Terrorists Vs. Chairs: An Outlier Story

The other day I needed to know how many people were killed by chairs, and while searching I came across the Washington Post’s You’re More Likely To Be Fatally Crushed By Furniture Than Killed By A Terrorist. It argues that worrying about terrorism is irrational, because terrorists kill fewer people each year than falling furniture, and nobody cares about that:

Consider, for instance, that since the attacks of Sept. 11, 2001, Americans have been no more likely to die at the hands of terrorists than being crushed to death by unstable televisions and furniture […] What accounts for the fear that terrorism inspires, considering that its actual risk in the United States and other Western countries is so low? The answer lies in basic human psychology.

I once saw the perfect response to this kind of argument on Twitter, but I forgot to screenshot it, so I’ll have to try to draw it from memory here.

One person posted a graph that looked something like this:

And somebody else edited it to look like this:

And whoa I had never realized before how sketchy it is to start your interval for recording the average number of terrorist attacks the day after the last major terrorist attack.

I mean, I know why people do it. It’s because September 11 was an “outlier”, and outliers should not be counted. Problem is, depending on your distribution, sometimes “outliers” are the only thing that matters.

Let me give an example. Suppose I’m trying to make an argument that earthquakes are totally not a problem for Haiti at all, that there’s no need to invest in earthquake preparedness, and that Haitian people who worry about earthquakes are stupid. I make a graph showing that since January 13, 2010, fewer Haitians have died per year from earthquake-related causes than from crazy furniture-related mishaps. This is totally 100% true. Look at those stupid Haitians, worrying about something that on average never hurts anybody!

(the Haitian earthquake of January 12, 2010 killed about 100,000 people)

I’m sure there are a zillion small Richter 1.0 and Richter 2.0 earthquakes in Haiti all the time. I’m sure our monitoring interval of January 13, 2010 to present picked up lots of these and correctly noted that they don’t kill anybody. The only Haitian earthquakes anyone needs to worry about are the outliers.

If you start your monitoring interval on January 13, earthquakes kill 0 people/year. If you start it on January 11, earthquakes kill 20,000 people a year. Neither of these is entirely fair – one is purpose-designed to maximize casualties, the other to minimize it. I don’t think there’s an obvious fair way to do things – the best solution would be extend the interval back to infinity, but then you get into problems like Haiti having fewer people back in the day, or Haiti not having risen out of the sea yet back in 4,000,000,000 BC. Maybe the best solution is to pick an arbitrary block of time like “the last fifty years”, or to do something very complicated like using the remote historical record to produce earthquake numbers and then combine it with modern populations to produce expected casualties.

The same is true of September 11. Start the interval September 12, and you get 5-10 terrorism deaths/year. Start it September 10, and you get 200. I don’t know when the best time to start it would be. If I had to choose something, I would say maybe 1985, when jihadist terrorism got started after the Soviet invasion of Afghanistan. But someone else could choose 1776, or 2000, based on similarly arbitrary criteria. And it would all be irrelevant – September 11 either made terrorists more ambitious, made security forces more watchful, or both, and so probably changed the calculus for good.

Granted, even when you include September 11, terrorism isn’t the worst thing, and people probably do overestimate it. So forget terrorism. On average, the flu kills something like 20,000 people worldwide each year. That’s a lot, but not apocalyptically much. If you go back year after year, the average stays at something like 20,000/year, right up until you get to 1918, when about 100,000,000 people died. So flu deaths over the last century average about 1 million/year. But three years from now, average flu deaths over the last century will average about 20,000 year. A death rate of only 20,000/year might make our current efforts to contain the flu seem excessive compared to other diseases. But a death rate of 1 million/year makes them look if anything the opposite.

Even worse: did you know that giant asteroids kill about a hundred people per year, on average? This is admittedly an odd definition of “kill” and “average” given that no human being has ever been killed by a giant asteroid. But given that giant asteroids strike Earth about every ten million years, and an asteroid strike today might kill about a billion people, on average giant asteroids kill about a hundred people per year.

Actually, un-forget terrorism. I have a friend who is very in favor of the War On Terror, and he argues that the problem with terrorism isn’t the average suicide bomber who kills three people. It isn’t even the 9-11 hijackers who killed three thousand people. It’s the group that steals a nuke and kills three million people. Just as “on average” a hundred people die each year from giant asteroid strikes, maybe “on average” thirty thousand people die each year from nuclear terrorism. All you’d need for this to be true is one nuclear attack per century. And that’s as bad as an average flu season!

The thing about falling furniture is that there’s probably not going to be a furniturepocalypse where suddenly millions of people all perish at once after being struck by a really really big desk. Furniture is constant. Terrorism isn’t. The whole point of black swans is that we pay too much attention to constant risks and ignore the outliers, especially the outliers which outlie so far that they haven’t happened yet. That’s true whether it’s terrorism, earthquakes, pandemics, or AI.

I worry that someday many years from now, terrorists are going to have some improbable victory which is even more destructive than September 11. I worry that uncounted people are going to die. And I worry that ten years later, someone is going to post on Facebook about how “From the day after ISIS nuked London through today, on average fewer people per year have died of terrorism than from hilarious accidents involving bedside dressers!”

This entry was posted in Uncategorized. Bookmark the permalink.

402 Responses to Terrorists Vs. Chairs: An Outlier Story

  1. ikea, doombringer says:

    > Maybe the best solution is to pick an arbitrary block of time

    There’s the “exponentially weighted moving average”, where you weight the current year at 1, and the year before that at d where d < 1, and the year before that gets weighted d^2, and the year before that gets weighted d^3 and so on to infinity.

    But then you’ve replaced your problem “how do I choose the size of my arbitrary block of time?” with “how do I choose the exponent for my moving average?” and maybe it’s not enough of an improvement to be worth the math.

    • DanielLC says:

      Well, you don’t have to worry about average number of flu deaths per year this century suddenly dropping from 1,000,000 to 20,000, so that’s something.

    • Alejandro says:

      Wait, is that a throwaway username crafted solely for this post? Because if not, it is the most perfectly relevant username ever for the first comment in a post discussing furniturepocalypse.

    • Izaak Weiss says:

      IMPORTANT: If you do this, you also need to multiply your entire sum by (1-d) to get the weighting right. Otherwise, you’ll get the wrong answer.

      Example: With d=.75 and 20 being the number of events per year every year since the big bang, we can type `sum 0 to inf (20 .75^n)` into wolfram alpha and it tells us that the answer is 80; this is obviously wrong.

      However, 80 * (1-.75) = 80 * .25 = 80/4 = 20

      • Emma says:

        Yes. When you calculate a weighted average you divide a weighted sum by the sum of just the weights. And

        \Sum_n=0^\inf d^n = 1/(1-d),

        so dividing by it is the same as multiplying by 1-d.

        Edit: Also, is there a way to use latex or unicode in comments? Specifically to write capital sigma and super- and subscripts?

        • Robert L says:

          edit: that works – it is & Sigma ; without spaces

          sub and sup tags don’t work.

        • Izaak Weiss says:

          Not everyone in the comment thread is a math person; I just wanted to make sure that this method wasn’t used badly, as without the normalization constant you will get a bigger number than you should.

    • person says:

      But then you’ve replaced your problem “how do I choose the size of my arbitrary block of time?” with “how do I choose the exponent for my moving average?” and maybe it’s not enough of an improvement to be worth the math.

      The lesson there is that the numeric value isn’t what’s important. In fact, having infinitely many distinct choices, each equally valid, you could not possibly think there’d be a “right” choice.

      But nobody cares about the exact numbers anyway. Numbers are used to extract some qualitative notion of the relative likelihood or rarity of events. And for any fixed basis d above, you get that. Fix any arbitrary d, then set some qualitative milestones such as the likelihood of lightning strikes in your exponentially weighted moving average with basis d, the likelihood of changing careers, the likelihood of getting in a car accident, the likelihood of a drought/tornado/etc. until you have a sense for how much different somewhat familar events are “worth” in this scale, and then by comparing the numerical value for some new thing like terrorist attacks, against those for familiar things, you have a sense of how proportionately more likely things are.

      (This works because although the exact numerical value varies with d, the RATIO of different things does not depend on d.)

      • Good Burning Plastic says:

        (This works because although the exact numerical value varies with d, the RATIO of different things does not depend on d.)

        Yes it does. If A has been getting rarer than it used to be and B has been getting more frequent, then the smaller d is, the more recent years will be weighed, the more frequent B will be calculated to be with respect to A.

    • Steve Sailer says:

      Here’s the One Percent Doctrine as explained by Dick Cheney (played by Richard Dreyfus) to George W. Bush (played by Josh Brolin) in Oliver Stone’s biopic “W:”


  2. FooQuuxman says:

    The thing about falling furniture is that there’s probably not going to be a furniturepocalypse where suddenly millions of people all perish at once after being struck by a really really big desk.

    This can be solved by shaping an asteroid into a Really Big Desk.

    What makes terrorism different from furniture is that the decision process of other people is intimately connected with it’s occurrence, how you respond to it can wildly change how much you get.

    • DanielLC says:

      > how you respond to it can wildly change how much you get.

      Isn’t that the point? Why worry about something if your response can’t change how much it happens? How much we worry about the flu can wildly change how much we get.

      • MugaSofer says:

        The difference is that it can change wildly. And you may not realize you’ve changed the amount of terrorism wildly until it’s too late, because terrorism consists mostly of outliers.

        • albatross says:

          It seems to me that there are two different things going on w.r.t terrorism deaths vs furniture deaths:

          a. Terrorism is an intentional action by a thinking entity that can change tactics or be convinced to give up and do something else.

          b. The impact of terrorist attacks in the west are dominated by the outliers (9/11), rather than by the average (nutcase with a Glock in a mall).

          Those two aspects need to be treated separately. Street crime fits (a), but not (b)–muggers are thinking beings who respond to incentives, but there’s not some low-but-unknowable probability that a small group of muggers will somehow manage to carry out a single mugging that makes the total damage done by muggings go up by a factor of 100 that year.

          On the other side, earthquakes and hurricanes are natural phenomena that aren’t being directed by intelligent adversaries, but whose impact is probably dominated by the outliers.

          • Steve Sailer says:

            Violence gets our attention more than accidents because violence is political.

            It has costs to the victims beyond just the first order effects. The costs of being victimized by violence traditionally over human history and prehistory were not just the immediate ones but the costs of losing resources to groups that use violence successfully to control resources. The losers don’t just die, they have less access to food and thus fewer descendants.

    • Matt H says:

      Terrorism is memetic and contagious. Order is fragile and can break-down. This is why it’s worth responding to. Not that I know how to respond. I can properly secure my tv to a stud, one TVs actions don’t influence another’s, at least not until the singularity.

      • Gudamor says:

        I badly misread this comment at first:
        “Terrorism is memetic and contagious. Order is fragile and can break-down. This is why it’s worth responding to. Now that I know how to respond: I can properly secure my tv to a stud”

    • Furslid says:

      How may people die from terrorism is not just linked to how many terrorists there are. I don’t think that terrorism has been less deadly because there are less terrorists in the world since September 11. The change also comes from how deadly each attack is.

      Falling furniture kills 1 or 2 people in each lethal incident. So if there are 10 falling furniture incidents a year, and they include unusually deadly incidents, there might be a death toll of 20. 10 die in a year with low deadliness furniture falls. So the main driver for how many people die from falling furniture can be how many lethal furniture falls there are.

      What does a year with an average number of terrorist attacks, but unusually deadly terrorist attacks look like? It doesn’t just have twice the number killed as a year with an average number of attacks of low deadliness.

      • Autolykos says:

        I don’t think that terrorism has been less deadly because there are less terrorists in the world since September 11.

        There seem to be, however, less qualified and competently led terrorists since September 11, 2001. Few of the more recent attacks seem to be planned by a central leadership, and judging from those that were (e.g. Mumbai) the quality of their personnel seems to have deteriorated from “skilled and patient operatives you can entrust with a complex plan that takes years to prepare for” to “village idiots you can hand a Kalashnikov or a suicide vest and point at the nearest crowd”.
        All of the more ambitious plans that I heard of failed horribly, and quite a few were infiltrated (and possibly orchestrated) by police or intelligence services right from the beginning, with the actual wanna-be terrorists far too incompetent to have pulled the plan off by themselves (like the Sauerland group in Germany).
        The fixation on suicide attacks may be good theatrics (and terrorism is mostly theater), but it has the tendency to burn quickly through your best personnel – something any sane organization will avoid like the plague. There are just so many more civilians than even semi-competent potential terrorists, that we can probably just sit back and let the police handle them like they would any other common criminal. They will run out of recruits way before they’ll make a noticeable dent in our population.

    • rossry says:

      This can be solved by shaping an asteroid into a Really Big Desk.

      Relevant: “The Vending Machine Murders”

  3. SamO says:

    The role of intention seems to be really important on this issue. Terrorism is worrisome merely because of the historical body count. There are multiple (citation needed) potential terrorists who would actively plan a large scale attack if given the opportunity. In my mind, part of the mechanism that limits future opportunities for terror attacks is actively worrying about them.

    • SamO says:

      Correction: Terrorism *isn’t* worrisome merely because of the historical body count.

    • Warwick the Wild says:

      This is the problem with estimating the value of a lot of security-related efforts. Did the zillion dollars you spent on security and intelligence successfully deterred all the thieves, spies, and terrorists, or was it all a giant boondoggle cooked up by security firms/agencies to make money/justify their budget?

      We can sort-of guesstimate with foiled attacks or some such, but even that seems pretty spotty, especially if you’re positing a deterrent effect (either by scaring would-be terrorists off or by shutting down supply chains enabling them).

      • Murphy says:

        Plus it can go in the opposite direction. In an effort to “deter” terrorists you send a load of jackbooted thugs to kick in doors and smash teeth all around the world.

        Amazingly the people involved seem to like you even less afterwards and the number of attempted terrorist attack just seems to go up rather than down.

        To top that off in the course of finding people to kick their teeth in you may find people are making deals with ancient unspeakable evils that have murdered more people that even the most ambitious outlier terrorists ever could dream of:


        • J says:

          The one thing I will never forgive Obama for is letting the CIA pose as polio vaccination health workers when going after bin Laden.

          • Douglas Knight says:

            You’ve fallen for propaganda. That story is completely false. It’s just a … CIA cover story. (Also, it was hepatitis, not polio.)

          • J says:

            Can you offer any proof? Here’s the Guardian, Scientific American, and NPR all affirming that the CIA posed as vaccination workers (the latter quoting the white house directly).

          • Douglas Knight says:

            What do you mean by “proof”? Does the Guardian offer proof? They say that they performed an “investigation,” but they just assert things with no explanation of how they could know. Maybe it’s just a CIA press release.

            All I can offer is Seymour Hersh, but surely he is a more reliable source than the White House.

          • Macbi says:

            Surely the CIA claiming to have impersonated vaccinators is just as damaging as them actually doing so?

          • Jaskologist says:

            Morally, is there really much difference between using vaccinaters as spies, and just convincing people that you’re using vaccinaters as spies?

          • Douglas Knight says:

            Morally, is there much difference between DUI and murder?

      • albatross says:

        Yep. And again I think there are two different effects there:

        a. The “wasted life insurance effect.” I’ve carried life insurance for many years now (I have a wife and kids who need my continued income). Since I’m still alive and writing this, clearly that money was wasted, right?

        That’s an error you could make if you spend a lot of ways to respond to / prevent terrorist attacks, but nobody tried one. That doesn’t mean the TSA’s porno scanners and blue-glove-groping routine is justified by definition, it just means you can’t definitively say “see, no attacks were stopped this year, therefore there’s no value.”

        b. You can’t detect most of the attacks you prevented. If the right set of visa applications to the US had been rejected, we might never have had a 9/11 attack, and we would never have any idea that our visa rejections had prevented it.

        It’s notoriously hard to tell if you’re preventing something with your precautions–think about statins or baby aspirin or quitting smoking. Those are all cases where we can gather lots of statistics on lots of people using / not using the preventatives and try to tease out subtle effects. I don’t know how we’d do anything like that for most of our war on terror activities.

        For the record, I actually think most of our (US) war on terror activities range from not very helpful to useless to actively harmful. But figuring out whether (and in which cases) I’m right looks pretty hard, in practice.

        • Warwick the Wild says:

          For the record, I actually think most of our (US) war on terror activities range from not very helpful to useless to actively harmful. But figuring out whether (and in which cases) I’m right looks pretty hard, in practice.

          I can’t say I disagree with that. I’m virtually certain that a lot of our security spending is useless theater (consider last year’s embarrassments to the TSA), and I think that is an area where you can test the efficacy of your methods. You might not be able to say whether or not there really is a threat, but you can reasonably say that if there was a serious threat, these measures wouldn’t have stopped it.

          • 27chaos says:

            I agree that it is theater, but terrorists seem drawn to choosing airports as targets because of the theater, which helps make them easier to manage, and civilians respond to theater by feeling safer about flying, so it’s not as useless as it might appear at first glance. Still fairly useless, admittedly.

          • John Schilling says:

            Terrorism is about hacking people’s risk-assessment heuristics to make them feel more afraid than would be merited by the material consequences of the threat. Is it really a surprise that part of the defense involves hacking people’s risk-assessment heuristics to make them feel safer than would be merited by the material effect of the security procedures?

          • Who wouldn't want to be anonymous says:

            When terrorists still manage to be successful despite the theater, people end up even more risk hacked by the terrorists. Naturally the state would have to expand the theater to un-hack people again. Recurse as needed.

            The problem is that theater is way, way, way more expensive than terrorism; and successful theater does not appreciably increase the cost of terrorism, but successful terrorism appreciably increases the cost of theater. Despite the imbalance of resources in the starting configuration, I suspect that the state would run out first.

          • John Schilling says:

            The problem is that theater is way, way, way more expensive than terrorism

            What’s your estimate for the cost of, say, restructuring the entire airline industry to fly with half the present number of passengers because the other half are too scared to fly?

            What’s the cost of a Worse Than Donald Trump administration, when the existing government and opposition both lose their legitimacy due to blatant inaction in the face of a persistent terrorist threat?

            Yes, yes, I know – nobody you know is afraid to fly on airliners. Indeed, all your friends have cut back on flying because of the intrusion and inconvenience of the TSA’s security theater. Nice bubble you’ve got there.

          • Glen Raphael says:

            What’s your estimate for the cost of, say, restructuring the entire airline industry to fly with half the present number of passengers because the other half are too scared to fly?

            If – as you seem to assert – an appearance of airline safety really matters to customers, then it is too important to leave that up to government! Rather, let’s let individual airlines and airports decide for themselves how much or little security theater to provide their customers. Let airlines and airports compete along the “appearing safe” axis. If most customers want lots of theater, airlines that provide lots of theater will gain market share and the ones that provide less will lose market share until “the market” decides how much is the right amount.

            More likely I suspect you would see market segmentation. Comfort Air might specialize in providing LOTS of theater – Strip searches for everyone! Bring back the nail clipper ban! – while Freedom Air specializes in serving people like me. Freedom Air terminals return us to a 1960s-era level of security where non-customers can walk out to the gate to meet incoming flights and passengers can bring a rifle as a carry-on. Freedom Air has lower ticket prices and faster end-to-end travel times because the time and money budget doesn’t include a TSA-caliber screening procedure.

            We’ll see who wins.

          • John Schilling says:

            If – as you seem to assert – an appearance of airline safety really matters to customers, then it is too important to leave that up to government! Rather, let’s let individual airlines and airports decide for themselves how much or little security theater to provide their customers.

            Do you understand that the single most important contributor to the perception of security is the uniforms with American flags on them? No matter how good e.g. and airline might be at providing actual security, and no matter how lazy and inept the TSA’s people may actually be, almost nobody will trust a bunch of venal mercenaries working for a greedy corporation the way they will a civil servant sworn to the defense of the nation.

            Some things, governments are better at than corporations. This is one of them. Again, not actually securing airplanes against terrorists (they might be better at that), but being trusted to secure airplanes against terrorists.

            More likely I suspect you would see market segmentation. Comfort Air might specialize in providing LOTS of theater […] while Freedom Air specializes in serving people like me

            Yes, and the easiest way to win that competition isn’t to provide actual security or even the illusion of security, but to put out scary advertisements about how dangerous Comfort Air is, about how Freedom Air only puts on a good show but they’re really as rotten inside as the TSA was.

            Congratulations – you’ve just incentivized American citizens who command something like fifty billion dollars in market capitalization, to devote their advertising resources and expertise to doing the terrorists’ work for them. ISIS thanks you for your contribution to the global jihad.

          • “Do you understand that the single most important contributor to the perception of security is the uniforms with American flags on them? ”

            Private security people can have uniforms with American flags on them.

            As you may know, there are airports, such as SFO, where security has been subcontracted to private firms instead of being done by TSA. Do you have any evidence that people feel less secure as a result?

          • Glen Raphael says:

            put out scary advertisements about how dangerous Comfort Air is

            I’m pretty sure the incentives don’t work that way.

            In most industries the health of the ENTIRE INDUSTRY is a major concern for participants in it and they are inclined to cooperate and share information to achieve that end. Airlines are even more inclined in this direction than most due to unusually strong network effects – the value to customers of being able to fly commercially is much larger when customers know they can get anywhere and are certain they’ll be able to get back than it is if one’s ability to get where you want to go and not get stranded is strongly tied to the whims of or continued survival of particular specific companies.

            Thus in times of special crisis (storms, volcanoes, crashes…) airlines with free seats have often been known to honor one another’s tickets (and settle up later) just to get everybody home. And in normal times they do stuff like code-sharing, partnering behind the scenes to increase the apparent size of their route network.

            The fact that falsely advertising “Company X planes are deathtraps!” would harm the health of the industry in general is obvious enough that no airline in their right mind would do it any more than one peanut butter company would try to gain market share by claiming other peanut butters contain rat poison. Peanut butter companies don’t want eating peanut butter to seem scary; neither do airlines want flying to seem scary.

            (My family ran a travel agency for much of my life so I’ve heard a lot of stories of both airlines and travel agencies helping out their competitors to serve the common goal of collectively keeping customers safe and satisfied.)

          • Glen Raphael says:

            @David Friedman:

            Private security people can have uniforms with American flags on them.

            Yup! Google image search confirms Covenant Aviation Security (CAS), the private firm that protects the gates at SFO airport instead of TSA, does indeed wear uniforms with American flag features.

            UPDATE: I can’t find a good picture – that round shoulder thing might be a regular TSA logo badge they’re wearing. Though there’s no reason they couldn’t wear some OTHER badge with a flag/bird motif. Badges are cheap…

          • John Schilling says:

            Private security people can have uniforms with American flags on them.

