: Can you say you did as much for peace as the guy who cooked Ariel Sharon's breakfast?)

    ReplyDelete
  58. Sir S:

    If we were comparing death rates of persons in various professions, you would have a point; as we are instead comparing only the rates of job-related fatalities, you do not.

    Yes, you can go ahead and adjust a subset of total deaths by a fraction for those edge-case deaths where it's difficult to determine whether the death was job-related, or assign a a fraction of a death on a case-by-case basis, but as dd said, for most professions this is not going to noticeably affect the end result.

    For fishermen, it's quite easy to determine whether or not a given death was job-related, for the vast majority of deaths. There is nothing ambiguous about "crushed by machinery" or "fell overboard."

    For Hezbollah, I included only deaths due to suicide bombing or direct military confrontation with the IDF; there is no ambiguity at all as to whether the deaths included in my estimate were job-related, regardless of "background" death rates. The problem is determining which suicide bombings were conducted by Hezbollah, and what percentage of the Lebanese killed in wars with Israel were full-time employees. This uncertainty is reflected in the results I came up with.

    For US Soldiers, dd and I both (indirectly) used official DoD casualty figures for the Iraq theater. One might argue that the DoD is lying about some of those deaths, but knocking maybe a dozen off of more than six thousand otherwise unambiguously job-related deaths isn't going to affect the end result very much. It is worth noting that military deaths outside the Iraq theater were not included, so the job-related fatality rate for US soldiers is actually going to be a bit higher than my initial estimate.

    The Prime Minister number is the only one likely to be affected substantially by uncertainty over whether a given death was job-related or not, as I think dd took liberties when he included Pitt the Younger and Pelham, (but not Palmerston, as he wouldn't have been in the climate that killed him if it weren't for the job).

    The correct thing to do there, though, is not to go faffing about with your precious SMRs, which won't do much to illuminate the issue, but to present a result that reflects the underlying uncertainty, ie, as a range from 625-1250 per 100k (for the initial estimate, I can't keep track of all the changes in the roster dd has made since).

    I'm still at least half-convinced that you're having an elaborate Lancet-nostalgia laugh here, instead of just being too dim to distinguish death rates by profession from work-related death rates.

    ReplyDelete
  59. I see your point Robotslave, and for hezbollah you might be able to argue that one doesn't have to adjust the rate for the background rate of the population its drawn from. But in general daniel is discussing all-cause mortality (e.g. strokes, dying of a cold when you're in Russia, etc.) in a group of people who have been self-selected for healthiness (sick people generally don't run for these positions). Given all-cause mortality in the background population is also quite high, you need to disentangle this.

    This problem even still exists for hezbollah. Quite a few men of glorious martyring age in Lebanon have died of rocket attack even though they aren't part of Hezbollah; even more so in Gaza (if we were considering hamas) or Iraq. Similarly US soldiers actually have a pretty high rate of death due to deliberate shooting or vehicular injury when they aren't working, and injury is the biggest killer of men in the fishermen's age group, at work as well as at play. So to compare all these groups we need to consider how much they would have died anyway.

    For example, daniel said 10% of presidents died. If the death rate for an equivalent age cohort of ordinary men is 5% then we have an SMR of 200. But given the confidence interval on those 4 deaths is 1 to 10, we actually have a death rate of 2.5% to 25%, giving a 95% confidence interval for the SMR of 50 to 500, where 100 equals parity. Not many conclusions to be drawn there...

    In other news, I'm pretty sure that the BMJ or the ANZJPH have published an article showing that the most dangerous place in the world is the set of a television soap - people in those shows die at a staggering rate.

    ReplyDelete
  60. Sir S:
    No, daniel only counted assassinated presidents (and then added in W. H. Harrison); if you counted all deaths in office, you'd need to include Taylor, Harding, and Franklin Roosevelt.

    ReplyDelete
  61. plus the 12 wounded ones...

    the point still holds, with murder among the reference population as the reference rate. The choice of reference population is pretty arbitrary, and a decent study of these things would probably compare all-cause mortality in both groups as well, to see if the murder rate was consistent with a generally higher death rate (maybe presidents live dangerously in general - I'm sure Taylor, Harding and Roosevelt all died of Syphilis).

    ReplyDelete
  62. the point still holds

    No, it doesn't.

    Assassination in every case would not have been attempted had the office-holder not been holding the office.

    You've got a really bad case of "everything looks like a nail" there, Sir S.

    Either that, or you're drawing out the dead-horse joke to absolutely heroic lengths, here; I suspect this is actually the case, as you've gone right back to conflating confirmed profession-related fatalities with all fatalities for persons employed in a given profession (in the cases of Hezbollah and Fisherfolk). Your nick is more than a little suspect, too, fwiw.

    I don't begrudge a man a bit of fun along the lines you're following here, but it's considered good form to pull up stakes and vanish once the locals have figured out what you're up to.

    ReplyDelete
  63. I'm not sure what you're getting at with comments about my nick, robotslave, and even the most cursory review of the link attached to it would probably serve to disabuse you of any silly notions about dead-horse lancet jokes (whatever they may be - i confess myself at a loss).

    While it's true that the assassinations wouldn't have happened were the presidents not presidents, it's also true that if the presidents weren't presidents they wouldn't be so protected from street crime, and might not be working. And presidentiality is a full-time business (didn't Lincoln die out of office hours...?) so the general rate at which people his age died of injury and crime needs to be considered.

    I've heard it said that young Americans get shot at a pretty high rate in the US. Putting them in the Army overseas (e.g. Japan) removes that risk for them, but shifting them to Iraq doesn't. Any Lancet denialist can tell you, however, that Iraq is no more dangerous than Washington DC. So we need to consider those balancing risks before we just go ahead and present the death rate (particularly since being a soldier in Iraq is also a full-time risk). This is what SMRs are for.

    ReplyDelete
  64. Surely if four US Presidents had really been killed, we would expect to see at least twelve who had just been wounded

    Another suspicious fact. All of the presidents surveyed lived on an arterial street and spent a lot of time in public places. Can these results really be fairly extrapolated to the full presidential population?

    ReplyDelete
  65. Further, should we actually be looking at incapacitation in office as well as actual death?

    In which case FDR definitely counts towards the US total, as having run for his fourth term while aware that he was unlikely to survive it.

    ReplyDelete
  66. Not having read the comments, but having searched them in vain for the words "astronaut" and "shuttle", I am wondering whether the risk of taking space shuttle flights was considered? With 14 deaths in a total of 1137.17 flight days, this is a pretty deadly occupation.

    Data: http://en.wikipedia.org/wiki/Space_Shuttle_program#Flight_statistics

    ReplyDelete