            Do you understand that I was not being 100.00% literal in that statement?

            Private security people cannot reliable cause their customers to believe they are dealing with uniformed military personnel or law enforcement officers. Any such deception will not hold, and will likely backfire if attempted.

          • Glen Raphael says:

            Private security people cannot reliable cause their customers to believe they are dealing with uniformed military personnel or law enforcement officers.

            First off, you’ve provided no evidence that customers care about this. Do you have any? I fully grant that my own impression that “government employee” suggests “probably incompetent” might just be something that happens in my own bubble, but how do you know your contrary impression that “government employee” means “reassuring” is any less bubble-prone?

            Any such deception will not hold, and will likely backfire if attempted.

            Um. We gave you the example of San Francisco Airport. Has that “deception” held? Did it backfire? At last count there are 23 airports using private alternatives to TSA (up from 5 initially) and many others are considering switching despite a painfully drawn-out approval process to do so (the switchover takes a full year). According to the article I just linked:

            airports that have switched to private firms say they consider the contractors more responsive and better able to adjust staffing to address traffic surges and lulls.

            The article does not mention “customers think they’re less safe” as a reason for concern – should it?

            Recently there was a single week in which roughly 6,800 American Airlines passengers missed flights due to TSA delays. TSA has started recommending people arrive THREE HOURS early for domestic flights (up from a mere TWO last year) to accommodate screening delays. Is that enough theater or would we “feel safer” to add even MORE random delay to a transport method whose chief attribute used to be speed and efficiency?

          • At a slight tangent on the subject of TSA vs private firms …

          • Who wouldn't want to be anonymous says:

            The problem is that theater is way, way, way more expensive than terrorism

            What’s your estimate for the cost of, say, restructuring the entire airline industry to fly with half the present number of passengers because the other half are too scared to fly?

            I was speaking of the cost to commit terrorism, not the cost resulting from terrorism. That under the “terrorism is mind-hack” model, in addition to the theater you better actually be doing something effective because they can bankrupt you by forcing ever more intrusive security measures. (The implication being that terrorism is a whack-a-mole; when we’ve beaten Islamic terrorism–if we beat Islamic terrorism–there will just be new and novel things people are upset about and willing to blow things up over. So adopting theater is to set yourself on the path to certain bankruptcy.)

            But since you bring it up, I fail to see how it is different than any other asset bubble groaning on the verge of collapse after decades of government distortions in the market. Yes, yes, of course we’re obligated to run around screaming and tearing our hair out that airlines have invested in too many planes and a drop in passenger numbers will drive them all bankrupt and so ever increasing government measures to prop up passenger numbers are called for… or whatever.

            I mean, that’s really the root problem isn’t it? The real cost of flying is more than (some) passengers are willing to pay. That if those cost were internalized to the transaction VTC would eat the airlines’ business traveler lunch. Tourists would rather stay home and spend their money on video games and Real Dolls. Facebook and Instagram and Skype make it much less important for retirees to spend their pensions flying around visiting their grandkids in the four corners of the Earth. Affordable virtual reality headsets have even come on the market in the last year or so which could make visiting impressive landmarks less important than it used to be. Skype could easily be combined with virtual reality headsets, too, now that I think about it.

            I am pretty sure those things are called substitute goods, and have been enabled by the march of technology. Airlines need to compete with them, or become as obsolete as transatlantic passenger ships. Government subsidising their security isn’t the answer. That’s what all the talk about “uniforms with flags on them” is about after all.

          • bean says:

            There is a tacit agreement throughout the airline industry, manufacturer and customer alike, not to compete on safety, at least in front of the public. Some of this is because of concern for the general reputation of the industry, but some of it is also because the bar has been set so high by the government. But I’m not sure the agreement would hold in the face of attempts to privatize the whole system.
            In even the best possible case, totally privatized security will be destroyed by the next 9/11. Airliners hitting buildings hurts the building, as well as the airliner. Enforcing some minimum level of security/safety on large, fast-moving pieces of metal is a good thing, because if you don’t, said level will be set retrospectively by mob pressure in the aftermath of a disaster. If the system works right, when there is a disaster, the mob pressure is diverted and the experts can fix actual problems. To some extent, the problem with 9/11 was that it was big enough that the experts couldn’t deflect mob pressure like they normally do.

    • Subbak says:

      In my mind, part of the mechanism that limits future opportunities for terror attacks is actively worrying about them.

      You can have an argument that you can limit terror attacks by enacting efficient countermeasures, that often have very bad side-effects for individual liberties (by efficient I mean something like what Israel does, not what the TSA does). I think the loss of freedom is definitely not worth it, but I have to concede that it is possible to reduce occurrence of terrorism by being sufficiently authoritarian (but you have to do it right, authoritarianism does wrong only has drawbacks). Obviously to think of those measures, and to implement them, someone has to worry about terrorism.

      However, people who are not doing that should never worry about terrorism. Because the more people worry about terrorism, the more efficient the terrorist attacks are. After all, the point of terrorists is to use fear to advance their political agenda. If they made you afraid, then they’ve succeeded at their primary goal. Now obviously you can’t will yourself into not being afraid of terrorist attacks, but it’s better to act to reduce the global fear inspired by terrorists. This reduces the impact they really care about. While it does not directly affect the “number of people killed” impact, hopefully not seeing panicked reaction might deter terrorist actions and indirectly affect that.

      • scav says:

        Interesting. So it’s a game where each side is trying to maximise a different thing.

        We obviously win by preventing deaths, because we don’t want people to die.

        But terrorists can’t actually *win* by killing any plausible number of people: they are outnumbered by a gigantic ratio. Even if every potential suicide bomber killed 1000 people, the end game is 0 suicide bombers left and still about 7 billion non-terrorists mourning the dead and getting back to their lives.

        They only “win” by provoking a terrified response and by our making an unforced error in how we deal with them. Although of course, they don’t actually win even then. Western capitalist liberal democracy goes on apace. The terrorists just get to feel like they achieved *something*, enough to recruit more angry young men.

        • albatross says:

          To what extent do terrorist attacks recruit more terrorists and thus lead to more attacks? And what role does media coverage and political reaction to the attacks play in this?

          I strongly suspect that mass-shootings are affected by previous mass shootings and heavy media coverage of them, but I don’t know to what extent the same is true of terrorist attacks. (And there’s overlap–the Orlando and San Bernardino attacks were pretty classic mass shooting attacks, but done by people who were sympathetic to the goals of ISIS and perhaps were actually motivated by the thought of carrying out a suicide attack on their behalf.)

      • Maware says:

        The point is not “worrying” it’s “thinking and planning effectively.”

        The problem is that people don’t think at all, something happens, and we get a brief, reactive response. Over time, people complain about the effects of that response, and we slowly slip back into not thinking or doing anything. Then the next outlier incident happens, and the cycle repeats again.

        Then we get stupid articles like the WaPo one. If anything, we should think about terrorism more, and plan effective measures against it while taking into account the need for personal liberty. But stupid people who immediately go back to “oh, see? terrorism kills less people than chairs over time!” aren’t doing that, they are perpetuating the cycle. And it’s not just that, either. So many other issues are driven by the reactive-forgetful cycle.

        • Sok Puppette says:

          The thing is that there are in fact no effective measures short of creating a totalitarian dystopia. And even that might not be effective. It’s really easy to cause havoc in a universe that favors entropy. And, whatever kind of world you create, SOMEBODY is going to hate it enough to be willing to do that.

          On edit: I think I have to retract this because I find I can’t clearly articulate the standard of “effectiveness” that I’m using. Basically none of the measures I’ve ever seen taken, or even suggested, would prevent a plausibly competent actor from causing mass damage, or reduce the size of that damage. But I can’t say that they don’t raise the bar above the capabilities of a plausibly INCOMPETENT actor, and we’ve seen that there are a lot of incompetent terrorists out there.

          • J says:

            We do the things that actually help, like reinforcing cockpit doors. We just also do lots of other stupid, counterproductive, destructive, expensive things too.

          • Simon says:

            Aum Shinrikyo were flush to the eyeballs with cash, had competent engineers and chemists working on the problem, weren’t really being watched that carefully and still failed to do that much harm with the sarin attack on the Tokyo subway.

          • Sandy says:

            Aum Shinrikyo launched the Tokyo subway attack because the police were closing in on them and they thought a distraction would help them get away. It was an improvised, last-minute thing.

      • idjit says:

        My two thoughts on this.

        One, their primary goal is not to cause fear, it is to see the change they want. Yes the ideal situation for a terrorist is, have goal, cause fear, achieve goal. But it seems to me, particularly if you look at America, this does not work? Mostly you get, terrorists have goal(change in policy?), cause fear, drones. Unless their goal is something like “I want more drone strikes” then terrorism is not super effective as a tool to achieve an end as used against Americans, even if it is causing them fear.

        Two, I imagine a large number of ‘terrorists’ are not actually terrorists at all, as per the understood definition. They are not making a rational and calculated terror attack in the hopes of changing western foreign policy in the middle east from (blank) to [blank]. It seems to me that they are just living out something of a power fantasy? Like the last pockets of human resistance setting off bombs to kill a few of their new alien over lords. The aliens are never going to give them what they want, and the aliens are never going to ‘lose’. The humans just keep attacking them out of a sense of revenge, and justice, and all that jazz.

        • Zakharov says:

          Going by revealed preferences, terrorist leaders want to increase the size of their organisations, while grunts want a sense of meaning in their lives.

  4. mogden says:

    The problem with terrorism is that crafty people are trying to invent new schemes to increase casualties far beyond what we have seen so far (for example: bioterrorism). This potential risk enhancement is not occurring with wooden furniture.

    • Loquat says:

      Also, there are relatively simple precautions you can take to reduce your risk of being killed by furniture – securing heavy items to the wall is commonly recommended and doesn’t have much downside unless you enjoy frequently rearranging your furniture. There aren’t a lot of things an ordinary citizen can easily do to reduce their risk of being killed by a terrorist without changing their own lifestyle.

      • Simon says:

        And that’s because in insurance terms your risk for being killed by terrorists is considered de minimis, i.e none. What could you possibly do to mitigate a personal risk of none? Of course that’s no consolation if someone you know picks up the losing lottery ticket.

  5. I don’t have a solution, but…

    A couple years ago, I assumed that the probability of an event occurring each year had a flat prior between 0% and 100%. Then, using just the last time it occurred as evidence, I computed the posterior distribution. If I remember correctly, the expected value of the posterior distribution was 2/(n+2). In other words, if an event happened 10 years ago, it had a 1-in-6 chance of happening next year.

    Now, I don’t think the odds of another 9-11-esque terrorist attack is 1-in-9, but that’s because I have additional knowledge.

    I also agree with ikea, doombringer. This seems like an ideal place to use an exponentially weighted moving average, though just what that decay constant should be is a super interesting question. The other thing to think about is how much we can do about it.

    Presumably, it’d be quite difficult to reduce the number of furniture-relatd deaths. However, reducing cold-related deaths by preventing huge outbreaks using vaccines seems much more doable and utility-incresing even if no huge outbreak would’ve happened anyways. To me, the big question concerning terrorism isn’t so much how likely I am to die in an attack, but how effectively we can reduce that probability. This is a difficult question to answer, to say the least!

    [Edit, SamO said it better than I did: “part of the mechanism that limits future opportunities for terror attacks is actively worrying about them”]

    • doubleunplussed says:

      What additional knowledge do you have?

      I would have thought a one in nine chance of a 9-11-esque attack per year sounded plausible. Events an order of magnitude or two smaller in terms of deaths in developed countries seem to happen yearly or so, so 1 in 9 per year for a large event doesn’t sound that crazy.

      • The Nybbler says:

        Interesting idea; if we look at the smaller-scale terrorist events, do we see a consistent power-law distribution? If so, we can estimate the larger-scale events.

        • doubleunplussed says:

          …and you’ve just nerd sniped me. BRB writing Python script to fit power law to wikipedia data.

          • Douglas Knight says:

            Try a database

          • doubleunplussed says:

            Thank you, much easier.

          • doubleunplussed says:

            Here’s the result:


            Power law fit doesn’t look half bad. If you take it seriously and extrapolate you get:

            Attacks killing a thousand (± 25 %) people: once every 7 years = 142 deaths per year
            Attacks killing ten thousand (± 25 %)people: once every 250 years = 40 deaths per year
            Attacks killing a hundred thousand (± 25 %) people: once every 6000 years = 16 per year

            (the ± 25 % comes from the binning of the histogram)

            So the rare events are not contributing significantly to the overall expected death rate, as you can see in this plot, which is the same as the above except each point is scaled by its number of deaths, so we get to see the contribution of each size of attack to the overall death rate:


            If you use the power law and add up the contributions from each event size to the overall death rate (i.e integrate it), you get a grand total of 13700 ± 3200 deaths per year.

            This is slightly higher than the 7581 deaths per year average that’s in the raw data, and I think that’s probably due to this power law reasoning – the effect of expected but as of yet unobserved events. But it’s not much higher. The vast majority of deaths from terrorism—even if this power law is true—come from the smaller attacks of 10-100 deaths.

            But I don’t see why you couldn’t find yourself in a situation where the power law describing the severity and frequency of events had such a fat tail that the integral didn’t converge. In that case you would be in the bizarre situation of expecting an essentially infinite death rate from a problem, even though a possibly small rate had been observed so far.

            I’m totally on board with the idea that this might be people’s subconscious reasoning behind being more afraid of terrorism than the simple stats indicate. Perhaps our monkey brains know about fat-tailed risk, and overriding those instincts with supposedly cool reason is actually the error. As is almost always the case in clashes between system 1 and system 2, the irrational things our brain does usually have reasons behind them. The monkey brains just happen not to have the right numbers being input to them in this case.

            There’s also the meta-level that we don’t know it’s a power law, there is still a lot of uncertainty in what the probabilities of events that have never happened are. In some sense if this uncertainty about uncertainty is also fat-tailed, the fearful people might be making the right call after all.

          • doubleunplussed says:

            To add to this, the best fit power law above had an exponent of alpha = -1.5±0.1.

            That is, frequency_of_attacks ~ exp(-1.5 * number_of_deaths).

            A power law’s integral does not converge if the exponent isn’t less than -1. If the frequency of attacks were like exp(-1.0 * number_of_deaths), then the expected number of deaths per year would be infinite. But that exponent is five sigma away from being the best fit to the data, so is probably not the case. Phew, the probability of terrorism having an infinite kill rate is only five sigma (one in 2 million).

            But just out of interest, below is the same two plots as my above post, but with the exponent forced to be -1.0:

            Freq vs severity: http://i.imgur.com/5mZbrf5.png
            Contribution to death rate vs severity: http://i.imgur.com/K3kRYiF.png

            It’s a worse fit, to be sure. The second one in particular looks pretty bad. Thankfully the frequency of terror attacks in reality drops off faster than their severity. But appreciate how bizarre this is – the first plot has a line that does not look ridiculously at odds with the data (though this exponent is five sigma away from being the best one, as mentioned). And yet that line describes a probability distribution with an *infinite* expected number of deaths per year. Basically that every year you are gambling with some probability of all of humanity being killed, and the only reason your naive average death rate from recorded events so far doesn’t reflect that is essentially due to it being too small a sample.

          • Good Burning Plastic says:

            It looks like you would get a better fit with an exponent close to 1 but with an exponential cutoff.

          • Jacob says:

            I would encourage the mathematically inclined to read some of Talebs published papers on the subject, he dives into a lot of details about fat-tailed distributions. Some of my thoughts:

            1) doubleunplussed noted that the fit had a slightly higher average than the raw data. This is actually a trait of power laws. Naive interpretations will tend to underestimate the average.

            2) Fat-tailed distributions are difficult to fit, because they are most important in the tails (where the outliers are), but that’s also where we have the least amount of data. If we have some reason to believe the data is an exact power law that’s not the biggest deal, but for terrorist attacks and war deaths, we just use that form because it’s mathematically convenient. It’s extremely possible that the “true” distribution is different.

            Nate Silver talks about this in the “Signal and the Noise”, regarding the distribution of earthquake. It’s empirically observed the frequency follows a power law…except maybe at high intensities, which is where we care the most. The difference between expecting a Richter 9.0 earthquake every 500 years vs every 5000 years is a difference in building codes, but two grad students crunching the data slightly differently might get each answer.

            3) Keep in mind when looking at power laws on log scales, small visual changes can mean dramatic changes in death tolls.

            4) Regarding existence of the mean…yeah that’s interesting. Financial modellers often use a gaussian distribution for price changes, Taleb thinks the Cauchy distribution is more appropriate. It looks similar near the center but has fat tails…and no mean. I don’t know how to interpret this, it definitely makes the Von Neumann–Morgenstern utility theorem hard to apply.

          • haishan says:

            Please don’t try to derive parameters for a power law by fitting a linear regression to log-log data.

            Also, check other heavy-tailed distributions! Might a lognormal distribution be a better fit? (I’d do it myself but I’m supposed to be working…)

            In any case, I agree with Chalid that fitting a model without some good reason to believe in that model will give us, at best, a very rough approximation of the probability of a large-scale attack.

          • Douglas Knight says:

            Jacob, are you endorsing Taleb over other technical papers on fat-tailed distributions?

          • Decius says:

            You can’t have an infinite tail on expected deaths because it’s capped at human population.

            What’s scary is the nonepsilon chance that there’s a terrorist attack that kills everybody.

          • Just want to say, this is why I love SSC! Thanks for the analysis!

          • Anthony says:

            So looking at your plot, if you started with data for 20 deaths or more, the exponent would be higher, but so would the intercept. I’d think it possible that terrorist attacks with low numbers of casualties might be under-reported, as many might be (innocently) considered “ordinary” crime by the authorities, or easier to cover up by authorities who don’t want to report a terrorism problem.

          • Murphy says:


            I’m not sure the bias would be in that direction:

            When many countries have special legal powers for terrorism cases it can be tempting for police to call every damned thing terrorism.

            People heckling politicians and people having anti-war marches have had anti-terrorism laws thrown at them:


            Also lets not forget those same powers being used to remind journalists that their family are fair game to be targeted if they get too uppity:


            In a world where councils are willing to pursue littering, dog shit and people smoking in non-smoking areas under anti-terrorism legislation under-reporting of terrorist attacks seems unlikely to be any kind of likely problem.


          • Titanium Dragon says:

            To be fair to the infinite deaths possibility, we can’t actually rule it out. It is consistent with the Fermi paradox, after all.

        • Chalid says:

          While this sort of thing is interesting, I don’t think there’s much reason to trust the distribution “out of sample” on very large attacks.

          Some processes tend to generate power-law distributions, and it’s easy to see that in sciences and the like. And perhaps one could come up with a justification for a power-law within a certain class of terror attack. But the really big events aren’t necessarily going to be generated by the same processes as the little events.

          I think the thing to do would be to look at the various mechanisms that generate power laws, see which ones might explain a power law in deaths from terrorist attacks below 10^3 deaths, and from that think about whether it’s logical to extrapolate to 10^7 deaths or not. (I rather doubt it.)

          • scav says:

            Is there a power-law distribution of radius-of-effect for things that a terrorist could plausibly use to kill people? There’s obviously an upper limit to that radius, so the number of kills is then limited by population density of the target.

          • Ilya Shpitser says:

            Agree with this. You can fit lots of models to lots of data, that tells you very little, and “background knowledge => model fits well” is much more sensible than “model fits well => ex post facto explanation.”

          • vV_Vv says:

            I think the thing to do would be to look at the various mechanisms that generate power laws, see which ones might explain a power law in deaths from terrorist attacks below 10^3 deaths, and from that think about whether it’s logical to extrapolate to 10^7 deaths or not. (I rather doubt it.)

            I assume that human population density already follows a power law because of reasons.

            The more populated targets tend to be better defended, but if the terrorists manage to overwhelm the defenses, for instance by figuring out a novel attack tactic (e.g. flying an airliner into a skyscraper), they can usually kill some near-constant fraction of people at the target, yielding a power law effect.

            An attack causing 10^7 deaths would most likely involve WMDs. This is less likely than an attack causing 10^3 deaths, but not 10,000 times less likely, hence a power law is plausible at that scale.

      • K says:

        A new 9/11 attack is very unlikely to get similar body counts, simply because of 9/11. Cockpit doors are now reinforced, and pilots are not going to open them because a terrorist threatens a flight attendant with nail scissors. At most, you will bring down a few planes, for a body count in the hundreds.

        • Emma Casey says:

          A new 9-11 attack wouldn’t target a plane though. It would target a football stadium or a cruise ship or something else we’re not planning for.

        • Murphy says:

          Blowing up a dam, releasing a biological weapon, some novel attack using microdrones, other.

          Airplanes are extremely unlikely to be the vector of any notable new terrorist attacks.

        • Furslid says:

          You’re oversimplifying. The lesson of 9/11 is that there are small areas in our society where a minimal amount of force can have a disparate impact. These areas are difficult to anticipate and defend.

          Terrorists could buy people with multiple drug resistant TB tickets to the superbowl, comicon, burning man, etc. Terrorists could attack dams. Terrorists could contaminate food warehouses. Terrorists could attack electrical infrastructure.

          Modern society has a lot of vulnerabilities. Clever and evil people can attack them. This is the lesson.

          Planes were vulnerable and we can be safe by protecting them isn’t the lesson.

          • Mary says:

            There is a limit in that terrorists, unlike the Joker, aren’t doing it for giggles. They will be limited to things they can plausibly claim.

            Mind you, that’s not MUCH of a limit.

          • Publius Varinius says:

            @Mary: I don’t think terrorists are limited at all to things they can claim. Sure, terrorists need to do some things they can plausibly claim for publicity reasons, but they have no reason to refrain from doing unclaimable things to weaken their enemies.

          • houseboatonstyxb says:

            they have no reason to refrain from doing unclaimable things to weaken their enemies.

            Remember the 60s-70s? When damage was done, half a dozen groups claimed credit. This tied up the US in trying to find which one really did it. The more suspects, the more Iraqs.

            Plus which, claiming credit got attention to groups no one had heard of before or since.

          • Publius Varinius says:

            The Brazilian terrorists arrested in July were apparently planning to carry out a biochemical attack.

          • Mary says:

            “they have no reason to refrain from doing unclaimable things to weaken their enemies.”

            Sure they do. They use up resources that could be devoted to things that they could claim. They really, really, really want to claim credit because otherwise most of the value of terrorism is lost.

          • Publius Varinius says:

            Russia will not build any airplanes. That would use up resources they could spend on tanks.

            Terrorism is merely a tool that our enemies use against us, and not the terminal goal of said enemies.

      • Good Burning Plastic says:

        I would have thought a one in nine chance of a 9-11-esque attack per year sounded plausible.

        Well, there’s the fact that there was not a single such attack in the US in quite a long time before 2001. (But certain relevant things have changed meanwhile.)

        • Milo Minderbinder says:

          Well, there was the first WTC attack in 1993, which only killed 6 and injured a thousand but could have brought down both towers if the terrorists had only parked the van in a different spot.

    • Scott Alexander says:

      That doesn’t seem like a good idea, because if there are a thousand different disasters that each happen once a millennium, you’ll end up obsessing over the last one to happen. For example, you might think that giant earthquakes strike Haiti once every 4 years since it last happened six years ago, whereas in fact maybe that’s a once a millennium event and next year a giant tsunami will strike Brazil.

      This is clearly silly in terms of natural disasters, but if it’s across broader categories it might not be so obvious. If 9-11 is a once-in-a-millennium level of terrorist success, maybe we’re spending the next twenty years or so worrying a lot about terrorism for no reason.

      • wintermute says:

        It’s even worse than this: using time-weighted averages completely strips you of the ability to consider certain long-term or skewed distribution events.

        The Yellowstone supervolcano is an interesting example. On a time-weighted scale, it’s not that big a deal, even if you adjust for the fact that humans didn’t exist last time it went off. But it’s a rare, high-impact event, and even worse it’s cyclic. So using time weighting, it will only be safe when you are worried about it, and only be dangerous when you aren’t worried about it.

        And yes, obviously that’s silly to do, but it means a time-weighting system hasn’t actually given us any power to make a threat comparison here.

      • Garrett says:

        This is one of the things that Bruce Schneier has been talking about and one of the reasons that I got into EMS. Trying to harden specific targets like airplanes is mostly futile because attackers can simply pick another target like shopping malls. Instead, by beefing up our response capabilities to all events (such as EMS) we are able to better respond not only to any arbitrary terrorist attack, but natural disasters as well.

      • Good Burning Plastic says:

        If the only thing you know about Event A is that it happened once X days ago and never in the last (X-1) days, then the probability it will happen tomorrow is 1/X. If you have more information you can shift this probability up or down.

  6. 27chaos says:

    I don’t think people are making comparisons between terrorism and furniture with the expectation that we remove 100% of our attention from terrorism.

    If you want to see how ineffective our current spending on terrorism is in more detailed arguments, I recommend you read some Bruce Schneier. Actually, now that I think about it, his work is right up your alley anyway. He’s a very interesting to read smart technological libertarian blogger.

    • K says:

      It is ineffective in deterring terrorism, but that’s not its primary purpose. The purpose is to make people *feel* more secure (and although I tend to get nervous among armed guards – obviously something very dangerous must be likely to happen – it appears to work for some people), and to demonstrate that the government is doing *something*.

      • MugaSofer says:

        Don’t arguments like the one Scott discusses in the beginning also serve this function?

      • Subbak says:

        I’d argue that on the contrary, it serves to make people feel more scared. Politicians, especially right-wing ones (but not only them, far from it), will often use fear (of terrorism, of crime, of immigrants…) to get people to vote for them. If everyone is constantly talking about terrorism, worried about terrorist attacks, and reminded of the threat by constant pointless security checks, that helps them get elected, and it helps them justify their more authoritarian measures (think Patriot Act and all its friends everywhere in the world).

  7. Dr Dealgood says:

    I’m not a statistician by any means, but I could have sworn that there were methods of statistical analysis specifically for analyzing large low-frequency events.

    Ok, yeah according to Professor Wikipedia the term is “Rare Event Modelling” or REM. Typing that into google gives me a bunch of pages all pointing to or citing this paper ‘Logistic Regression in Rare Event Data’.

    So this is a sort of good news / bad news situation. The good news is that we can totally account for this sort of thing and provide meaningful comparisons between the likelihood of deadly furniture accidents and terrorist strikes. The bad news is that nobody will because it requires a level of statistical numeracy beyond that of the average scientist, much less the average pundit.

  8. Lowell says:

    Fun related fact: On average over the last 22 years, land in Florida has earned about $0.03 per acre per year in cocaine falling from the sky.

  9. Leonard says:

    Your points are interesting, but there’s a much more fundamental difference between chairs and terrorism as sources of danger: we derive great utility from the existence of chairs, which far more than balances out the negative utility of the very rare chair maiming or death. By contrast, we gain no positive utility from terrorism, or at best the positive utility we do get from it is dwarfed by all the deaders. Or the positive utility is unseen. (I.e., if you buy the “blowback” theory then it’s actually a price we are paying for driving the Soviets out of Afghanistan or something… worth it for sure!)

    This is why nobody really cares whether you start at Sept 10 or 12: even 23 deaths per year from terrorists is too much.

    It’s also why nobody makes a big point of it that 32000 people per year die on the roads. Sure, we do try to minimize it; we care. But we all know that the ability to drive around is of massive utility to us, and we accept that there are risks.

    • Mengsk says:

      This really isn’t a fundamental difference here. The reason why tolerate 32,000 road deaths is because we don’t have good (i.e. inexpensive) ways to prevent those deaths. Most of the low-hanging fruit has been picked in the area of road safety, which means that way is the only remaining way to save lives is “spend a lot of money for little marginal benefit”, “invent new safety technology” or “change the way we use cars”. Currently, we none of these are feasible, so we tolerate the 32,000 road deaths.

      A similar formulation works for terrorism. Although we don’t benefit from terrorism, we may well decide that we’re willing to tolerate 23 deaths a year from terrorism if it means that we get to pass on the dis-utility associated with implementing a horrible, costly preventative measure (such as deporting all Muslims).

      • eh says:

        Thought: by looking at https://en.m.wikipedia.org/wiki/List_of_countries_by_traffic-related_death_rate, and possibly comparing the government spending on road safety, you could approximate the market value of life in each country: very high in Finland, high in the US, vanishingly small in the DRC. With better data you could probably approximate a QALY, too.

        Does that make effective altruism a kind of arbitrage?

      • K says:

        Don’t go overboard rationalizing this. Something like 20 000 people died from the tsunami that hit Fukushima, but instead of building a slightly taller fence, Japan phases out nuclear power – at tremendous cost financially (need to replace that power), politically (dependence on importing fossils), and globally (if you believe IPCC). Even if nobody died from radiation, and a few simple modifications (putting generators on higher ground – or better, replacing old reactors by the coast with newer ones in safer places) could have prevented the meltdown.

        Humans worry about spectacular events, human-created events, things we don’t understand, things we can blame somebody for. Everybody forgets the tens and hundreds of thousands who die in tsunamis and earthquakes, but harps on about a handful that died in a plane crash, or the fact that somebody could have almost died in a reactor accident.

        • Mengsk says:

          I definitely think that people in general are more sensitive to the impact of rare, but spectacular events (like plane crashes). But I think there are reasons why it’s difficult to coordinate systemic interventions to address “chronic/mundane” hazards that go beyond people underestimating their impact. They tend to be the sort of problems where it’s easy to pass the buck along to the next administration, or implementing a solution would require the coordination of significant lifestyle changes or be vulnerable to free riders.

    • Squirrel of Doom says:

      There isn’t much positive utility from terrorism, but there is from the innocent bystanders that would be killed if we try to destroy it.

    • Murphy says:

      Talking about the utility of terrorists is like talking about the utility of only the 50 chairs which failed lethally while ignoring the millions of other chairs. Obviously those 50 chairs have almost no value, people sat on them a few hundred times then they caused a death.

      Obviously those 50 chairs are a menace and have almost no redeeming qualities.

      To be more reasonable you have to talk about the utility of the population from which the terrorists come. The rest of the worlds population provides massive utility to america, saudi arabia provides massive utility in the form of oil, the fact that 15 mostly saudi people killed a few thousand americans is dwarfed by the utility of all the things americans do with Saudi oil. or at least it is if you’re using the same scale where the road deaths are acceptable.

    • Jacob says:

      The war on terror has cost about $1.5 trillion since 2001. That money could have saved many lives by providing food, medical care, legal aid, police, education, and so on. The average statistical life is valued around $10 million, so $1.5 trillion = 150,000 lives. Because of a loss of 3,000. Doesn’t seem like a good deal.

      • Anonymous says:

        The Clean Water Act has cost about $AWholeBunch trillion since 1972. That money could have saved many lies by ListOfPreferredPoliciesHere, and so on. Because of a river fire… and no deaths. Doesn’t seem like a good deal.

        I hope you can see why this is terrible reasoning… for things like environmental regulations and military activities.

        • Saint Fiasco says:

          ListOfPreferredPoliciesHere isn’t arbitrary. Jacob seems to be arguing in favor of policies that save more lives per unit of money.

          The only incorrect part is comparing the 150000 hypothetical lives to 3000 actual lives. It should be 150K hypotheticals compared to a hypothetical number of deaths in a world where the War on Terror did not happen.

        • Murphy says:

          Jacob was pointing to the typical value of a statistical life. in general if a government intervention that’s supposed to be saving lives is saving less lives than that for a higher cost it doesn’t matter what it is or what the alternatives is, you’re probably spending more than you should on it.

          But there’s deeper problems.

          it’s so fantastically easy to wave you hands and declare that had we not spent all the money then the Klingons/terrorists would have killed us all.

          You need to deal with the problem of tiger replant rocks.

          How do you tell the difference between when you’re paying for something that actually works or when you’re paying for a rock which keeps tigers away.

          If your country has spent 1.5 trillion on tiger repellent rocks ever since a load of people were killed by a truck full of escaped tigers you need a better metric than “the head of the department of tiger prevention who’s best friend owns the company Tiger Repellent Pebble Co says that the rocks are our only option for dealing with them and we should keep spending just as much on the rocks” to decide if it’s really been worth it.

          That same guy and his friends super double plus swearing that their opinion is that we shouldn’t cut back on pebble acquisition doesn’t cut it.

          Pointing out that there have been almost no tiger attacks ever since every street was lined with the pebbles also doesn’t cut it, especially if we were in the same situation on every day prior to the day a truck full of tigers crashed.

          tigers might indeed be a threat, it might be important to stop more trucks full of tigers crashing but with so few tigers we need some concrete proof that the pebbles really are what are keeping us safe if we’re going to spend a non-trivial fraction of the entire worlds wealth on them.

      • K says:

        If you can estimate the number of lives saved and the cost of the various programs, you can get a monetary quantification of the “spectacularity” of various types of death.

        If terrorism would result in 5000 deaths since 2001, and spending $1.5T reduced it to 2000, this gives a spectacularity index of $500 million per life. (If I got the numbers right).

        The estimates are probably too hard to make meaningful – e.g. the wars in Asia have many other causes and intents than preventing terror. But generally, the amount of money we are willing to spend to prevent a death varies tremendously, and if it could be calculated more precisely, would give a good measure for how we consider some causes of death more serious than others.

  10. doubleunplussed says:

    Maybe we should fit a pink noise distribution to the past occurrences within a large block of time, and report the “average” as that of the distribution, rather than that of occurrences. That way the reported average will take into account our best estimate of the size of the tail on the distribution, which is really the problem here.

    Edit: I see Dr Dealgood beat me to it with a Wikipedia link to rare event modelling. I disagree that this is beyond people’s skill set. As a mathematically apt person it seems to me that most statistics presented to the public are incredibly primitive metrics of things even though the people producing them are definitely capable of more sophisticated analysis. I think the problem is how comprehensible the statistic is to the public, rather than what the researchers are capable of. Which is also misguided perhaps, the public also doesn’t know the difference between median and mean, so their understanding is something you didn’t have in any case. Might as well present the more sophisticated metric and call it an “average”, people will not know the difference anyhow.

    • Autolykos says:

      Totally agree on this one. Statistics should be done right, or not at all. And they would still provide value to people who don’t understand the method. If they are mathematically semi-literate, they can look the method up on Wikipedia and check whether it was applied correctly (not that anyone except a few nerds would do such a thing). And if they are completely math-illiterate, they have to take it on faith anyway – but at least they get the right numbers.

    • TrivialGravitas says:

      I think you misunderstand the degree to which people are capable of being bad at statistics. Its not that the general public is bad at stats, or that journos are bad at stats, or that scientists are bad at stats.

      Mathematicians are bad at stats in my experience. This doesn’t apply to statisticians of course, but to people in less practical fields of math. Stats aren’t hard really, but they aren’t intuitive either. Maybe those journos can do the analysis, but they can’t explain it.

      • Franz_Panzer says:

        I am a mathematician .
        I concur.

        Statisticians know statistics. Other mathematicians are certainly capable of learning statistics, but few of them do.

    • doubleunplussed says:

      Anyone interested should see my comment above:


      In which I’ve fitted a power law to the frequency vs severity curve of recorded terrorism attacks.

  11. Likewise, most governments don’t go full-blown Pol Pot, Mao Zedong, Lenin/Stalin, or Hitler, but every once in a while they do, and unbelievably large numbers of people die. This can happen even in a democratic state (Weimar Republic).

    If you haven’t read R. J. Rummel’s Death By Government, you should. In the 20th Century, governments killed 262 million people in mass murder and genocide. That’s an average of 2.6 million per year over the course of the century — another number that vastly exceeds the numbers killed by terrorism globally. That number excludes soldiers killed in wartime, noncombatants killed during attacks on military targets so long as the primary target was military, and actions taken against armed civilians during mob action or a riot.

    Book on Amazon

    Rummel’s Website

    • nm. k. m. says:

      On the other hand, for most of its lifetime, Weimar government was on the brink of becoming failed democracy, and in 15 years, it failed, to the surprise of very few.

      The argument that posits that “governments go full-blown ” ignores most of the reality of how genocidal tyrannical governments actually rise into power. Usually, with a (rapid, maybe violent) overthrow of the previous government.

    • Tatu Ahponen says:

      Rummel’s methodology is incredibly flawed and his numbers are wildly inflated compared to all serious studies on the individual genocides and other mass killings he refers to. For instance, that number, 262 million? It used to be 174 million… and a big part of the reason for the change then Rudy just went “Whoops, I forgot to include the deaths by colonialism in my original analysis, I’ll add in 50 million extra deaths due to that” – apparently just pulling that number out of thin air.


      • The fact that Rummel did not originally include deaths due to colonialism is not an argument against including them. Are you saying that he is not allowed to correct oversights? What is your specific argument against including the deaths European colonization of Africa caused?

        Even if Rummel overcounted by a factor of 2, the death toll is enormous — still orders of magnitude greater than the global death toll from terrorism.

        • Fahundo says:

          I think Tatu’s complaint was more that the 50 million number seemed to be something like a guess.

        • The large increase in his total that I remember was due to his conclusion that the famine during the Great Leap Forward was democide not accident. He discussed his reason for changing his mind on that in some detail.

        • Tatu Ahponen says:

          Not only that the number was a guess, but the whole “I just forgot about colonialism!” thing talks about how arbitrary his method is.

          • Could you quote Rummel saying that his number was a guess?

          • Tatu Ahponen says:


            “As a result of this research, I’m willing to estimate that over all of colonized Africa and Asia 1900 to independence, the democide was something like 50 million.”

            His research here seems to consist of looking at some evaluations of Congo genocide and then, indeed, making an estimation – also known as a “guess”.

            This is not by far the only problem with Rummel’s numbers, as the link I posted also notes.

    • Cerebral Paul Z. says:

      I’m planning to get a copy just so I can tell the magnificent bastard I read his book.

    • Cypher says:

      That does lump together deaths caused by different forms of government with different ideologies, which is like lumping together all deaths caused by electrical power plants.

      I certainly do think that it’s justified to demand far higher standards of evidence from Communists and Fascists (literal) for roughly this reason, however.

    • Simon says:

      And what figures does he use to quantify lives saved due to government activity? It’s going to be quite a large number I suspect.

      • Much harder to measure, since it requires you to know what would have happened without that activity.

        Rummel isn’t looking at arguments about excess mortality due to regulation and the like. He is counting people actually shot, deliberately starved to death, and the like–where it is clear that they died because of deliberate state action.

        • cassander says:

          Rummel’s numbers run a lot higher than most other estimates. He assigns Stalin a body count nearly double that of Conquest, for example.

  12. Scott Alexander says:

    Right, but then when you find that it doesn’t (eg terrorism, flu) what do you do?

  13. Mengsk says:

    It seems to be that a lot of the confusion here is caused by the fact that very unlike events are being grouped together under the umbrella label of “terrorism”. Although the the San Bernardino shootings, 9/11, and a hypothetical nuclear attack on Istanbul would all be connected by a common affiliation of radical Jihad, the preventative measure you’d take in each case are radically different.

    I suppose the heuristic would be “When calibrating our response to different hazards, we should compare the death rates of hazards where the preventative measures are comparably expensive”.

  14. Jill says:

    I agree with Thomas Redding here.
    “To me, the big question concerning terrorism isn’t so much how likely I am to die in an attack, but how effectively we can reduce that probability. This is a difficult question to answer, to say the least!”

    What can we do about it, is the question. And is what we are doing about it now helping, hurting or making no difference whatsoever?

    27chaos mentioned this writer, Bruce Schneier

    Billions Wasted on Anti-Terrorism Security

    Good Article on Airport Security

    Billions Wasted on Anti-Terrorism Security

  15. Alex Zavoluk says:

    Don’t forget the question of “what can we do about it, and how much does it affect the number of deaths”? Furniture deaths may be caused by simple stupidity on behalf of user, which seems expensive to reduce (massive campaigns about the danger of standing on chairs?)

    Terrorism is also expensive to fight, but terrorists may respond directly to what we do, in the sense that if we do nothing, terrorism will keep ratcheting up.

    • Fahundo says:

      but terrorists may respond directly to what we do

      Now I’m reminded of a Key & Peele sketch about a group of terrorists brainstorming ways to attack the USA , and at every turn they’re exasperated by the fact that the TSA has anticipated their next move. First they give up on a plan which would require a 5-inch blade, because 4-inches isn’t enough to hurt anyone, then they lament how they can’t bring 3.5-oz of liquid onto a plane, because 3.5-oz is the magic amount of liquid which makes it possible to unleash untold destruction. Then a terrorist comes up with a plan to make explosives that look like toothpaste tubes, but it won’t work if he’s only allowed to bring travel size toothpaste.

      • Alex Zavoluk says:

        I certainly don’t mean that everything we do is effective, but rather that if we had done nothing at all in response to 9/11 there would have been more attacks because terrorists would have been emboldened by our inaction.

        • Saint Fiasco says:

          Does that happen? Terrorists escalating because their enemies give them the silent treatment?

          • Edward Scizorhands says:

            If absolutely nothing was done in response to 9/11, I’m sure terrorists would have started doing it weekly.

          • ThirteenthLetter says:

            Ask the Israelis; there’s a strong argument that they’ve gone through exactly that scenario multiple times. Ignore the attacks, the attacks escalate. React violently to the attacks, things get quiet for a while.

          • Alex Zavoluk says:

            Are there examples of functional states not responding at all to terrorism?

    • albatross says:

      How would we tell if our fight against terrorism so far has made the expected impact of terrorist attacks on us larger or smaller? I rather suspect that our military interventions have made things worse overall, but I don’t know that for sure.

  16. Bram Cohen says:

    A much less silly comparison is the one about deaths caused by forms of power. The overwhelming majority of all *immediate* deaths caused by power were from the failure of the Banqiao Dam, which killed 171,000 people in 1975. Without that one accident hydro looks very safe. With it it looks fairly mediocre.

    The 9/11 attacks were hardly singular unexpected events. There was an earlier attack when I went to high school two blocks away which took out the basement of one of them. They were uniquely vulnerable and attractive targets for terrorist attack, and the lessons which should be drawn from them are about predicting predictable events rather than not predicting unpredictable events.

    • Scott Alexander says:

      Yes, this seems like the strongest argument against (non-thorium) nuclear power to me.

      • Douglas Knight says:

        Similarly, the Titanic is the strongest argument against ice cubes in drinks.

      • TrivialGravitas says:

        What are you possibly imagining nuclear could do that would be that bad?

        Chernobyl would have to happen 10 times a year to do as much damage as fossil fuels (mainly coal) do in the US when everything goes right. We would need 26 Chernobyls to meet the worst single chemical disaster (the Bhopal fertilizer plant) and 285 to meet the above mentioned dam collapse.

        And that’s based on plant designs that were pants on head retarded (no system for containment if something went wrong) run by bureaucrats who would be kicked out for incompetence if they worked at the FDA. It can’t happen in a western plant even if terrorists successfully attack it, because even a damaged containment system would block the worst of it (demonstrated by Fukashima, which isn’t expected to kill anybody at all who wasn’t actually inside the plant). But I’ll sport you the worst possible numbers, because why not, its not actually as bad as not taking the risk at all.

        • Well, the Fukushima evacuation did kill 45 people though through exhaustion, disruption of medical care, etc. And the number of expected cancer deaths is a matter of dispute. Still, even a plant that Fukushima that exploded isn’t that much above a median world coal plant in terms of deaths per megawatt hour generated.

          • Gbdub says:

            Can’t find the article at the moment, but I remember reading that the Fukishima evacuation demonstrably killed more people than staying put would have been statistically expected to. (Then again I think that only accounted for the radiation that was actually released, at the time there may have been an unknown risk of a larger more dangerous release of radiation)

          • Douglas Knight says:

            Fukushima may have only killed 45 Japanese, but it killed 100x as many Germans.

          • onyomi says:

            There’s also the issue of rendering parts of a country without a lot of habitable land uninhabitable (or habitable*) for who knows how long.

            *But I wouldn’t want to live there because I’d worry, rightly or wrongly, that I’m subtly increasing my cancer risk. Then again, lots of people continue to live in Beijing despite the coal mine quality air.

        • wintermute says:

          Bear in mind that after the chain of bad luck at Chernobyl, we actually got off pretty lightly. If the (risky, suicidal) effort to open the drainage valves had failed, Chernobyl itself might have been 10x or 100x worse.

          Of course, even spotting those numbers nuclear power looks pretty good. Chernobyl-bad disasters are substantially less likely than they used to be, and even a 100x catastrophe would only have matched a decade of coal deaths.

        • Diadem says:

          The “Chernobyl was badly designed” argument I think is a very weak one. Chernobyl was designed to be extra safe. The designed contained fatal flaws, true, but who’s to say that our new ‘safe’ designs don’t contain fatal flaws?

          I mean, sure, if no one makes any mistakes, then no very serious disasters are going to happen. But the worst case scenario is always a scenario where someone does screw up, horribly.

          And Chernobyl wasn’t a worst case scenario either. It wasn’t a full meltdown. They came close to one, but they managed to avoid that. The disaster would have been far worse if it had been a full meltdown.

          That all being said: In a comparison between nuclear and coal or oil, nuclear wins by a huge margin. Even if our modern designs are far worse than we think, and even if the average nuclear disaster is far worse than disasters so far have been.

          We need nuclear power. Solar and other green sources simply won’t cut it.

          • John Schilling says:

            Chernobyl was designed to be extra safe.

            A nuclear reactor designed to be extra safe, would have complete primary and secondary containment structures. The RMBK-1000 has only an incomplete primary containment. It also has a core made of coalgraphite, which is about as “safe” as it sounds for a thing that is supposed to get hot but never burn.

            Chernobyl was designed to be infallible. That never works. Things that are designed to be safe, should be designed to fail rarely but must be designed to tolerate failure. Neither Chernobyl nor any other RMBK-1000 nuclear reactor was designed to be “extra safe”.

          • bean says:

            Chernobyl was designed to be extra safe.

            What are you talking about? I’m not suggesting that the designers set about with ‘lack of safety’ as a goal, but ‘designed to be extra safe’ implies that safety was a really special consideration during the design process. It wasn’t. Online refueling for plutonium production was. This lead to a design with positive temperature and void reactivity coefficients, which would be unacceptable in the west, at least as implemented. (Yes, the CANDU does have a positive void coefficient, but it’s much smaller and the reactor doesn’t run as hot.)

          • Mary says:

            If you design things to be correct all the time and completely neglect robustness (coping when things go wrong), you will discover that the universe knows a lot more different times than you thought it did.

          • If you design things to be correct all the time and completely neglect robustness (coping when things go wrong), you will discover that the universe knows a lot more different times than you thought it did.

            I think this is true, but I can also see why people would want to wash their hands of the whole deal.

            Maybe there’s always some Final Destination-esque danger lurking in the background, which breaks through all your design systems and ends up killing a million people.

            Then you’re always patching the last Final Destination disaster, the experts tell you “we’re finally safe,” only for another crazy sequence of events in 20 years.

      • Bram Cohen says:

        It’s a misconception that a malfunctioning nuclear power plant could produce a nuclear explosion. What it can possibly do is irradiate the immediate vicinity and release a bunch of radioactive isotopes into the environment, most problematically Iodine-131. While that could be very bad, is several orders of magnitude less disastrous than people assume. Also while liquid fluoride thorium reactors have a number of very deep inherent safety features, and are a design I advocate for, all modern nuclear plants are much better than the older ones. A happenstance benefit of thorium reactors is that they hold onto Iodine well instead of it escaping into the atmosphere, which is a nice extra safety mechanism on top of the more inherent ones.

    • wintermute says:

      This does raise the question of self-disrupting probabilities.

      There’s no dam in the US that could reproduce the (direct) Banqiao death toll – I don’t think even catastrophic failure of the Hoover would approach that. And part of the reason is that people have learned not to build dams such that catastrophic failure would kill huge numbers of people.

      Pointing to the previous WTC attack feels a bit fuzzier – it was the same target, but the security measures to prevent each incident don’t overlap. Worse, terrorism is ‘anti-inductive’ because it comes with an opposing player. Attempts to learn from past instances have to contend with attackers trying to invalidate those lessons. So there’s some validity in saying “all we can expect is that we won’t expect the next attack”.

      • Bram Cohen says:

        I’m not saying that there aren’t good engineering countermeasures for both dams and buildings! Buildings in particular can (and now are) made with the structural supports encased in concrete, so the thermal insulation is much harder to remove and they can survive an airplane flying into them (other than the floor or two which are taken out directly, of course).

        • CatCube says:

          Encasing structural steel members in concrete for fire protection is something that used to be done in the 30s (the Empire State Building, for example). It’s not generally done now. If you’re going to add that much dead weight, easier to just go ahead and design the structure as reinforced concrete.

          And “taking out a floor or two” is a major contributor to the collapse of the WTC–the floor diaphragm was bracing the exterior columns to reduce their effective length and prevent buckling.

          • Bram Cohen says:

            Reinforced concrete is what I meant. The world trade center in Beijing, for one, is built with concrete, as a direct result of worries about WTC-style attacks, and I believe that’s increasingly the trend these days.

        • Titanium Dragon says:

          TBH, I’m skeptical that any tall building like that can be made to survive 9/11. The problem is that the uncontrolled fires above the height of hoses combined with probable loss of water pressure in the sprinkler system means that most likely you’re going to end up with the building on uncontrollable fire for a long time, which combined with the structural damage is going to probably trigger progressive collapse of the structure sooner or later.

          There was enough time on 9/11 to evacuate the buildings; making it so that the building collapsed after 10 hours instead wouldn’t have made much of a difference.

    • albatross says:

      Is there some reasons the WTC was uniquely vulnerable to terrorist attack, relative to other large buildings?

      • Alex Zavoluk says:

        Possible reasons Bram could be referring to:

        It was a private civilian complex, not a military base or government building, which would likely have more security.

        It had lots of people and practical value, but was also heavily symbolic (ie not a monument and not just any random office building).

        It was very important for the economy.

        It was very difficult to get emergency responders high enough up on the building.

      • Bram Cohen says:

        As another commenter said it’s extremely large, expensive, and symbolic. But it also had the problem that heat insulation of the structural supports was done with spray-on thermal insulation. Since then gigantic buildings have gone back to being built with concrete and rebar and have being able to survive an airplane fully loaded with fuel flying into them as one of their core design criteria.

    • Titanium Dragon says:

      You’re forgetting the Oklahoma City Bombing, which killed well over 100 people. Different group, sure, but we shouldn’t forget that 9/11 wasn’t even the only major US terrorist attack within that decade time span.

  17. Markus Ramikin says:

    For a problem in which outliers is the only thing that matters, I’d take the last two major outliers, and count from the day after the earlier one, to the day of the latest one. Ignore the time from last outlier to now. Wouldn’t this be is a fair, non-maximised way to measure the “average” impact of the last outlier? Extend to several to get a less random “average”.

  18. Incurian says:

    After reading the previous post, and re-reading “All Debates are Bravery Debates,” I started to think about how a lot of disagreement boils down to whether you’re concerned about the average case or the extreme case. Often there are unstated assumptions that one is more important than the other, and I think figuring out the relative importance of the average and extreme cases would be beneficial to most arguments.

    • Gbdub says:

      That’s a very interesting point, although I’m not really sure every case has someone arguing for the average side. More often we see people arguing about the extremes one way or another and the average gets ignored.

      E.g. Gun control – the most common gun related deaths are people intentionally shooting themselves and criminals shooting other criminals in ones and twos, mostly with handguns. But the hottest debate surrounds comparatively very rare mass shootings and the semiautomatic rifles often used in them.

      E.g. Police violence – we’re throwing riots over police shootings of black people as evidence of racism, but it looks like shootings are much less racially disproportionate than other, lower level police use of force.

      Certainly these aren’t original points on my part – there are certainly discussions around the “average” events in both examples, but that’s not where most of the mainstream back and forth is focused.

      • Incurian says:

        I think you are correct that these debates may often center around the extreme cases. What I meant to say was that regardless of what gets the most attention, the positions held by each side in a debate might be broken down into average vs extreme. I also think in many cases it’s not clear which is which, and it’s almost never stated explicitly.

        To continue with your gun control example, one side focuses on the extremes of mass shootings while the other side focuses on the average case of responsible gun owners.

        Or perhaps another way of looking at it is that one side does not acknowledge (or is unaware) of the average case of responsible gun owners and believes that mass shootings ARE the average case. If that’s so, then establishing which is average and which is extreme is probably very important to the debate – but because both sides have unstated assumptions about what is extreme and what is average they tend to talk past each other.

        Or maybe they both may correctly identify the average and extreme cases, but don’t acknowledge the relative significance they assign to each. “The small number of tragedies are outweighed by the many benefits of an armed citizenry” versus “these egregious tragedies are too high a price to pay for the supposed benefits of an armed citizenry.” Often both sides refuse to concede the other side has any kind of point, when really they ought to broadly agree on the facts (where they are clear) while disagreeing about the relative weighting of the externalities.

        To put it another way, these kinds of debates ought to focus on the frequency and magnitude of X, and a cost/benefit analysis of solutions for X. Probably the reason they don’t is that both sides suspect the other side is arguing in bad faith and that any concession will doom them.

      • “we’re throwing riots over police shootings of black people as evidence of racism, but it looks like shootings are much less racially disproportionate than other, lower level police use of force.”

        Race is a red herring here. Even if police shootings had no racial bias at all, we would still have a huge problem — American police kill far more people, on a per capita basis, than police in any other developed country. I don’t remember the exact numbers, but it wasn’t 20% more, or 60% more, it was several times, maybe 10 times more, than what is typical for European countries.

        • John Schilling says:

          American criminals shoot policemen more often than in any other developed country as well – and it’s not because they are responding to unprovoked attacks by policemen.

          As has been noted here repeatedly in the past, the United States is home to several extremely violent demographic groups that are largely absent in other “developed nations”. In at least one case because friendly, peace-loving Europeans encouraged them to go settle in America, Or Else. One can debate whether it is a bug or a feature, but it isn’t likely to change any time soon. This is where the violent-tempered people of the developed world wind up living.

          If you see it as a problem, your choices are to go somewhere else, to convince tens of millions of violent people embedded in violent subcultures to go somewhere else, or find a way to turn those cultures peaceful. And we’ve had lots of smart people trying that third option over the past century or two; so far they’ve all failed.

  19. K says:

    I think in general, the cognitive bias here is worrying too much about very rare events with spectacular consequences.

    One case in point is nuclear power. It seems evident to me (data/elaboration available on request) that, assuming IPCC is roughly correct, the only way to avert the looming climate crisis is a large scale global deployment of nuclear. Yet, few seem willing to go down that road. If you look at the history of nuclear power, there has been exactly one accident where people have died from radiation – in contrast to e.g. medical use of radiation, where accidents have killed a number of people, and nobody cares much. The question is how much statistical weight we should give Chernobyl – and in general, we can’t really estimate probabilities from a single event like that. You could of course look at current reactor designs and operating procedures and realize they are much, much safer, but most people aren’t going to bother. Instead they will attempt to avert climate change by growing their own tomatoes in their window sill, and blame somebody else when things go to hell.

    • Chris H says:

      Since you’re offering, I at least would like elaboration on the necessity of nuclear power to contain climate change. What target are you going for and what assumptions are you making? Thanks!

      • Edward Scizorhands says:

        1. People will not accept significant impacts on their lifestyles. You can say that they should or they will have to, but they won’t do it unless you force them at gunpoint.

        2. It works. There’s not “studies show that…” or “scientists predict that…” or “experts agree…” Nuclear has been demonstrated to work, and can produce a lot of grid power. Even if it’s just baseline power, that gets you well over 50%.

    • Callum G says:

      Nuclear would be a great option for reducing the GHG from power generation, especially because it would be less disruptive to the way we live compared to other methods. Also, it would get rid of coal which WHO reckons kills up to a million people each year(!). I think the explicit association with of nuclear with weapons paired with the thought of an explosion causes a lot of the paranoia.

      There are a few drawbacks to nuclear though. To start with we might now have enough economically recoverable reserves to cover the change. With current technology the Nuclear Energy Association estimates that we have about 200 years worth of resources at current consumption. Worldwide nuclear adoption would have to increase by a factor of 30 to cover fossil fuels so the reserves mightn’t be enough. There’s also issues with the long time it takes for a new plant to move through planning to operation as well as the massive cost of decommissioning plants. Gen IV reactors are more likely to be the ‘saviour’ if humanity is to go down this route.

      • Jiro says:

        “Reserves” are a stupid way of measuring how much fuel we have. The trick is that it usually means “how many sources do we know about”, but people stop looking for sources when the existing sources seem like they’ll last a while. “Reserves” are rarely an estimate of how much we actually have remaining in the ground including estimates for sources that will be found when people start looking again.

        • Douglas Knight says:

          Callum said “resources,” not “reserves.” But the answer to that question is at least billions of years.

          • Callum G says:

            If all reactors are fast breeder reactors and if all uranium can be extracted from all seawater. At the moment seawater extraction is still quite expensive and experimental, and even if that changed I imagine accessing the seawater in the bottom of the Mariana Trench would prove difficult.

            I’d like to see more breeder reactors but alongside renewable energies as they can be cheaper and more readily deployed. Either would require more government intervention at this point.

      • roystgnr says:

        Is it safe to guess that WHO is including poisoning from home coal stove fumes among those deaths? If so then any centralized power supply would fix the problem, not just nuclear.

        Hmm… looks like indoor air pollution deaths indeed make up more than half of the WHO estimate, but not the ~90% I was expecting. I still suspect that replacing personal coal use by even centralized coal power plants (with more complete burning, scrubbers, etc) would make a huge dent in the remainder. Nuclear would be even better, but might not be as feasible in the short term in areas which are still so poor that an indoor coal-burning stove is a popular idea.

      • gbdub says:

        Reserves would be higher if we go to breeder reactors that reprocess used fuel – the reasons we don’t are mostly political rather than technical.

        Reprocessing would also reduce decommissioning and waste management costs.

        One of the biggest expenses is the assumption that we need to store waste in something that can reliably contain for like 10,000 years. Why? In a hundred years we will almost certainly have a better way to contain or eliminate waste, or we’ll be dead and won’t care, so build a 100-year storage facility and monitor it for leaks.

        • Callum G says:

          Interesting point with the storage, especially considering at the moment it’s a massive cost with the likes of yucca mountain. A lot of infrastructure surrounding nuclear (actually surrounding power plants for the matter) suffers from NIMBY and that’s a great barrier to change. I’m not sure many people would want less secure nuclear storage in their town/county/state.

    • Maware says:

      Uh, of course, there’s really very few nuclear power stations at all. There’s only 100 in the USA. The question is when you start increasing that number and rolling them out on a large scale, and if we can maintain the same safety standards without the particular fear people have of nuclear energy. Especially considering the levels of corruption many countries have, and the shoddiness of infrastructure, and the massive instability of elected government.

      • Garrett says:

        There’s been a lot of research and design work on nuclear power plants in the past 40 years or so, such as the AP1000 reactor.
        Thus the available safety standards for new power plants would increase in comparison to the older plants.

    • Douglas Knight says:

      exactly one accident where people have died from radiation

      Where do you get your numbers? There were 3 separate incidents in Japan in the 90s where power plant workers died from radiation poising.

      • bean says:

        Doing industry is inherently dangerous, and in the US at least, the nuclear industry is the most safety-conscious industry around, period. How many people died in the US during that period? How many died at other power plants? How many died from falling off of roofs while installing solar panels?

      • K says:

        You’re absolutely right – I just went over the list on Wikipedia of radiation accident, and included only accidents with five or more fatalities. Of those, Chernobyl is the only power station accident, the rest are (mostly) medical, and (some) nuclear subs or weapons production. Sorry about the inaccuracy.

    • Titanium Dragon says:

      The problem is that we don’t really want non-developed countries to have nuclear reactors, because that would allow them to build nuclear weapons. This is an entirely sensible concern.

      That said, the developed world universally going over to nuclear power would be a good thing.

  20. Albatross says:

    For several reasons, I like both graphs and for these kinds of events I think you use the the day before and the day after the extreme example to understand the numbers.

    The first graph, with 23 deaths a year gives us a good average. For example when more than fifty people died in Orlando it tells me that terrorism might be on the rise. And has been correctly pointed out, there is much more variability in the terrorism numbers and much more effort being spent on reducing terrorism compared to chair deaths.

    This is where the second graph, with its 3,017 tells us that terrorism is fundamentally different than chair deaths. This tells me we probably never should have compared terrorism to chair deaths in the first place because lots of people die from chairs every year but there are years with zero terror deaths on US soil. Similarly there are no recorded years with 3,000+ chair deaths.

    So “look at chair deaths” would have lots of meaning when stacked against “deaths from cars backing up – around 200 a year, 17,000 injured.” But chair deaths versus terrorism deaths isn’t a fair comparison in that OSHA can’t get through the year with nobody dying from chairs, but they have mitigated it pretty well.

    Thus the two graphs tell the story, just like the flu and earthquake two number sets tell a story. In fact, looking at both sets we can make a strong case for flu vaccines and earthquake resistant architecture. But what should the ratio of spending be? Comparing flu deaths to earthquake deaths gives us interesting numbers.

    My point is this – saying there have been zero world wars in the last fifty years OR saying there have been two in the last hundred years are both a bit misleading. The proper way to describe it, I think, is to say it has been about 70 years since the last world war AND to say there have been two in the last hundred years. And it doesn’t make any sense to compare world wars to chair deaths. Instead we can just look at them in a vacuum – just like terrorism. Why? Well, if you compare world war deaths and flu deaths then it looks like flu deserves more money and attention. But the correct answer is that we don’t want a world war OR the flu and so how much we spend avoiding shouldn’t be determined by the comparison. Boosting flu prevention to 2% of GDP might not be cost effective, and flu budget sized military budgets might not prevent world war three. They are, like chairs and terrorism, apples and oranges.

    Even comparing similar data sets like war deaths to dictator genocide is problematic. I mean at some point Ukraine was invaded because of a lack of military spending, but Assad is a problem in part due to excessive military spending. Both outcomes are undesirable. But than telling us that unchecked military power is bad, and having a military big enough to defend your country is good the stats don’t really tell us what those numbers should be. Just because the flu and malaria kill different numbers of people doesn’t mean we shouldn’t prevent both of them. Aside from our personal charitable donations, it simply isn’t necessary to rank these things… consider:

    1. Chairs. 115 people. Despite several chairs per capita. OSHA seems an appropriate level cost benefit analysis when we consider other workplace hazards like farm equipment. And we kinda already have OSHA in place for the more dangerous events. So reducing chair deaths is really mostly sunk cost in general workplace safety.

    2. Terrorism. 23/year, up to 3,017. Ok, maybe we shouldn’t go to war over that. What would be an appropriate response? 150,000 more diplomats? Hmm. Well, the diplomats have a lot more benefits than just terrorism – they also improve trade and help human rights. Worth it.

    3. Earthquakes – worth building codes, especially when comparing cities hit with equal level quakes. The earthquakes didn’t kill most of those Haitians. Terrible building standards did.

    4. War. The UN stopped counting after 250,000 people were killed in Syria in 2010. After the Iraq War, the UN estimate of 30,000 was eventually tallied at somewhere near 1,000,000 deaths. Based on migration patterns and similar conflicts, I suspect there have been 3 million people killed in the Syria region, 7 million wounded and roughly 40 million refugees. Here the difference between reasonable estimates and the UNs count ought to drive pursuit of better information. But when people wring their hands in ten years saying nobody knew about the genocide in Syria, I would say the UN deliberately looked the other way. Invading Syria with four million troops sounds extreme and would cost a fortune. But it might save a million lives even if the conflict results in hundreds of thousands of war dead.

    5. The Flu. This one the cost benefit is very easy. Get a vaccine.

    6. Genocide. See also Terrorism and War. Here we can see why it might be a good idea to build a coalition instead of having one nation build up to four million troops. This is also a good argument against having Iran, Turkey or the Kurds do all the fighting. One of those militaries getting sufficiently large to finish the job in Syria could result in problems for their own citizens later on.

    • cassander says:

      >4. War. After the Iraq War, the UN estimate of 30,000 was eventually tallied at somewhere near 1,000,000 deaths

      this is incorrect. A few studies that did not tallie deaths, but made elaborate assumptions based on interviews got numbers that high. In terms of bodies found, the number is around a quarter million.

  21. a n o n says:

    Maybe you’re worried about terrorists nuking London, but that’s not what the Washington Post is answering to. People are worried about being shot by a terrorist when they get out of their home.
    Also, the fact that even including 9/11 makes deaths from terrorism only 30 times more abundant than deaths from furniture is still an argument in favor of the Washington Post’s stance.

  22. Macbi says:

    One lesson to draw from this is that anti-terrorism efforts should focus on preventing the tail risks. Securing the cabin doors on aeroplanes was an excellent idea because it prevents people from crashing planes into buildings. But screening people’s luggage only prevents them from destroying the plane itself. Clearly it’s bad for a plane to be destroyed, but since it happens so rarely worrying about it is like worrying about chairs.

    (I’m talking from a western perspective here. Things are different if you live in Iraq or Nigeria.)

    • Alphaceph says:


      At the moment I don’t expect terrorists to be well-organised enough to do anything really dangerous, and our intelligence agencies have the advantage when terrorists try to communicate with each other, which correlates with the more high impact attempts.

      • gbdub says:

        Then again, there have been at least a couple known attempts to bring down airliners (and probably a successful one in that plane full of Russian vacationers) in the past decade (e.g. the shoe bomber and underwear bomber).

        If those are successful, they kill probably ~300 people each, and one or two of those working every year starts to make air travel look pretty scary.

        • Alphaceph says:

          600 people per year is bad, but not that bad.

          “Tail risks” IMO covers the really big stuff, such as terrorists getting something into the water supply.

  23. Tekhno says:

    I’m not worried about terrorism itself much, so much as the reaction to it.

  24. Alphaceph says:

    What is a reasonable estimate for the chance of an Islamic terror organisation pulling off > 10,000 deaths in the developed world over the next 10 years in a single incident?

    It seems like it’s rather low in the next 10 years. They’re not competent enough?

    If I bet against that at 20:1 against would anyone take the other side?

    • Salem says:

      20:1 has implied odds of 4.7%. That implies that we ‘average’ at least 47 deaths/year in Western countries from Islamic terrorist action on an unprecedented scale – and presumably far more, as 10k is a minimum not a cap.

      • Alphaceph says:

        > 10k is a minimum not a cap.

        yeah, let’s say I assign a 3% probability to 10,000-50,000 deaths, 1% to 50,000-100,000 and the remaining 1% trailing off with a fixed expectation per order of magnitude until you hit 7 billion.

        Still, will anyone take the other side on any of those? I’m trying to get a feeling for reasonable probability distributions.

        • Mengsk says:

          We can probably make this a little more specific. The only way i can think of for a terrorist to kill more than 10,000 people at once is with a nuclear or biological weapon. You could probably talk about the odds of either of these things happening and multiply by the likelihood of either of these things killing less than 10,000 people

          • Alphaceph says:

            > The only way i can think of for a terrorist to kill more than 10,000 people at once is with a nuclear or biological weapon.

            – An attack on a sports stadium with significant numbers of automatic small arms or explosives might achieve 10k dead

            – bringing down several packed skyscrapers,

            – chemical in the water supply (not biological – merely a chemical poison).

            – mass drone attacks, e.g. to create a firestorm

            – a mass attack with many terrorists with small arms on a particular city, such that the police are temporarily outnumbered

            – infrastructure destruction that devastates the economy, e.g. taking out all the power stations in a country could easily lead to thousands of deaths in the resulting chaos

          • throwaway says:

            “mass drone attacks, e.g. to create a firestorm” – enough drones to cause firestorm?


            “based on World War II experience with mass fires resulting from air raids on Germany and Japan, the minimum requirements for a firestorm to develop are considered by some authorities to be the following: (1) at least 8 pounds of combustibles per square foot of fire area (40 kg per square meter), (2) at least half of the structures in the area on fire simultaneously, (3) a wind of less than 8 miles per hour at the time, and (4) a minimum burning area of about half a square mile.”

            40 kg per square meter if far too high to be doable with drones.

      • Alphaceph says:

        > That implies that we ‘average’ at least 47 deaths/year in Western countries from Islamic terrorist action on an unprecedented scale

        9/11 already exceeds that. What my betting odds imply is that 9/11 involved unusually competent adversaries who were a bit of a one-off. ISIS has not been particularly effective in its attacks, despite the easy ride we have been giving them.

        • Salem says:

          But this is separate to 9/11 stuff, which “only” killed 3000. Your numbers suggest that in addition to the deaths we observe from terrorist attacks large and small, there are another >50 deaths/year in Western countries from as-yet-unobserved very-high-scale terrorist action.

          In other words, there is an invisible terrorist attack every year, as bad as the 2005 London Bombings, and maybe much worse.

          • Alphaceph says:

            I think the penalty term for a 10k death event every 100 years should be higher than a 100 death event every year actually. Look at 9/11 and the bad decision that followed.

            Spectacle is definitely something terrorists should aim for, but its hard to pull off.

  25. Rob says:

    This is why Taleb is always getting angry about people using stats like that – I think “the empiricism of idiots” is the term he uses.

    He tries to expand it to say that violent death may not truly be declining through modern history because we’ve just shifted it to a couple of huge black swan risks of nuclear war.

    Regardless, if it was the black swans we were truly worried about, it would suggest that we should shift our focus of anti-terrorism spend. More on securing nuclear storage sights around the world, less on airport screening.

    • Albatross says:

      Makes sense. This is one reason I appreciate the Iran deal. The US supports a lot of horrible regimes like Turkey, Saudi Arabi and China. At least with the Iran deal we were able to establish an environment with some accountability in terms of nuclear security.

      • ThirteenthLetter says:

        At least with the Iran deal we were able to establish an environment with some accountability in terms of nuclear security

        What makes you believe there is any accountability here? If tomorrow Iran announced that due to American provocation they were tearing up the deal, do you really think there would be any consequences?

  26. utilitarian troll says:

    I’m in favor of rationalists promoting concern for terrorism because I expect a lot of measures that help guard against terrorist attacks also help guard against existential risks posed by small groups of malcontents. It’s easier to talk about terrorism than existential risk because it’s a concept the public is already familiar with. This strategy isn’t dishonest either–just talk about “black swan terrorist attacks” in the form of terrorists e.g. engineering some kind of deadly virus and killing millions or billions. Yes, I do believe this concern plausibly justifies a light global surveillance state in the form of the sort of stuff Edward Snowden complained about.

  27. B says:

    Yeah, the nuke argument is the standard answer to the “terrorism is a trivial problem” argument, remember Sam Harris making it in the Joe Rogan podcast at some point (though still think that while it is a problem worth spending time into, a lot of people blow it at least an order of magnitude out of proportion).

  28. Link Dumper says:

    Your flu death numbers are really unfair. Looking at a within-country plot of flu deaths you see 1918 was only 4-6x as bad as usual.
    Info on contagious disease death
    Flu deaths in the US
    Worldwide there are 250k to 500k deaths per year from epidemics, so 1 mil/year from rare events isn’t as big a change. Admittedly, most flu deaths in developed countries are old people, who are pretty fragile to start, so maybe epidemics are more worrying if they kill young people.

    It would be cool to see analysis on the distribution of earthquake deaths. Haiti’s the biggest, but still typical of big earthquakes. Big earthquakes, plus their resulting tsunamis and fires, are most of the deaths. Doubleunplussed, want to try?
    Worst earthquakes
    Nb, poor countries suffer more from disasters
    Poverty and natural disaster death

    A somewhat interesting post overall, and I agree that meteor strikes, financial collapse, AI, and nuclear terrorism are still potential problems.

    • Murphy says:

      I’d be a bit cautious about the GA link. I’ve found a fair few of their articles to be simply wrong. I’m not saying those graphs are incorrect… just to double check the numbers before basing statements off them.

      While some of GlobalResearch’s articles discuss legitimate humanitarian concerns, its view of science, economics, and geopolitics is conspiracist. The site has long been a crank magnet: If you disagree with “Western” sources on 9/11, or HAARP, or vaccines, or H1N1, or climate change, or anything published by the “mainstream” media, then GlobalResearch is guaranteed to have a page you will love.

  29. Wency says:

    We are doing the vast majority of what we can to manage the risk of chair death. Every time your parent or teacher told you, “Don’t lean back in the chair, you could crack your head open,” a measure was taken to reduce chair death. Also, every time someone fell off a chair due to their own stupidity but didn’t die, a lesson was learned and a possible future chair death prevented. People are constantly doing things every moment of every day to reduce the risk of death by furniture. Managing chair death is largely a personal matter. As I write this, I’m sitting properly in a comfortable, well-built chair.

    If chair death is something you worry about, you can take even more measures to minimize your personal risk while still living a mostly normal life. Get a standing desk. Sit on the floor when you eat dinner. Chairs, being inanimate objects, do not actively try to circumvent the measures we take to reduce chair death.

    Terrorists, on the other hand, are a community problem — there’s not much we can do to manage terrorism as individuals without, say, avoiding life in a major city, which may involve major social and career disruption for us. Also, terrorists actively work to circumvent the countermeasures we put in place, so there is an inevitable arms race of sorts, and there is always going to be an inclination to discuss how we manage that arms race.

    • Rosemary7391 says:

      I think people do tend to worry more about things they can’t control. As you say, chair death is trivially easy for me personally to reduce the risk of, mostly by one time decisions (eg putting the thing together properly rather than quickly, or repairing it when broken). How can I reduce my risk of falling subject to terrorism? Or my loved ones? Realistically, very little. I can sneakily fix my friend’s chair, I’m unlikely to convince him to go live in the middle of nowhere with me. (Incidentally, I wonder which of harm to self or harm to loved ones people fear most? If the latter, that’s another reason people fear it more, even more out of their control).

      As such, people are going to talk about terrorism more and want “society” to be actively doing something to reduce that risk. One time decisions (akin to fixing the chair) won’t work when someone is constantly trying to break your chair.

  30. David Pinto says:

    Thanks you for this article.

  31. CorporateLawyer says:

    I’ve always thought the biggest difference between terrorism and natural disasters or flu or chairs is that terrorism is an indication people want to kill you! The “don’t worry about terrorism because its statistically unlikely you’ll die” is basically saying yes, people want to kill you, they just don’t have the capability to do so. But how reassuring is that? If a guy in prison writes letters about how he is going to murder your family, do you feel okay because he’s in prison and can’t harm you? Or do you worry that maybe he will motivate someone else to kill you?

    It’s the same reason why we worry about small geopolitical conflicts. If China sunk one submarine in the Pacific, we wouldn’t discount that because only a few dozen people died. We would freak out because it would be an indication that even more conflicts is to come in the future .

    I find it funny that Bryan Caplan, who is all about signaling in higher education, fails to miss the point with terrorism. Terrorism is a signal that people want to hurt you, and that they would kill you if they had the chance. You need to take these threats more seriously than natural disasters because people’s capabilities change. The U.S.’s could weaken, ISIS could strengthen. Nothing is preordained

    Wait sorry, one thing is preordained: people love to garner status by lecturing others how stupid they are to be focusing on terrorism instead of chair deaths, and how enlightened the lecturer is to be moving beyond such simple human cognitive biases. That is preordained.

    • Desertopa says:

      If a guy in prison writes letters about how he is going to murder your family, do you feel okay because he’s in prison and can’t harm you? Or do you worry that maybe he will motivate someone else to kill you?

      I think that reasonably depends on the context. If the guy is, say, a mob boss, then of course you should be worried. But if the guy is an impulsive idiot, and is openly writing letters about how he wants to kill you as a manifestation of his usual level of caution and forethought, then his being in jail is a perfectly adequate security mechanism and he’s likely to remain there a long time.

      On a societal level, the prospect of China deciding on open conflict with us is scary. The prospect of a tiny cult with thirteen members and thirty grand in assets deciding on open conflict with us is not very scary. Spending lots of money to hedging against a conflict with China is fairly intelligent, spending lots of money hedging against a conflict with the tiny cult isn’t.

      Terrorist organizations fall somewhere in between (on an absolute, rather than log scale, certainly much closer to the tiny cult than to China.) Yes, it’s true, their capabilities could change. But their abilities are highly unlikely to change suddenly without us noticing, and money spent on security isn’t the same as money spent trying to limit their growth (empirically money spent trying to limit their growth hasn’t offered very good returns so far.) Besides, money spent on security is to a large extent predicated on the assumption of a certain level of resources anyway. Investments in measures to prevent people, say, leaving bombs in crowded areas and detonating them, are not going to help stop people firing an ICBM at us. If the people who you once worried about planting improvised explosives become capable of launching ICBMs, the money you spent securing yourself against improvised explosives isn’t an investment into the now-salient endeavor of preventing them from firing them at you.

      • Jiro says:

        on an absolute, rather than log scale, certainly much closer to the tiny cult than to China

        How is using an absolute scale instead of a log scale a remotely sensible thing to do here? If you’re using an absolute scale, terrorists could kill 49% of the number of people that could be killed in a war with China and they would still be closer to the tiny cult than to China.

        • Desertopa says:

          Having less than half the resources at their disposal of China, an open conflict with them is probably going to result in a lot less than half the damage of an open conflict with China. The severity of the conflict doesn’t scale linearly with relative power.

      • Liskantope says:

        I think that reasonably depends on the context. If the guy is, say, a mob boss, then of course you should be worried. But if the guy is an impulsive idiot, and is openly writing letters about how he wants to kill you as a manifestation of his usual level of caution and forethought, then his being in jail is a perfectly adequate security mechanism and he’s likely to remain there a long time.

        Speaking as someone who was once in a very similar situation, I can say that this makes sense on an intuitive level to me. When I was a teenager, another teenager I knew (who fit the description of “compulsive idiot” much better than “mob boss”) was in prison for a while and wrote a description of having killed me in a letter. I remember feeling not terribly worried at the time. On the other hand, it made my parents very stressed, but I think that was mainly due to their fears of him getting out (some 15 years later, that guy’s still in prison but I’m long gone from where we grew up).

        From a rationalist point of view, situations like this aren’t very dangerous, even while it can be extremely unsettling to know that someone is actively imagining killing you. But I think the main reason why terrorism garners what seems like a disproportionate amount of concern is simply that a major terrorist attack is upsetting in a spectacular way: there is something much scarier about the idea of 100 people being killed by one action of a smaller group with no way of knowing how to avoid their fates than, say, 100 people being killed in separate car accidents within wider area on a particular day.

    • Edward Scizorhands says:

      Agency is really important.

      People get much more upset if they found out I took $20 from their wallet than if they found out they dropped $20 and it blew away.

      • ThirteenthLetter says:

        Exactly right. Who was it who said that even a dog knows the difference between being stumbled over and being kicked?

        Even aside from the highly valid point that, unlike a poorly built chair, someone attempting to kill you is not going to give up after one failure, arguments of the form “don’t worry about terrorism, chairs kill way more people!” are doomed to failure because they boil down to “stop being human beings, guys!”

        • Doctor Mist says:

          We are hardwired to detect and punish defectors, of which a terrorist is an extreme example. I sometimes wonder if that wiring is actually counterproductive in the modern world, where our fate is so thoroughly tied up with complete strangers. If our natural reaction to a terrorist event was “In other news, 30 people died tragically today when one of them exploded” would there be more suicide bombers? Fewer? Any?

          • ThirteenthLetter says:

            Maybe there would and maybe there wouldn’t, but you might as well posit that human beings wouldn’t gossip, or be self-interested, or get distracted, or enjoy music; it’s equally likely to happen.

          • Jiro says:

            Just like regular people usually aren’t rational, neither are terrorists. Terrorists generally aren’t very good at reaching their stated goals, but they’re pretty good at fanaticism, feeling important, self-sacrifice, signalling to co-religionists, gaining power over members of their own group by instilling values in them that lead to those other things, and generally all sorts of other social things that humans do that don’t achieve their stated goals. Killing yourself to destroy the infidels is looked on as praiseworthy regardless of whether this has any effect towards actually destroying the infidels as a nation.

            I would expect that if nobody reacted to suicide bombing, there would be no differences whatsoever in how often terrorists use suicide bombers.

          • Doctor Mist says:

            you might as well posit that human beings wouldn’t gossip, …

            Heh. Of course. I wasn’t advocating policy, just idly musing.

        • Simon says:

          The internet concludes that it was said by Oliver Wendell Holmes. Good phrase.

  32. Desertopa says:

    With regards to the friend afraid of terrorists stealing a nuke…

    This isn’t where the real danger lies. Not because there’s some specific doomsday scenario which is much more destructive, but because the big danger lies in scenarios which our security forces haven’t already thought of and made preparations against. Paranoia against people trying to steal nukes is built into our security infrastructure, in a way that, before 9/11, paranoia against people hijacking planes and flying them into national monuments wasn’t.

    But this is one of the reasons that anti-terrorism security doesn’t make for a very good investment. I have a particularly devious and morbid imagination, and I’ve had a few people suggest it would be a good idea for me to work for the Department of Homeland Security. But this is a terrible idea! If you’re constantly trying to guard against all the possible methods that could be used by the most devious people, you’re reducing the risk of those black swan events, but this is way more expensive than investing in something like “make nukes safer,” and the vast majority of these defenses will probably never be used. Say a terrorist measure can kill 300,000 people, and there’s a 0.1% chance of them ever trying it in the next century, and a 50% chance of it working if they try. If we spend 50 million dollars a year combating it, we could bring the chance of them successfully pulling it off down to .1%. That’s 50 million dollars a year saving approximately one and a half people a year. The chance of a massively destructive terrorist attack isn’t concentrated in something obvious like “stealing nukes,” it’s distributed across the space of a large number of possible plots, all of which are individually highly improbable. That’s why trying to safeguard against these kinds of events is almost certainly a terrible investment.

    At least naturalistic outlier events aren’t anti-inductive. If you spend resources preparing against earthquakes, nature isn’t going to hit you with a plague instead to avoid wasting resources on a preemptively foiled attack.

    • Maware says:

      Now apply this to AI Risk, which this community obsesses over and which doesn’t have a single case of actually existing.

      • Desertopa says:

        Sometimes you have to work from models, rather than frequency analyses. There’s no shortage of society-shaping events that have occurred in the last hundred years which never occurred before. Our best models of terrorism suggest that it’s more dangerous than a frequency analysis of terrorism over the last decade (because its highest impact is in outlier events,) and possibly more dangerous than a frequency analysis including those outlier events (because terrorists are trying for maximum impact, and if they can pull off a plot to do much more damage than they have with the previous attacks, they’ll take it,) but we don’t have a model under which terrorism being civilization-threatening is realistic.

        If we don’t worry about things that have never happened before, regardless of evidence of how likely such events would be to occur, then this also means we have no reason to worry about, say, nuclear exchanges, because they’ve never happened. There are zero real life examples of escalating conflict from a nuclear attack, in one hundred percent of cases where nukes have been used, there was no retaliation. But this isn’t very comforting when you consider that in all existing cases of nuke usage, either there was no other country in the world with the capability for nuclear retaliation, or the nukes were being used in tests rather than attacks and there was nothing to retaliate for. Considering that we’ve had some very close brushes with outright nuclear conflict, the lack of standing examples certainly doesn’t tell us that we should model their plausibility as low.

    • Macbi says:

      This is true. It suggests that we should make all-purpose preparations that will defend us against many different attacks. For example we would want to make sure that emergency personnel could communicate with each other even if the usual phone network collapses. I imagine that many good protocols like this are already in place.

      • Desertopa says:

        This is what separates useful security infrastructure from, essentially, throwing money into a bottomless pit. Since most of it goes on behind the scenes though, it doesn’t have much impact on the appearing-to-do-things front. Most of the public discussion around terrorism defense essentially revolves around security theater.

    • Aegeus says:

      This reminds me of Bruce Schneier’s argument about how to handle terrorism: Don’t invest in fighting a specific plot you’re worried about, invest in intelligence to discover which plot the terrorists are thinking of.

      Reinforcing security at the most popular targets is still a good idea – not every terrorist is creative and devious, Al Qaeda kept on trying to hit airplanes for years – but you have a much smaller range of targets to protect if you try to stop plots at the planning and preparation stages.

    • SUT says:

      The problem with the stolen nuke isn’t purely or even primarily tragic loss of life. It’s what we have to do as a country, and what we have to do as individuals and “investors” who vote with their feet.

      a.) Nuclear defense involves retaliation, we will have to mass murder innocent people who have nothing to do with the act, all based on an obscure isotope analysis of the fallout. Or will we need to do 5-20 “Iraqs” simultaneously and initiate regime change and occupy the states which are most anti-American.

      b.) Not only will we lose the total economic value of the one city hit, but the desirability of living in a downtown metropolis anywhere will plummet for anyone “with something to live for” [sure people who move to detroit to do urban gardening will move to 42nd street, but not average finance guy with a wife and kid] and industries, the arts, sports will move out into the safety of sprawl. It will be the devaluation of taxi-medallions times about a million.

      c.) the security response of millions of new customs inspectors and immigration monitoring will make DHS/TSA look quainter than the Pony Express. Can’t wait for the most popular job of the 21st century to be parcel inspector.

      Etc, etc. In sum, the feedback response from an act of nuclear terrorism in the West is *THE* primary fork in the 21st century. If there’s one thing that would be desirable for our monstrously large federal gov’t to pursue as a goal, [and something only said gov’t can acheive] it would be to avoid going down that fork. There’s no turning back.

      • Desertopa says:

        The feedback from nuclear warfare would be dramatic, but the feedback from nuclear terrorism which the terrorists don’t have the capacity to readily repeat is likely to be much less, and besides which, preventing people from stealing nukes and using them is much easier than preventing regimes which can build their own from using them.

        But terrorists can provoke similar levels of escalation to that which would be caused by nuclear terrorism with any sort of attack that causes a similar level of destruction.

        • Mary says:

          Except that feedback is the desideratum. They would probably try, for that exact reason, to hit twice in rapid succession.

          Note that we used it twice in Japan — a “we’re not joking” sort of thing. (And used up our store in the process.)

  33. Will Everett says:

    Hey Scott,

    I just wanted to mention that you’re comparing the worldwide flu deaths in 1918 to the average US flu deaths per year. According to the WHO, worldwide there are on average 250k-500k influenza deaths per year. I’m not sure why the US gets 5-10% of those when we make up less than 5% of the population, but maybe we just have more old people and the flu is more likely to kill the elderly?


    • mobile says:

      We Americans are really good at attributing deaths to the flu.

      • Murphy says:

        Also there’s lots of common causes of death which just aren’t a thing in the US. TB, malaria etc. If you don’t die of those then you eventually die from whatever is left.

    • SlightlyLessHairyApe says:

      Yes, the flu is more likely to kill the elderly and otherwise immunocompromised folks. This is also a statement about the relative efficacy of antivirals as opposed to antibiotics — a patient in ICU (with no AD) that has a bacterial infection will get potent IV antibiotics. If that patient ever gets the flu, however, we can slow it down but ultimately the immune system has to clear it.

      So we’ve kind of squeezed a lot of those deaths (some of which were, sadly, inevitable) into the flu because that’s something we can’t fix too well.

  34. bean says:

    Your friend has a bad calibration on the destructiveness of a typical nuclear weapon. Modern cities are built much more strongly than those of Japan in the 40s, and terrorists are not that likely to be able to get one into a nice airburst. Also, buildings would get in the way, and most methods can’t account for that. Nukemap estimates 757,000 fatalities and just over a million injuries from a 340 kT B-61 in lower Manhattan. This is probably rather high, because of the shielding effects of the city. For a 10 kT weapon, I get 103,000 fatalities and 213,000 injuries.
    The real threat is bioterror, which is a lot harder to control, and could increase these numbers by orders of magnitude.

    • Chalid says:

      As someone who lives and works near a prime nuke target, I would be interested in hearing more about this. Do you have any recommended reading?

      • Eric Rall says:

        “The Effects of Nuclear Weapons” by Samuel Glasstone and Philip Dolan. It’s widely available online (one source here). Chapter 5 — Structural Damage From Air Blast is probably the most directly relevant section.

        The short version is that wood-framed residential construction and unreinforced masonry construction (typical building types in WW2 Japan, and not too different from modern US suburbs) are very vulnerable, while steel-framed and reinforced concrete buildings (common modern urban construction types, especially for commercial and industrial buildigns) hold up much better.

    • The Nybbler says:

      The deaths from a terrorist nuke in NYC are unlikely to be largely from the blast; it likely won’t be an airburst. Rather, it’ll be a surface blast in New York Harbor, killing people in Brooklyn, Jersey City, and lower Manhattan with the blast, but spreading a lot of fallout over heavily populated regions (which particular regions depending on which way the wind is blowing; mostly Brooklyn and Queens on a typical day)

    • Edward Scizorhands says:

      Society could break down if some unknown group sets off just one nuke. What if they announce they have a second, even if they don’t, and are going to use it in the next three days?

      • SUT says:

        Or worse, use the threat of the second (under “some city”) to extort us into allowing aggressors free pass in a campaign of conquest and genocide on their neighbors (e.g. MENA) or even on our allies (e.g. Japan)

        • Edward Scizorhands says:

          Not knowing who did it would make everything worse. I’d rather Russia successfully shoot a nuke at NYC than having an unknown group detonate one in NYC. We can bargain with Russia and strike back and they have something to lose.

        • bean says:

          I suspect that in that case, said country would end up glowing very, very quickly.
          Of course, if our nuclear forensics capabilities are still intact, we’ll know who was involved pretty quickly. (I’m not sure about this. 20 years ago, it would have been certain.)

      • ediguls says:

        I want to read this thriller.

    • J says:

      Yeah, I came here to post the nukemap. Detonating a “crude nuclear terrorist weapon” in the middle of San Francisco kills (cue trombone) 2900 people, and the acute radiation radius is 500m. I count about 7 blocks from end to end.

      • John Schilling says:

        The calculator seems accurate as it goes, but I’m not sure why they imagine the yield of a “crude terrorist nuclear weapon” would be 0.1 kilotons. The simplest nuclear weapons, whether gun or implosion assembly, mostly either produce a yield of ~10 kilotons or don’t work at all. Getting 0.1 kilotons requires either a rather sophisticated design, one which is also inefficient and wasteful of scarce fissile materials, or it requires an unlikely edge-case failure scenario. I note that “Nukemap” also includes an entry for a 10 kT improvised HEU device; I’d recommend using that even if you expect the hypothetical terrorists to be using plutonium.

        I’d also recommend assuming terrorists are smart enough to detonate their bomb at, say, the corner of 1st and Market in the densely-populated Financial District, rather than the geometric center of the somewhat irregular city. That leads to 95,000 fatalities with a 10 kT device, and still 10,000 dead if you insist on imagining terrorists can only get a 0.1 kT fizzle.

        • J says:

          Are you sure? The 3 NK tests, which are probably the closest thing to a terrorist improvised bomb, are 0.5kT, 6kT and 10kT.

          But let’s grant the 100k fatalities. That’s about 1/8th (~12%) the population of SF city proper, and 1.4% the population of the bay area as a whole. So even if the terrorists get a NK-state-of-the-art bomb to your metropolitan area (wp lists 62 metros over 1M pop), to the very center of it (which presumably is where the feds would put their radiation monitors), your chance of death is 1-10% (and way lower if it happens in a randomly selected metro area).

          I stopped stressing out over terrorist nukes when I plugged an airburst of the largest current US H-bomb (B-83, 1.2MT) into nukemap in downtown SF and saw that it doesn’t even make it to SFO.

          • bean says:

            Are you sure? The 3 NK tests, which are probably the closest thing to a terrorist improvised bomb, are 0.5kT, 6kT and 10kT.

            My understanding is that everyone was surprised that they managed to fizzle the first one the way they did.

          • John Schilling says:

            The 2006 North Korean test was almost certainly a failure of the device which then achieved 4-6 kT in 2009 – and there are unconfirmed that they preannounced the yield to the Chinese in a “don’t be alarmed but…” sort of way, indicating that this was the design yield.

            Of the eight nations which have openly tested nuclear devices, seven achieved approximately the design yield on the first test and the design yield was within a factor of two or so of the 10 kT “easy setting” on all but one of those. Possibly some of those nations had complete failures with no nuclear yield before the first reported test. One out of eight, North Korea, got a sub-kiloton fizzle. As bean says, that surprised people.

            That’s not something you can plan on, or expect from someone else. It’s the equivalent of, say, a presidential election being decided by a recount, or an association football game being decided by penalty kicks. If you play to win, you’re much more likely to just plain win or just plain lose; if you play for the cliffhanger ending, you’re most likely going to just plain lose unless you are very good.

            If you’re worried that someone might try and detonate their very first atom bomb in your city, you estimate the effects of a 10 kiloton groundburst and you estimate the cost of cleaning up after a dud, and that should cover at least 90% of the possibilities.

            And see my post below regarding the true measure of terrorism. If you stop the calculation with, “…and there’s only a 1.4% chance of my being killed in the blast”, you’ve ignored most of the potential harm.

  35. mobile says:

    If you’re going to cherry pick a date to exclude Haitian earthquake deaths, consider going out a few weeks, say, to March 2010. It may take a long time to die from an earthquake-related injury, or for some deaths to be discovered, or there may be deaths from aftershocks that occur several days after the main quake.

  36. SlightlyLessHairyApe says:

    Just to geek out on the data/stats question, perhaps the right thing to do is to call out our uncertainty explicitly by grabbing all the sliding 5 year averages and treating them as a big ensemble (weighting right at the ends where there’s less overlap). In other words, if the question is “how many terrorism deaths do you expect over the next ten years”, then we should posit that reasonable upper/lower bounds would be the highest/lowest ten year window in the recent past.

    So we can say that expected terrorism deaths per year are in the range of (10,600). That’s a large range, but to me that’s a feature in that it strongly highlights the year-to-year variability in the underlying data.

  37. Vegas says:

    I see even more dramatic illustration.. Gun grabbers point to German government as example how disarmed population resulted in less shooting, but expand time horizon and you see German government killing tens of millions of disarmed people.

  38. Skef says:

    Congratulations! I think this is the first time The Browser has shared a post of yours that doesn’t have something to do with psychiatry. You now have authority in their eyes beyond your expertise.

    • Anonymous says:

      I’m not sure how to navigate the Browser, but what about this or this?

      • Skef says:

        Huh – I guess I missed those when they came around (although I don’t remember if I knew about this site in 2014). I subscribe to the RSS feed and just noticed that they seemed to pick up all the psychiatry articles but not the others.

  39. RCB says:

    I think the whole premise here is flawed. My view is that the total number of deaths is not a good way to order “how much one should care” about an issue.

    From an actionable standpoint, the real question is “which issue, if devoted attention and money, would result in a greater number of lives saved?” It’s about return per unit effort or expenditure.

    While falling furniture might kill more people than terrorists in any period of time, it might also be that efforts to reduce terrorism will save a lot more lives than efforts to stop furniture from falling. Perhaps furniture deaths are all freak accidents that all but the craziest laws and regulations would prevent. I’m not certain this is true, but it sounds right.

  40. MichaelT says:

    The problem with your friend’s argument is that most of the resources expended on the War on Terror, as well as most of the curtailing of our civil liberties, are directed at preventing small to medium sized terrorist attacks, not the large black swan ones. How can the Department of Homeland Security, who can’t even find 10% of banned contraband being brought on planes, stop someone from detonating a nuke in Times Square? How do our military adventures in Afghanistan, Syria, Iraq, Yemen, etc. stop ISIS from getting nuclear technology when none currently exists in any of those countries?

    • Sandy says:

      The obvious answer would be that it requires a lot more organization, communication and planning to smuggle a nuke into Times Square than it does to conceal contraband on a plane.

      A nuclear plot is a lot more conspicuous, making it easier for DHS to detect and act on. It’s not like you can just buy a nuke on Amazon, after all. Nuclear weapons programs are heavily guarded, monitored and regulated.

      How do our military adventures in Afghanistan, Syria, Iraq, Yemen, etc. stop ISIS from getting nuclear technology when none currently exists in any of those countries?

      ISIS aren’t the only terrorists with nuclear ambitions. al-Qaeda and the Taliban are quite open about their desire to get their hands on Pakistan’s nukes, and that country is not exactly known for its firm clampdown on terrorist groups.

      Also America has nukes parked in Turkey, which is right next door to Syria, and 4 more degrees will get you to Kevin Bacon.

    • Aegeus says:

      I’d argue that large-scale attacks are thwarted by overseas adventures more than small-scale ones.

      Large-scale terror attacks take planning, coordination, and supply chains, which can plausibly be destroyed by bombing. Terrorist who are hiding from drones are terrorists who aren’t out shopping for uranium. But the smaller “lone-wolf” terror attacks only need a suicidal guy with a gun (or even a truck), so there’s very little chance to stop them before they launch the attack.

      • Edward Scizorhands says:

        You now have me pondering the theory “overseas adventures stop the large-scale attacks, but encourage the small-scale attacks, and this is worth it because just one nuclear bomb could end American civilization.”

        I’m not saying I agree with it. Just that I have to think about it.

      • houseboatonstyxb says:

        @ Aegeus
        the smaller “lone-wolf” terror attacks only need a suicidal guy with a gun (or even a truck), so there’s very little chance to stop them before they launch the attack.

        So, how long before a small or medium size group obtains a nuclear bomb to put in their truck? Quite a few years before they could get the bomb to the US. But if terrorists start nuking European/UK cities in retaliation when the US does something they don’t like … how long til Europe/UK get fed up with it, and what may Europe/UK do then?

        • Saint Fiasco says:

          how long before a small or medium size group obtains a nuclear bomb to

          I honestly don’t know. Aren’t nuclear bombs very hard to get?

          I think you would need to be a nation-state or have a nation-state sponsor to get one and at that point MAD applies. After all, Afghanistan had a tenuous connection to 9/11 and look what happened to them. If those terrorists had a nuclear bomb Afghanistan would have been nuked too.

          • You can crudely create a nuclear weapon with some probability of success. However, there is a high chance of failure, and you’re probably topping out at 10kt.

            You also need the fissile material, which is NOT easy to get. There are over a dozen reported cases of stolen fissile material, but you need a lot of fissile material.

            Successive nuclear blasts are not going to happen unless Pakistan immediately collapses and someone steals all the HEU.

  41. Sly says:

    Did anyone else find this to be weirdly one of Scott’s worst posts? No, you do not get to count hypothetical non information in your average calculation. Meteors kill 0 people a year on average, not a hundred – and saying that they kill 100 is simply an error.

    • J says:

      My personal beliefs are skewed heavily toward “the best response to terrorism is to not overreact”, so I’m sure that colors my judgment. But yeah, while he makes a good point about tail risk, I feel like it’s a big omission not to address the absolute scale: even if you start counting from 9/10/2001, death by terrorist is still ridiculously unlikely.

      • SUT says:

        The whole point of what we learned is that we didn’t take it seriously enough.

        Every bookshelf in the country has Tom Clancy’s Debt of Honor, which portrays a terrorist crashing a plane into the capital. People at the flight school found it odd that certain students didn’t want to learn to land the plane. And there was the entrenchment of Al Qaeda leadership in a mountainous safe space with increasingly brazen and sophisticated acts being executed thousands of miles away, e.g. USS Cole.

        Even if you’re not sympathetic with the current War on Terror, it very likely could have been prevented from reaching its current state, including Bush’s wars, and Obama’s drones, from all ever happening if we had truly evaluated our risk and acknowledged the ruthlessness of the enemy. If we could have thwarted the attack, we wouldn’t need to respond.

        • J says:

          That’s hindsight bias. Lots of books have people getting mugged when they go into big cities. Some people get suspicious when they see people in big cities who look like the stereotypes they saw on TV and in books. Some people actually do get mugged. None of that tells us very much about whether we’re being too tough or too lenient on crime as a society, or how much risk there actually is of getting mugged. And even if we emptied out all the cities and bombed all the stereotypical mugger-looking people to death, people would still get mugged once in a while.

    • TomFL says:

      100 year on average is in fact technically correct over a long enough time period. The statistics for large asteroids is fairly well known and death tolls from large strikes can be estimated. So a crapload of people will be killed in very rare strikes.

      But the statement is misleading. One should instead say 10,000,000 will die in a large asteroid strike that occurs about once every 100,000 years. The prior statement infers 100 die every year.

  42. Graeme says:

    If I may point out the obvious, the more influence humans have in precipitating/mitigating an event, the more complex analysis becomes.

    Let’s look at the simple one: volcanic eruptions/earthquakes/tsunami are regular and recurring. You can approximate the probability and severity, estimate the odds of it occurring over any future period, etc, etc. It only gets complicated when you have to analyse the costs, namely the damage done to a city and it’s people, with and without mitigation, but even that is relatively easy compared to human precipitated disasters. Call these N=1.

    We can’t STOP tsunamis, we can only mitigate the consequences. By contrast, nuclear disasters are much harder to deal with. They don’t occur with some sort of blind externalized periodicity, our actions effect both the severity and the frequency of disaster. The hydro dam example is a good one, because of the complexity involved in examining both the probability of the event AND it’s consequence. With hydro, you can mitigate most effectively by making sure you don’t put villages in the flood zone, but there may be compelling reasons to do so and the probability of failure is highly dependent on civil engineering practices.

    The good news is that people DO make a study of these things. As structures fail, we can examine the causes, and determine whether they are applicable elsewhere and adjust our views accordingly. In the light of Chernobyl, Three mile and Fukushima, we are much better positioned to evaluate the risk, even if those events themselves were rare and unpredictable. Call this N=2

    Terrorism is particularly difficult to analyse. Other entries in this class of problem would be disease epidemics, war, crime, and to a lesser extent famine (and Economics). Like structural failures, both the cause and effect are due to human factors, but unlike structural failures, the failure is caused by active, responsive agents. Call these N=3.

    The conditions experienced by a building are constant (or rather variable, but within predictable statistical constraints). Buildings are subject to an external set of rules; we might change how we build them, but we don’t change how they fall down. They are essentially deterministic.

    By contrast, level 3 problems create response and are themselves responsive to attempts to mitigate. Terrorists may come up with new ways to attack (use of an airliner as a weapon is quite the innovation, when you think about it), or they may be motivated not to bother. The level of terrorist threat determines our response, our response determines the terrorist threat, and endless complexities spiral outward. The crux of the matter is that we can examine level 1 and level 2 problems at any point in history, and we get an apples to apples comparison. The factors are known, or at least knowable to some degree of statistical certainty. By contrast, you cannot calculate the risk of a bombing in 2017 because you can’t calculate what the odds were in 2001, even in retrospect. Unless you can create a comprehensive, accurate, and predictive model of human behavior, you are not going to come up with a cogent answer.

    Which makes certain comparisons very dangerous. You can’t talk about the odds or casualty rates of terrorism because we literally don’t know what they are, and historical data does not reflect the current reality. There have been no major US attacks in the last 14 years: was this due to the security apparatus? Or did the probability of occurrence change?

    That said, we if we pretend level 3 problems are level 1 or 2 problems, we get conspicuously awful and misleading statistics, as Scott’s essay so nicely demonstrates. If we want to do better, and we need to if we are to cogently allocate resources to fighting problems of this type, we need to look at both sides of the problem. We must make a study of more than just attempts vs outcomes, but also the sociology underlying the system. At a bare minimum, if we can identify a set of characteristics that occur reliably in successful terrorists, we can at least begin to estimate the scale of the potential threat.

    • Uncle Ilya Kuriakin says:

      Graeme says: “Terrorism is particularly difficult to analyse.”

      Many people (including me) have long agreed with Graeme’s point, as illustrated by these three resources:

      • President Harry Truman’s presentation to General George Marshall of the Distinguished Service Medal and Gen. Marshall’s acceptance of it (audio recording, 1945).

      • The CAIN (Conflict Archive on the INternet) of the Irish Peace Process, including Michael McKeown’s Database of Deaths

      • The Parents Circle-Families Forum (PCFF), an organization of Palestinian and Israeli families who have lost family members in the conflict.

      Michael McKeown’s Database of Deaths, in particular, represents on the one hand, the apotheosis of SSC-style number crunching. And yet on the other hand, the CAIN of the Irish Troubles, when appreciated as a whole, brightly illuminates the post-quantitative post-rational cognition that effective peace processes require.

      And the PCFF’s articulation of anguished hopes speaks even more directly to the vital role of post-quantitative post-rational cognition in successful peace processes.

      Historically speaking, it’s not easy to better Marshall’s 1945 reflections upon practical methods for the construction of peace and prosperity; reflections that provided the foundations of the Marshall Plan (1948-1952) — a plan that was much-reviled by the far-right of that era — which led to Marshall’s Nobel Peace Prize (1953):

      (audio here) To the soldiers of the Army, you still in uniform and you who have already returned to your home, I, as one citizen and one comrade, express my deepest gratitude.

      You were the greatest protective force this nation has ever known. In its direct hours you carried the might of America into action. You gave to the United States its rightful prestige among nations. And no one knows better than you that you did not do it alone. You know how tremendous was the contribution of the nation as a whole, of the millions of laboring men and women whose productive efforts supplied your munitions, of the factories of this great nation and the science and management that guided their operation, of the farms, and of the plain citizen who put your needs above his own.

      But of all these efforts yours was by far the greatest. You faced death and swallowed fear, endured the agonies of battle and of hearts torn by loneliness and homesickness and starvation for the normal life you loved. Yet you took it — all there was to take anywhere on the battlefronts of the world. And you had the strength and will to give it back, give back much more than your enemies could take. You know that those who stayed behind were no different than you. Had they been out front and you behind, all would have served just as you did. That is the genius of America; that is the strength of a free people.

      Most of you know how different, how fortunate is America compared with the rest of the world. That is something those at home cannot fully appreciate.

      Today this nation with good faith and sincerity, I am certain, desires to take the lead in the measures necessary to avoid another world catastrophe, such as you have just endured. And the world of suffering people looks to us for such leadership. Their thoughts, however, are not concentrated alone on this problem; they have the more immediate and terribly pressing concerns–where their next mouthful of food will come from, where they will find shelter tonight and where they will find warmth from the cold of winter.

      Along with the great problem of maintaining the peace we must solve the problem of the pittance of food, of clothing and coal and homes. Neither of these problems can be solved alone. They are directly related, one to the other.

      It is to you men and women of this great citizen-army who carried this nation to victory, that we must look for leadership in the critical years ahead. You are young and vigorous and your services as informed citizens will be necessary to the peace and prosperity of the world.

      To belabor the not-so-obvious, the history of conflict resolution shows plainly, that a too-strict insistence upon the numerical accounting of terror and suffering, and the consequent unending arguments associated to that accounting, have far more commonly occluded paths to peace and prosperity, than illuminated them.

      Do you want to effectively illuminate paths to peace? Begin by listing the deaths associated to a conflict. All the deaths. But don’t categorize the deaths as justified (or not) versus terrorist (or not). Just list them. That’s the start.

      PS  On many medical services, a Morbidity and Mortality Conference serves many of the same purposes as a Truth and Reconciliation Commission, and runs under similar guidelines.

      A sincere question (whose answer is not known to me) is this: do psychiatric services commonly host Morbidity and Mortality Conferences? If so, what lessons are learned? If not, why not?

      • Graeme says:

        While I agree with what you are saying in general, the phrases “Post-Rational” and “Post-Quantitative” concerns me a bit. We should not give up on rationality, or trying to quantify the damage, we just need to recognize that the current toolbox is insufficient to the task. In particular, average deaths per year is utterly inadequate.

        The hardest part of this stuff is that resources are finite, and allocation is generally by consensus (even if that consensus is collected under threat of lethal force). We will never know how many people would have died in Vietnam the US had been a little weaker, nor how many if they’d been stronger. Would the world have fallen to nuclear war if the US pulled it’s military budget out of missiles and put it into schooling and public health? It’s frustrating, when you know exactly how much a life is worth ($3500 USD is one less dead child), but you don’t know how valuable airport scanners are (I think they’re worthless and expensive, but I’m no expert).

        It’s even harder when you realize that people LIKE having a military and are willing to coordinate to have one, but are pretty ambivalent about saving children in the third world (revealed preferences). I am not sure you could rally people around saving 30 million lives in Africa. Actually, I’m 100% sure they wouldn’t: no elected politician has ever proposed axing 100 billion dollars from defence to hand over to the third worlders. If they did, they would be torn apart immediately for leaving America defenseless and selling out their people to foreigners.

        So yeah, we need better metrics than deaths. But this stuff is HARD. One of my favorite quotes:

        For every complex problem, there is an answer that is clear, simple, and wrong

        • Uncle Ilya Kuriakin says:

          Graeme observes: “The current toolbox is insufficient to the task.”

          Extending humanity’s cognitive toolbox is exceedingly dangerous in practice and hence, inherently contentious in principle, isn’t it?

          Today’s counter-Enlightenment finds its disciples (AFAICT) chiefly among those who seek reasons to reject progressive modernity’s campaign of toolbox-extension (both specifically in respect to anti-terrorism, and generally in regard to all realms of human enterprise).

          “Modernity+” (as it might be called) has at least four active initiatives:

          The First Peace Initiative of Modernity+: Rationality  (with a social emphasis)  See for example, Saaty and Zoffer “Principles for implementing a potential solution to the Middle East Conflict” (Notices of the American Mathematical Society, 2013).

          The Second Peace Initiative of Modernity+: Parcellation  (with an empathic emphasis)  See PubMed surveys of the rapidly evolving STEAM-enterprise (e.g., using PubMed/MESH terms like “Connectome/methods”[MAJR] or “Brain+Mapping/methods”[MAJR]”).

          The Third Peace Initiative of Modernity+: Freethinking  (with an Anabaptist/Spinozist emphasis)  For a conservative-friendly historical grounding see Leszek Kolakowski’s “Dutch seventeenth-century non-denominationalism and Religio Rationalis: Mennonites, Collegiants and the Spinoza connection” (2004). For a parallel strand of feminist-friendly hilarity — equally in the comedic and philosophical senses of “hilarity” — see Amélie Rorty’s good-hearted “Spinoza on the pathos of idolatrous love and the hilarity of true love” (2009).

          The Fourth Peace Initiative of Modernity+: Medicine  (with a regenerative emphasis)  Here a good starting point is are the survey articles that were collected in the theme issue of The Journal of Academic Medicine for 2009: “How should academic medicine contribute to peace-building efforts around the world?”

          Against Modernity+’s diverse cognitive, social, philosophical, and STEAM-initiatives the counter-Enlightenment has embraced a futile strategy of abusive “Thwhack-ye-Mole” … while the fast-evolving cognitive toolset of Modernity+ gains inexorably in STEAM-power and global social dominance.

          Which is good, isn’t it? 🙂

  43. John Schilling says:

    OK, the biggest problem with both of those graphs is that they are counting the number of people who have died from terrorist attacks. That’s about like counting the number of people who have died of rape – it happens, but it’s not the point, and by any rational metric it is a small fraction of the cost.

    The purpose of terrorism isn’t death, it is terror. There have been successful terrorist attacks which have caused plenty of terror without killing anyone, usually by design. And even factoring in the attacks that do cause death, the bulk of the harm is not in the shortened lives of the relatively few direct casualties but in the reduced quality of life for everyone else. Rather like rape.

    Granted, reduced quality of life due to chronic terror is not something that can be easily quantified and put on a graph. But it’s the one thing that absolutely does need to be on that graph.

    • albatross says:

      The poster child for this is the anthrax attacks. They caused a huge amount of panic, and yet if the attacker had sent the same spores in the same envelopes to the same people, but with no reference to them containing anthrax spores, I doubt anyone would know what had happened. Probably more people would have died, but nobody would have connected the deaths and realized there had been a terrorist attack with bioweapons.

    • TomFL says:

      The panic from a dirty bomb will be off the charts. However it appears that these can actually be cleaned up fairly effectively given some effort. Radiation levels will still be higher after cleanup but the risk fairly low. This will convince absolutely no one to move back in.

      Realistically they just need to explode a tiny dirty bomb with measurable radiation for the same effect. Most people’s understanding of radiation levels is non-existent. I think they should measure everything in banana-rads.

      • Macbi says:

        In fact the government will force people to leave like they did in Fukushima. Otherwise low house prices would induce smart people to move in.

        • Jiro says:

          Two types of people would move in:
          1) people who are smart, and by whose standards moving in is a good deal
          2) people who are stupid, and by whose standards moving in would be a bad idea, but who would choose to move in anyway. (Note that saying “it would be their revealed preference to move in” is essentially equivalent to saying “there is no such thing as stupid”.)

          Actually, there are more, overlapping, types:
          3) People who are children and can’t control where to move to, but who have parents who are stupid and by whose standards moving in is bad
          4) People whose presence makes life easier for other people by whose standards it would be better living near the first group far from Fukushima than living near them near Fukishima. In other words, if your employer moves near Fukushima, you may be stuck.

  44. Anon says:

    Thinking more about it, I’d add people aren’t concerned necessarily that they’ll be killed by a terrorist as much as that someone frequenting a high profile target will be killed by a terrorist. Say ISIS manages to detonate a crude nuclear weapon (100 ton yield) at Grand Central. Nukemap gives a fatality estimate of roughly 45,000. Now what were the relative odds during arbitrary time period x of a person *who had visited Grand Central* being killed by a terrorist vs. by falling furniture? Of course an individual could simply avoid Grand Central, but then we’re in a situation where someone acting rationally must avoid one of the most culturally significant locations in America to get their safety back down to falling furniture levels. This would seem to justify the fear and outrage we observe in response to terrorism, yes?

  45. J Mann says:

    Some random thoughts:

    1) At one level, it’s OK for people to be more upset than some deaths than others. We do our best to achieve 0 airplane deaths, not because airplanes are bigger killers than chairs, but because air crashes are more upsetting. See also police shootings of black suspects, soldiers dying in overseas wars, victims of mass shootings, etc.

    In every case, Mister Spock might be more influenced by a “if you want to save lives, outlaw motorcycles” argument, if if people are more upset by plane crashes or terrorism or police shootings than by chair and motorcycle deaths, then that factors into the utility equations as well.

    2) I’ve been astonished for a while about how bold some of the facebook comparisons (terror to mass shootings or whatever) are about starting the clock on 9/12/01. You’d think if you were going to cut 9/11 as an outlier not likely to be repeated, you’d start a few years after 2001 just to look fair, or at least 1/1/02.

    • Montfort says:

      As for 2, they make it obvious because if you think 9/11 is an outlier, then you also think excluding it is fair, so why would you try to conceal it?

      • J Mann says:

        It seems to me that if you include the terror-free period right up to the last even that you want to cut, you’re throwing off your average.

        Theoretically, you could have started at any point in the period between 9/11/01 and the next terror event, but by grabbing that whole period, you get the lowest average you possibly could.

        I’m no statistician, but it seems to make more sense either to include a much longer period and explicitly exclude 9/11 or normalize it to what you think the number should have been, or to use a point halfway between 9/11 and the next event.

  46. mjg235 says:

    We should also note that the tail risks of a major terror attack are not only measured in extremely high death toll, but substantial loss to infrastructure and heritage. Say ISIS grabs one of our nukes in Turkey and takes out New York City. The loss of life will be in the millions, which is astonishing enough, but we would also lose Columbia, NYU, the New York Stock Exchange, the New York Times, our best deepwater port infrastructure, the vast majority of the fashion industry. I could go on.

    These are a lot of the things that structure modern life, and I shudder to think what we would do were they to disappear in an instant. It is far more difficult to quantify, but definitely a part of the risk analysis for this sort of event.

  47. Jacobian says:

    Didn’t Nate Silver write in The Signal and the Noise that terror attacks, like earthquakes, are distributed according to a power law? I think for terrorism, attacks that are 10 times as deadly happen 10 times less often.

    This means that there’s an equal number of casualties from each “band” of terror attacks: the hundreds that kill a few people, the dozens that kill dozens and the 9/11 that kills 3,000. If 9/11 can happen roughly once every 20 years, an attack that kills 300,000 happens once every 2,000 years. Maybe that’s too long to worry about, or maybe were due 😛

    • J Mann says:

      There’s a big discussion upthread on this – just search on “power law.” Here’s a good article on Aaron Clauset, the guy whose research Silver discussed in his book.


      TLDR: Clauset thinks power law fits terror pretty well, and that this implies that the risk of a catastrophic attack is small but much larger than people think, but he doesn’t think he has enough data to be confident in his conclusions.

    • Douglas Knight says:

      You’re saying that it’s distributed by a particular power law, not just some power law. I don’t think Silver is that specific, but his source, Clauset and Young, say that it has a parameter of 2, not 1.

    • TomFL says:

      Terrorism isn’t physics. Extrapolating terrorism to natural laws is a bit of a stretch and seems like academic whacking off. Modelling death by wars in 1950 would have shown that huge numbers of dead were highly likely in the next 50 years and the advent of nuclear weapons would have been further proof of the likelihood of that happening. Very convincing…in 1950.

      I’d like to see what his model would have been on Sept. 10th 2001 and what his predictions would have been for a catastrophic terror attack the next day.

      I can see where the ability of small groups to commit catastrophic damage is controlled by their will and means to do so. Cavemen don’t have the means to commit nuclear holocausts. The local nunnery will never have the will. Beyond that I think modelling this is worth exactly zero.

  48. Jacobian says:

    Scott, you’re not worried about the giant chair of Yur-Nagansa that will topple on top of the world killing us all? Outliers!

  49. Patrick says:

    I don’t understand why death by chair accident is being held up as something silly. People are really bad at falling off stuff. Especially as they get old. Every time an old lady falls down there’s a chance that their card just got punched. It’s a real thing.

    That being said the cynic in me wants to run this chair graph alongside “margin of police shootings of unarmed African Americans above the expected value given population percentage” and see if they’d print it.

    • Chalid says:

      Also, toddlers are occasionally killed by pulling heavy pieces of furniture down onto themselves (say, having it tip while they are climbing it). There was a major IKEA recall because of this.

      As the parent of a kid who likes to climb, “death by furniture” features much more prominently in my daily nightmares than “death by terrorism.”

  50. Mary says:

    One notes that unlike the flu and the furniture, terrorism has a motive to escalate.

  51. Sabril says:

    How many people are year died from lynching during the Jim crow Era? Very few but the problem was worth taking seriously because the aim of lynching was to terrorize a population. So too with modern terrorism. Because the victims are not just the dead and injured. It includes everyone who now feels uneasy on a clear autumn day in nyc.

    These people who downplay terrorism by counting the dead, I wonder how they distinguish lynching? I doubt they can do it. I doubt they can distinguish any crime where much of the harm is psychological

    • cassander says:

      FWIW, less than one, after the 1940s.

    • pku says:

      But talking about how terrorism is rare and unlikely to harm you does a lot more to address this part of it than, well, pretty much anything else.

      • sabril says:

        But talking about how terrorism is rare and unlikely to harm you does a lot more to address this part of it than, well, pretty much anything else.

        Well that’s a different issue. But do you have any evidence to back up your claim?

  52. DaveL says:

    My friends and I made an app that uses mortality stats to show you what you’re likely to die of (built in <2 days, only works well on mobile)


    Video: https://govhack-storage.s3-ap-southeast-2.amazonaws.com/2016/GovHackWhatAreYouAfraidOfv2.mp4

    We just went back and used all the data we had (eg all terror attacks in database and shark attacks since white settlement). Outliers and the future isn't guaranteed to be like the past etc… But I still think it would probably be better to have a fear of fatty foods and lack of exercise etc than more visceral yet less common ones like sharks and terrorists.

    • John Schilling says:

      Where did you get the mortality stats for “killed as a Suspicious Loner by one of the neighborhood vigilante groups that sprung up when the legitimacy of governments failed after one too many terrorist attacks (even though the terrorists didn’t kill that many people)”?

      As Scott said, it’s the outliers that matter. If you’re wondering whether you should be afraid of terrorists, “how likely am I to die in a terrorist attack” is not the most important question.

  53. sabril says:

    I think it’s pretty likely that this Andrew Shaver is just virtue-signalling by downplaying the crimes of non-westerners. If some right-wing White Nationalist group started lynching one or two blacks or gays a year, would he whitewash it by arguing that you are far more likely to be killed by a piece of furniture? Or would he be marching in the streets, screaming for Something to Be Done? I know where my money is on that bet.

    • pku says:

      I don’t know about Andrew Shaver specifically, but in general, “terrorism is statistically negligible and shouldn’t be as scary as it is” is used by the more moderate/practical left rather than the SJW left – that is to say, the people who don’t want to justify or excuse it, but want to talk calmly and rationally about how everyone should stop panicking about it. The same group who are most likely to argue that both KKK members and people who riot during BLM protests are marginal phenomena that shouldn’t terrify us, rather than pick a side between them.

      • sabril says:

        I don’t know about Andrew Shaver specifically, but in general, “terrorism is statistically negligible and shouldn’t be as scary as it is” is used by the more moderate/practical left rather than the SJW left that is to say, the people who don’t want to justify or excuse it, but want to talk calmly and rationally about how everyone should stop panicking about it.

        I’m a little skeptical . . . can you give me three examples of pundits who use the “body count argument” to downplay terrorism AND make a similar argument about crimes where the perpetrators are white and the victims are non-whites?

      • Protest Manager says:

        Could you let us know when those same people use the “small body count” argument against “Black Lives Matter”?

        Because of the terrorism body count is “no big deal”, neither is the police body count. Esp. when compared to the number of blacks killed by other blacks.

        • Saint Fiasco says:

          Terrorism is a big deal when it consistently targets a specific subset of a population. For example when terrorists persistently attack tourist attractions in poor countries, it can have a significant effect on that country’s economy because tourists don’t feel safe there, making the job of the tourism industry very difficult.

          If police treat black people badly that makes blacks feel less safe, making the job of the police more difficult.

          • John Schilling says:

            But terrorists treating the class of “American people” badly can’t make American people feel less safe, have any significant economic impact, affect the legitimacy of governments, etc?

            You might argue that terrorist attacks shouldn’t have such broad effects, but they do.

    • Macbi says:

      I can’t speak for the author’s motivation, but I don’t think your analogy is good. It’s not that terrorism isn’t worth fighting, it’s that it isn’t worth spending a trillion of dollars fighting. I would be happy for the police to spend millions of dollars to take down a White Nationalist group like the one you describe, but if the price tag headed up towards the billions I’d object. If government regulations on chairs cost a few million a year then that’s fine by me if it saves some lives, but if they cause the price of the average chair to double then they’ve gone too far.

      • sabril says:

        I can’t speak for the author’s motivation, but I don’t think your analogy is good.

        Well are you disputing that — just like the case with lynchings — there is a good deal more at issue with terrorism than just the body count?

        t’s that it isn’t worth spending a trillion of dollars fighting.

        That’s a bit different from the point I was addressing.

        Let me ask you this: Do you agree that more should be spent to fight crimes or other societal ills which cause massive psychological harm & physical harm than crimes or societal ills which cause the same amount of physical harm but far less psychological harm?

        • Macbi says:

          I agree with you in principle, but in the case of terrorism it seems that politicians are intent on increasing the psychological harm as much as possible for their own ends. It would be better if as well as trying to stop terrorism we also tried to assure people that they needn’t be afraid of it.

          • I don’t think anyone yet has mentioned Syndic, an old sf novel where governments have collapsed and been largely replaced by organized crime. The eastern half of the U.S. is run by the syndicate–undemocratic but a reasonably free and attractive society.

            The terrorists are the U.S. government in exile, based on islands in the Atlantic. They occasionally blow things up and the Syndic mostly ignores it. “A chemical factory blew up? That’s too bad–but chemical factories sometimes do that.”

            Which is very frustrating for the terrorists.

            There is a conversation near the end between the protagonist, who wants something done about the government, and an older and wiser relative who argues that doing anything serious about them will undercut the existing structure of the Syndic’s society–better to ignore them and put up with an occasional explosion.

          • LHN says:

            Though being unflappable in the face of repeated, varied attempts to determine what will provoke a reaction may not actually be possible for humans. An individual might be arbitrarily stoic, but I’m more skeptical that a government or society can be. E.g., it’s easier for a mob boss to be philosophical about a chemical plant than about his kid’s school.

            I haven’t read the book, so it may be the USG in exile has limits on what it’s prepared to do. But obviously there are real-world groups that don’t, at least anywhere within the plausible tolerance range of Western electorates.

          • I think it’s clear in the story that the damage being done by terrorism, while significant in absolute size, is tiny in terms of the total size of the society.

          • sabril says:

            I agree with you in principle, but in the case of terrorism it seems that politicians are intent on increasing the psychological harm as much as possible for their own ends.

            Can you give me 3 examples of this so I know what you are talking about? TIA.

  54. Njnnja says:

    When doing statistical analyses, outliers should receive close scrutiny to look out for data errors, or to find relationships or regime changes that are otherwise difficult to see, but there is no competent researcher who would say to remove outliers from an analysis simply because they are outliers.

    Outliers are not less important than “normal” data points (whatever those are). The idea that outliers should be removed is what researchers joke about when data don’t do what they want the data to do, not a legitimate approach to data scrubbing and validation.

  55. TallDave says:

    This reminds me of the question from the guy who is struck by lightning on Monday, who asks which day in the future is most likely to be his next to feature a lightning strike. The official answer was “Tuesday” (because each successive day has to account for the increasing probability of any of the prior days’ having already experienced the next lightning strike), and I pointed out technically this depended on the time the question was asked on Monday, because very early in the day not enough of Monday has passed that the greater amount of time in Tuesday overrides that “prior day” effect, and we could probably calculate that time from the likelihood of lightning strikes.

    And then Jimmy Wales liked my comment. It was the greatest day of my life [citation needed].

    So anyways, the most likely day for the next horrifying terrorist attack is always… tomorrow. Unless you’re up late, in which case it may be “today.” Enjoy that in your brain!

  56. Yossarian says:

    One other thing you shouldn’t forget is the existence of feedback loops in cause and effect chains. It would be better to compare terrorism with yet another highly publicized cause of death – airplane crashes.
    Sure, each airplane crash by itself kills a lot of people, but still, statistically, air travel is among the safest travel modes. And yeah, with the sort of media coverage these events get, there will be people irrationaly scared of flying. But what happens afterwards? The plane people get together, analyze the event, figure out what caused it and invent ways to make air travel even safer – using redundant systems, figuring out ways to make engines fail less often, making sure the pilots are professional and healthy, prepare contingency plans for what to do when something goes wrong, yada yada. This might cause the ticket prices to go up, but it’s also likely to actually make planes safer (and that probably among the reasons that we do have the safe air travel). So it’s a neutral, maybe even benevolent feedback mode.
    On another hand, once we start media hysterics about terrorism and start making too many waves – we give the terrorists the publicity, fame and notoriety that they want, so we are likely to get more people interested in terrorism as a result. Couple that with all kinds of unpleasant things and costs that an overly paranoid counter-terrorist program will inflict on us – and you get a feedback mode that is bad, no matter how you look at it. And once you get more terrorism, you get even more paranoia and media coverage, then you get even more terrorism..

  57. Doug M. says:

    Just for the record: asteroid strikes are rapidly becoming a thing we don’t need to be scared of.

    Why not? Because in the last fifteen years or so, we’ve mapped the orbits of every Near-Earth Object (NEO) with a diameter greater than a kilometer or so.(1) NEOs are the asteroids and short-term comets that cross Earth’s orbit. By definition, NEOs are almost the only thing that can possibly hit us. (2) And as of this writing, we’re pretty sure we’ve bagged every single NEO bigger than 1 km, and the majority of NEOs bigger than 300m. By 2030, we expect to have all the 300m NEOs as well and to have taken a solid bite out of the 100m NEOs.

    None of the asteroids that we’ve currently mapped are on a collision course with Earth. NEO orbits are chaotic over astronomical time, but on a time scale of decades to centuries they’re extremely predictable. So, nothing big is going to hit us in the next hundred years or so. *We know this*. We know it to a very high degree of confidence.

    What does “nothing big” mean? Well, the asteroid that killed the dinosaurs was probably around 6 km across. At the other extreme, the one that exploded over Chelyabinsk a few years ago — making a loud scary bang and breaking a lot of windows, but not injuring anyone — was around 25m across. The biggest rock that could plausibly “sneak up” on us in the next few decades would be around 300m across. That would hit with a force around 800 megatons(3), which would kill everything within about a twenty mile radius and might give us an 1816-style “year without a summer”, depending on when and where it hit. Unpleasant, but for most of humanity no more than an inconvenience — and let’s note again that we’ll have all the 300m asteroids mapped by the end of the next decade. When you start talking about 100m asteroids, well, that’s just a big random bang that’s not going to have any major effects outside of its immediate locale. (With regard to which, recall that 80% of the Earth’s surface is deep ocean, empty desert, or Antarctica.)

    Fear of asteroid strikes is becoming as irrelevant as fear of smallpox.(4) This is not widely appreciated, but it’s true.

    Doug M.

    (1) Scroll down to the last graph on this page: http://neo.jpl.nasa.gov/stats/ That’s the one for discoveries of big asteroids, over 1 km across. The surge around 2000 is when our telescopes and techniques got good enough to start capturing them consistently. See how it declines into a long tail? In 2000 we were finding new large NEOs every few days. In 2016, we haven’t bagged a new one since April. The current count is around 880; the best guess is that the actual number of large NEOs is well under 1,000, and that we already have around 95% of them.

    (2) In theory we could be hit by a long-period comet. However, this is a lot less likely than being hit by a NEO. Your average NEO crosses Earth’s orbit once every couple of years; your average long-period comet crosses Earth’s orbit once every few thousand years. (Halley’s Comet is an oddity — for a long-period comet, it has a very short orbital period.) In fact, the odds of a comet strike are so small that it’s possible it may never have happened once in Earth’s history. Comet strikes should look slightly different from asteroid strikes — a comet has a different composition, and is moving a lot faster. But every time we’ve been able to firmly identify the source of a strike, it’s been an asteroid. Long-period comets don’t seem to be a meaningful threat.

    (3) Assuming a porous rocky asteroid striking at average NEO orbital velocity and inclination, using the Imperial College impact calculator.

    (4) Wait, but what happens when we DO spot an asteroid that’s going to hit us? — Well, then we can probably move it. The more advance warning you have of a strike, the smaller the delta-V required to move the NEO off an impact course. With decades of warning, only a very gentle nudge — a few cm/s at most — will do the job. That’s well within our current capabilities.

    • J Mann says:

      Very interesting – thanks!

    • Edward Scizorhands says:

      Since you have the knowledge, I’ll ask: how do we know we can see the 1km NEOs? What if they have a low albedo? Do we have independent evidence (meaning, aside from “what we can see”) that low-albedo bodies are rare?

      • Douglas Knight says:

        The link says that they assume that the albedo of NEAs are like those in the asteroid belt, which is big and thus easy to study. It should be easy to model the likelihood of spotting an object as a function of albedo, and thus have good bounds on how common are dark asteroids, and thus dark NEAs.

      • John Schilling says:

        A very low-albedo, near-earth body will necessarily be a warm body – it will be absorbing lots of solar radiation, and then reradiating it in the infrared. One of our most prolific NEO detectors was an orbiting infrared telescope operating at the right wavelengths to look for the signature of such dark objects.

        As Doug notes, we have’t found 100.00% of them, but we do have the statistics to say that we have found >95% of them. The postulated remainder are mostly ones whose orbits are inconvenient for observation from Earth; any time they are close to us in space they are also close to the sun in angle and so tend to get washed out of the images. To get to 100% detection of near-Earth asteroids, you’d want to put a telescope somewhere in the vicinity of Venus looking outwards. There are proposals to do just that, but not yet flight hardware.

    • John Schilling says:

      The long-period comet threat is roughly one impact per 150 million years, but those impacts are the really bad ones – big, fast, and little prospect of detecting them in time to do anything but try and evacuate the continent or basin it’s going to hit. Typically 1 mile / 1.6 km diameter, 50-60 km/s impact velocity, and ~100,000 megatons effective yield.

      Whether we “need” to be scared of such things is perhaps debatable, but there doesn’t seem to be much point.

      • Psmith says:

        Both this and the parent comment are very interesting stuff. There is no great stagnation.

        Can you and/or Doug M. elaborate some on the long-period comets? Once every 150 megayears vs. none verified and possibly none ever?

      • Andrew says:

        John- you say 1 per 150 megayears, but Doug said never in Earth’s history (so at least a few dozen times rarer?). What’s the dispute, here?

      • John Schilling says:

        There are definitely large impacts – and we’re still mostly talking an order of magnitude smaller than dinosaur-killers – in Earth’s history. More than 1/1.5E8 years. But unless they are very large or very recent, we usually can’t identify the source of the impactor. So no known long-period cometary impacts.

        But we’ve been watching the skies long enough to get pretty good statistics on the frequency and orbit distribution of long-period comets, and we know how big a target the Earth presents. If the last couple of centuries are statistically representative of geological time, and if the fine details of comet trajectories are essentially random or otherwise uncorrelated with Earth’s orbit, the impact rate comes to ~6.5E-9 per year.

        • Doug M. says:

          There’s a minority of impacts where we’re pretty sure, or reasonably sure, that it was an asteroid. There’s a majority of impacts where we don’t know what it was. There isn’t a single impact that we’re reasonably sure was a comet.

          Cometary impacts should have a somewhat different signal. Comets are mostly ice, so they should produce weaker isotopic and abundance anomalies, or none. That’s stuff like the iridium layer, or the layer of chromium with an extraterrestrial isotope signature. They shouldn’t leave an impact field (actual chunks of extraterrestrial material lying around) which asteroids sometimes do. They hit faster, releasing more energy and more heat, so they should produce a somewhat different mix of impact minerals — impact diamonds like the ones found (and mined) at Popigoi crater in Russia, weird stuff like stishovite which is a kind of super-quartz that can only form at very high pressures and temperatures, that sort of thing.

          That said, if the frequency of long-period cometary impacts really is around once every ~150 MY, then it’s not really surprising that we haven’t found a smoking gun. ~70% of the Earth’s surface is ocean basin, and ocean basins are geologically young — ocean crust generally gets subducted within a couple of hundred million years. You’d need to find an impact crater on relatively old, stable, near-surface rock — probably on a continental craton (core), someplace like Nebraska or Siberia. Such a thing may exist; we just haven’t spotted it yet. (Yes, we are still finding impact craters. All the time. They’re not necessarily obvious.)

          Doug M.

          • Doug M. says:

            A thing to keep in mind: the study of impact craters is still really young. I’m living in Tajikistan at the moment. There’s a large, suspiciously circular lake in the Pamirs a few hundred km east of me. It’s called Lake Karakul and it just recently (2013) got positively identified as a meteorite crater, with an age tentatively estimated as somewhere in the Miocene, 15 to 30 MYA. Those error bars are huge because the French team of geologists only had a couple of weeks to work. Lake Karakul is incredibly isolated, and there are maybe sixty days per year when it’s accessible by car — winter, it’s snowed in for months, and then in spring there are floods and landslides. (The local inhabitants consist of a single village and some very hardy nomads.)

            Lake Karakul is clearly a very large and important impact site, but we’ve only barely glanced at it. The French want to come back and do more detailed work, but the last I heard they were still waiting for a grant to come through. Maybe next summer…

            And then there’s stuff like the Eltanin crater and the “Fried Egg Formation”, which are seabed impact craters. (Well, probably. Nobody is completely sure about the Fried Egg Formation.) They’re under thousands of feet of water, and the Eltanin crater is also near Antarctica, under a chilly bit of ocean that’s regularly afflicted with horrible weather and lots of icebergs. So, detailed studies are not going to happen soon.

            TLDR: the study of Earth impacts is very much still an open field, with a lot left to learn.

            Doug M.

        • Tybalt says:

          “But we’ve been watching the skies long enough to get pretty good statistics on the frequency and orbit distribution of long-period comets”

          There is an active school of thought that complicates the picture:
          ‘Clube and Napier suggested that the outer planets occasionally divert giant [centaurs] (more than 50 kilometers in diameter) into the inner solar system into short-period orbits. Debris from the resultant disintegration of these giant [centaurs] can adversely affect the environment of the Earth. Dusting can block sunlight, resulting in globally cooler conditions. Impact events in the super-Tunguska class may result in not only heavy localized destruction but also the occasional “impact winter” or dust veil with global climatological effects.’



          • John Schilling says:

            Last time I checked, the Tunguska impactor itself was believed to be a fragment of a short-period comet, though the community may have changed its mind about that again.

            More generally, if the current population is typical, short-period comets are a more numerous threat over geologic time than long-period ones. Being perturbed by e.g. Jupiter into a short-period orbit is unlikely, but if it happens it gives a potential impactor many chances to hit the Earth vs. one for a long-period comet. It also tends to boil off the volatiles and break up the core, resulting in more numerous but smaller impactors.

            But the original discussion was about things we need to be scared of today. And today, we have a pretty good idea where all the large short-period comets are because we’ve made a hobby out of looking for them for rather more than a short period. Right now, there aren’t any big ones in potentially Earth-impacting orbits. In ten thousand years, we’ll have a new population of short-period comets and we’ll have to reevaluate that.

            In ten thousand years, I’m hoping that evaluating the entire population of short-period comets in some quaint little solar system will be something an enterprising Boy Scout can do as a civic-service project.

  58. Liskantope says:

    This post reminds me of a closely related question: assuming for the sake of argument that terrorist attacks or school shootings cause fewer deaths than something completely mundane that nobody worries about, should we go out of our ways to take measures to prevent them? This came up some months ago when there was a big argument on rationalist Tumblr where some people quoted statistics showing that a kid in school is more likely to die of a heart attack than a shooting. This was presented as an argument against gun control.

    My opinion is that this argument is not very effective. We already do a lot to improve children’s physical health, and there’s not that much more we can do in the sense that AFAIK there is no proposal on the table for this cause that enough people agree on. There are, on the other hand, many proposals on the table for the goal of reducing mass shootings, which are mostly just variants on one idea. People may disagree as to whether they would be effective at making people safer from guns at school or whether they might make things more dangerous, and that’s a debate that we should have. But I don’t see the fact that something else causes more deaths as a reason why people shouldn’t push particular strategies for preventing gun deaths, if they have what they think are good ideas. In general, “X causes less harm than Y” isn’t a good argument for not putting resources towards combating X, unless it directly draws resources away from the fight against Y. And in cases like this, where there is no clear path in the fight against Y, I don’t think we have to worry about resources being taken away from it.

    Take the oft-cited example of airplane crashes versus car crashes, for instance. I don’t believe that the main reason that we put so many resources towards making planes safer is primarily because plane crashes are more upsetting or the idea of flying is inherently scarier to the public. I believe that the primary reason for so many airline regulations is that we can make them and enforce them for commercial flights, while we are much more limited in what we can do to make private car rides safer (although there are some things we can do cars themselves or with driving laws, and we have gone out of our way to do them). The fact that cars are more dangerous than planes is irrelevant — because people have seen specific measures we can take to make flying easier, it makes sense to go ahead and take them.

    • Macbi says:

      Are aeroplanes more regulated than cars?

      I think the reason they’re safer (per passenger mile) is that between taking off and landing they travel in the air where there are very few obstacles. They also travel a long way and carry a lot of passengers.

      I agree with your point that “low hanging fruit” regulations should be enforced whenever we see them. However I suspect that fewer regulations are low hanging fruit that you would think. For example the life jackets in aeroplanes have never saved anybody’s life, but the crew has to go round every morning checking that they’re all still there (people steal them sometimes). If they didn’t have to do this air tickets would become slightly cheaper, more people would choose flying over driving, and lives would be saved.

      EDIT: My mistake, life jackets did save lives on ALM Flight 980, which makes this one of those difficult cases where you actually have to calculate if the regulation is worth the cost.

      • Liskantope says:

        I would say that commercial flying is far more regulated than driving. The pilot is subject to just as many rules as a car driver, and probably loses their job if they break them; there is no way to enforce driving laws this well. Also, there are more regulations for everyone involved in commercial air travel than could possibly be enforced for a private activity (e.g. driving one’s car to work). Yet even other forms of commercial have nowhere near this level of regulation; when traveling by bus or even by train, I’ve never had to put my luggage through an X-ray machine, turn off all electronics and stow my backpack in front of me at certain times, etc., and I’m sure the drivers are subject to fewer rules and looser procedures as well. Still, I don’t know exactly how much this contributes to the low rate of death by plane travel, and the reasons you gave are obvious factors as well.

        As for your point about life jackets, I suspect that nearly every rule introduced for commercial aviation arose from at least one accident where the rule might have saved some lives. But I would also say that some of this regulation has gone too far. For instance, the rule about keeping all electronics off during take-off and landing always seemed ridiculous to me (if I understand the relevant science right), although it came as a response to a single incident where someone using their mobile phone interfered with communications. Most airlines have only just recently begun to ease that rule back.

        (Nitpick: I don’t find it immediately obvious that removing safety measures like life jackets would increase demand for air travel, because it might make some nervous fliers more nervous… but probably not enough to overcome the power of lowered prices, and that wasn’t your main point anyway.)

        • Snodgrass says:

          I’ve had my luggage X-rayed before getting onto trains a few times (not a surprise for Eurostar, but a bit of one for the slow Malaga-to-Seville service … maybe the goal there is to prevent people bringing weapons onto the crowded station concourse, but the crowded station concourse adjoins a no-less-crowded large shopping centre).

      • John Schilling says:

        Are aeroplanes more regulated than cars?

        Yes, by far. Particularly for airliners, the only type of airplane most people will ever fly in, but even a four-seat Cessna and its pilot are subject to a much stricter regulatory regime than any automobile. Starting, on the aircraft side, with the requirement for an airworthiness certificate. There is no equivalent on the automotive side. There are requirements that an automobile has to meet, and the manufacturer can be fined for not meeting those requirements, but all they have to do is meet the requirements (or not get caught breaking them). The airplane manufacturer has to meet the ( requirements of FAR Part 23, and has to prove by test, analysis, and inspection, to the FAA’s discretionary satisfaction, that they meet those requirements. For the simplest airplanes, built in accordance with established practices, this takes years and costs tens of millions of dollars. The Tesla, were it an airplane, would probably still be awaiting approval while the FAA tried to figure out how to evaluate the safety of the battery pack.

        The airplane then has to be re-inspected every year, or every hundred hours if used in commercial service, to verify that it continues to meet those requirements. A process that bears no resemblance to any “safety inspection” your state might require for automobiles.

        Pilot training standards, even for private pilots, are much higher than for drivers. As are medical requirements; an actual physical exam by an FAA specialist is required. Both skills and health are re-tested every two years for private pilots; airline pilots have a very thorough medical exam every six months.

        Basic regulations like speed limits (yes, they exist) are Not To Be Broken. There is no tolerance for “Oh, I got a ticket, happens to everyone now and then”; a single recognized infraction is cause for a suspended license.

        As I mentioned I think in the latest open thread, the government recently relaxed parts of the regulatory regime for the smallest airplanes, no more than 1320 pounds, to something more automobile-like where as long as the airplane is built to industry best practices the manufacturer doesn’t have to specifically prove compliance to the FAA and as long as there’s nothing known to be wrong with the pilot’s health he doesn’t need the special FAA medical exam. Airplanes flown under this regime are about twice as likely to crash as similar airplanes flown under the traditional regulations, and I believe somewhat more dangerous per mile or per hour than automobiles. But, being the smallest airplanes, the danger is almost exclusively to the pilot and at most one passenger. The FAA’s mandate is specifically to protect innocent bystanders, which includes people who buy passage on a commercial transportation service but not necessarily people who go for a ride with Uncle Bob in his little flying contraption.

        • Rosemary7391 says:

          I never knew that you didn’t have to get your car tested for roadworthiness in parts of the USA. You do in the UK (once per year after it’s 3 years old). https://en.wikipedia.org/wiki/MOT_test

          I’d still agree that air travel is more regulated that car travel despite that. It’s just funny how little things that you assume must be true in other superficially similar places actually aren’t.

      • bean says:

        Are aeroplanes more regulated than cars?

        Yes, in the same sense that drugs are more heavily regulated than chairs. I work for a manufacturer of airliners developing repair documentation, and the regulatory side of things takes up the majority of my time. I’ve heard that it’s easier to put something in a person than it is to put a part on one of our airplanes. I’m not 100% sure that’s true, but the two are definitely of a similar magnitude. It used to be said that you knew when an airliner was ready, because the documentation weighed more than the airplane. These days, it’s worse than it used to be because we’ve gone all digital, and they haven’t worked out an appropriate replacement mechanism.

        That said, there has been serious debate over the oxygen system on airliners. It’s a hazard (improperly handled canisters have killed people), and so far as we know, hasn’t really saved anyone. But the public isn’t particularly interested in cost-benefit, so it stays.

    • TomFL says:

      It’s cost efficient to make airliners safer than cars. Doubling the cost of cars and road infrastructure taxes will increase car safety, but pricing a car outside of 25% of the population’s ability to buy is a non-starter. Society has decided to accept the death tolls for the benefit.

    • bean says:

      I believe that the primary reason for so many airline regulations is that we can make them and enforce them for commercial flights, while we are much more limited in what we can do to make private car rides safer

      This is true, but I think it’s only part of the story. The other part is exactly the terrorism vs chairs issue. Big airliner crashes happen a couple times a year, infrequently enough that they make the news. Car crashes happen every day, and only make the news when someone famous was killed in one.
      (Of course, at least in the US, everyone in the industry is really concerned about safety. There aren’t really good alternatives to cars if you’re worried about safety, but if the public starts to worry about airline safety, they’ll drive instead and the industry takes a big hit.)

  59. Petter says:

    > because terrorists kill fewer people each year than falling furniture, and nobody cares about that

    I don’t think that is true. IKEA and many other places sell safety things that you use to fasten your heavy bookshelves to the wall. I think this is a big concern for most parents with small children. When you visit the pediatrician, they constantly ask you whether you have secured your home. :-p

  60. Jack V says:

    I’m a bit torn.

    On the one hand, even if no-one had EVER been killed by terrorism, it makes sense to have some precautions against it, because knowing there’s nothing to stop you makes it more likely that someone who wants to do that, actually will. So I don’t think a straight-up chairs-vs-terrorists comparison is meaningful even if it comes out the way proponents intend.

    But I also think september 11 was unusually bad, and we should expect less terrorism, not another even-worse event to come along eventually. Or maybe, that running around without any shoes on killing people at random is not actually helping and maybe we need to do better at removing the causes of terrorism, and less at polarising pro- and anti- muslim sentiment.

  61. Maybe tangential and not really making a specific point but.

    Until the November 2015 attacks in France, the deadliest terrorist attack in France was the bomb attack and derailment of a train in january 61, which left 24 people dead (or 28, depending on the source).

    This was perpetrated by the Organisation Armée Secrète (OAS), a far right paramilitary organisation which at its height might have counted as much as 3,000 members, and was strongly opposing giving independence to Algeria. They also commited several assassination attempts (two of which failed only by an inch) against then president Charles de Gaulle.

    For a long time I thought the first attack I mentioned was an outlier, in the same pattern as recent islamic terrorist attacks. But after some research, it turned out that isn’t correct at all.

    Apparently the OAS commited *hundreds* of both blind and targeted attacks over a very small period of time; and while most of these attacks had only or two victims, a *conservative* estimate of the death toll attributed to the OAS reaches 2,200 deads for just the period going from january 61 to july 62 (some sources contend that the actual death toll might be significantly higher, since many of the victims were kidnapped and assassinated in remote locations with no claim being made by the OAS; French historians are still working through archives and testimonies at this point).

    And while most attacks where assassinations against a single person, a significant number still made over a dozen victims each time. Futhermore, after some research, it turns out that this is only the most deadly attack commited by the OAS on what is *now* French territory, but they apparently commited much deadlier attacks in Algeria, back when it was still French; it turns out some of those may have reach over a hundred dead.

    (comment abruptly ends here)

  62. Galle says:

    Our homework for this thread, clearly, is to come up with a semi-plausible chain of events that would result in the furniturepocalypse.

    • Chalid says:

      Probably a lot of people would die from falling furniture and shelves and the like in a really big earthquake. So we’d just need to convince whoever classifies these things to categorize these as “furniture” deaths as opposed to “earthquake” deaths. (“Cause of death: failure to use wall mounts.”)

    • Macbi says:

      Internet of things + hacking

  63. TomFL says:

    There ARE NOT universal methods for signal analysis. Sorry. Just stop trying. The best method to interpret data is to analyze the signal and noise, understand the underlying causes and pick the appropriate analysis method for your intentions. There is understanding what happened, and then predicting what will happen. Interpolation is OK, extrapolation is dangerous.

    I find just looking at the raw data tells me 90% of what I need to know. No statistics necessary. Any analysis that use sophisticated statistics is immediately suspect in my opinion. What drives me batty is when a “surprising” result occurs from an analysis using a single somewhat crazy data processing method. Perhaps the analysis is optimal, but I’d like to see the data analyzed using standard methods and it made clear why special methods were deemed necessary.

    People with an agenda choose analysis methods that suit their intended result. How many times have you seen a doomsdayer model an almost linear trend to an exponential whose knee is just around the corner and start yapping about near term tipping points.

    In terrorism it is the black swan that counts. The underlying data shows that people will intentionally commit the black swan event if given the resources, so it makes sense to try to limit their access to these resources. An important distinction is that these are not random physics driven events.

    People are too paranoid about terrorism generally, but those who dismiss it are deluded if they think we shouldn’t be looking hard for people trying to access biological weapons or nuclear material. The biological weapons the USSR had at the end of the cold war were terrifying, tons of weaponized vaccine resistant smallpox among others. They may or may not have disposed of it all, but there are people who certainly haven’t forgotten how to do it again.

  64. The Sprat says:

    Conservatives are familiar with this problem, because every six months or so, some wiseacre blogger or “journalist” will cook up a version of that first chart, except instead of terrorism vs. chairs, it’s Islamic terrorism vs. right-wing terrorism. And not only do these charts all strangely begin right there on 9/12/01, but if you dig deep into the data they also often count *ALL* non-Islamic terror attacks/mass killings as “right-wing.”

    (Also strange: when discussing the successful terrorism-prevention records of U.S. Presidents, 9/11 is suddenly included again.)

    Another confounder is relative lethality: there are many millions of chairs inside the U.S. [citation needed], but a comparatively low number of terrorists at any given time.

  65. James says:

    You’re trying to find a way to make an average meaningful in a system that’s designed to defy such things.

    The number of people killed via furniture is going to be roughly constant, because people (in aggregate) and furniture are roughly constant; any variation from the average is going to be small, and almost certainly random. Chairs are chairs; there’s nothing much that’s going to happen to make them more or less hazardous.

    Terrorism, in sharp contrast, relies on differences between normal life and the attacks. In areas where violence is the norm, terrorism gets no (or poor) results–during the events in the Ukraine, when Russia invaded/intervened (depends on your perspective, the info I got from folks in the country was too muddy to interpret), someone setting off a bomb would have gone unnoticed. In Israel it doesn’t disrupt life too much for anyone not affected by it. What’s one more explosion? It’s only in areas where violence is the exception, and a drastic one at that, where terrorism has any potential to make an impact. A bomb that kills 5 people in the USA makes the national news, because that’s probably the worst bomb attack that year. Terrorism is showmanship, and it’s the contrast that makes it work.

    Thus, looking at the average number of deaths due to terrorism is a nonsensical way to look at it. The most effective terrorist activities will be in areas where the average number of deaths per year due to terrorism is fairly low. I’m not entirely certain what statistical analysis would be superior, but the average is quite obviously the wrong choice. An average of deaths per attack (including and excluding those thwarted or which fail due to other reasons) could be an option: it captures the relevant data about how dangerous an attack is, while simultaneously providing a picture of how likely an attack is going to be to succeed.

    Further, furniture isn’t actively attempting to harm anyone (despite what you think at 2 am when your child wants a glass of water and you’re too lazy to turn the light on). Terrorists are. The risks associated with furniture are passive; they are known, and can be accounted for. The risks posed by terrorists are active, and anti-inductive: terrorists are actively trying to circumvent our measures to stop them, meaning that they are hoping to achieve a level of unpredictability impossible in inanimate objects. This element is completely ignored in these types of calculations; we’re supposed to view the two situations as perfectly equivalent, when in reality they are vastly different in kind.

    I have never been shot at by a hunter. More people have been injured in my family due to trees and tree limbs falling on them than by hunters. But when I’m in a woods I worry about the hunter I came across and not the trees, because the hunter can ACTIVELY CHOOSE to harm me. The tree can’t. And that matters from a threat-assessment perspective.

    As an aside, I’m highly dubious about the utility of averages in statements about giant asteroids and the like. The variance is simply too high (and the definition of “giant” is simply to great) for the number to mean anything. Plus, the vast majority of the rock older than the Jurassic is gone, and most of what’s left has never been thoroughly analyzed. Sea floor recycling has destroyed the majority of the sea floor since before the Jurassic (some Triassic stuff is left in the oceans, and some older stuff is on continents for various reasons). That’s a lot of data missing from these calculations, particularly considering the fact that we should expect giant asteroids to decrease in number through the lifespan of the solar system. Cyclicity in mass extinctions is even worse. We have very few data points, and even fewer once you look at originate rates and realize there are two distinct kinds of mass extinctions. All of this stems, far as I can tell, from a desire to fit past events in life’s history to popular theories. Once the Alvarez Hypothesis was confirmed everyone wanted to tie everything in life’s history to impacts, whether it fit or not. Basalt volcanism went through the same process. Global warming is now the pet one (I’m not saying anything about CURRENT global warming–I’m saying that people are ignoring everything but global temperature change when studying past events in life, just because it’s popular today). And I reject the notion that potential effects can be used to calculate actual averages in any case; it’s equivocation, pure and simple.

  66. TomA says:

    Scott, you should copyright “Furniturepocalpse” before some Hollywood screenwriter turns it into the next Sharknato hit series.

    • Deiseach says:

      Gives a sinister twist to all those cartoon deaths where guys get smashed by falling pianos – the pianos are doing it on purpose! 🙂

      de Maupassant has a story about sentient or somehow mysteriously animated furniture:

      I continued to listen for some time longer. I could distinguish now an extraordinary pattering upon the steps of my grand staircase, on the waxed floors, on the carpets, not of boots, nor of naked feet, but of iron, and wooden crutches, which resounded like cymbals. Then I suddenly discerned, on the threshold of my door, an arm chair, my large reading easy chair, which set off waddling. It went away through my garden. Others followed it, those of my drawing-room, then my sofas, dragging themselves along like crocodiles on their short paws; then all my chairs, bounding like goats, and the little footstools, hopping like rabbits.

      Oh! what a sensation! I slunk back into a clump of bushes where I remained crouched up, watching, meanwhile, my furniture defile past, for everything walked away, the one behind the other, briskly or slowly, according to its weight or size. My piano, my grand piano, bounded past with the gallop of a horse and a murmur of music in its sides; the smaller articles slid along the gravel like snails, my brushes, crystal, cups and saucers, which glistened in the moonlight. I saw my writing desk appear, a rare curiosity of the last century, which contained all the letters I had ever received, all the history of my heart, an old history from which I have suffered so much! Besides, there was inside of it a great many cherished photographs.

      Suddenly — I no longer had any fear — I threw myself on it, seized it as one would seize a thief, as one would seize a wife about to run away; but it pursued its irresistible course, and despite my efforts and despite my anger, I could not even retard its pace. As I was resisting in desperation that insuperable force, I was thrown to the ground in my struggle with it. It then rolled me over, trailed me along the gravel, and the rest of my furniture which followed it, began to march over me, tramping on my legs and injuring them. When I loosed my hold, other articles passed over my body, just as a charge of cavalry does over the body of a dismounted soldier.

  67. Montfort says:

    Gravatars are based on emails, right? Does anyone have a guess why mine appears to have suddenly changed? I’m reasonably sure my email is both entered correctly and unregistered.

  68. lambdaphagy says:

    This seems to be a hobby horse of the Post’s. Here, for example, is a graph showing us that in recent years, Islamic and right-wing terrorism have been of roughly equal concern. Beginning from the nice round start date of 2002.

    • Sandy says:

      I don’t get how they use data like this to say “The problem of right-wing terrorism is just as bad as the problem of Islamic terrorism!” when Muslims are only about 1-2% of the population — proportionately, if Islamic terrorism in the US is as bad as right-wing terrorism, the Islamic terrorists are punching far above their weight.

    • ThirteenthLetter says:

      Why, it’s almost as if they’re trying to push a narrative instead of inform the populace.

  69. Steve Sailer says:

    I think people are wired to worry about violence more than about other threats. I also think there may be good reasons for this.

  70. AL says:

    Terrorism often has a particularly strong effect on people’s behavior, and that effect tends to be particularly destructive on a collective/societal scale.

    Counting immediate terrorist kills as the only relevant metric fails to account for these other costs, which may be very significant indeed.

    • Saint Fiasco says:

      Aren’t more than half of those effects caused by the victims (over)reaction?

      We may be better off pretending that the immediate kills are the only consequence.

  71. sabril says:

    Professor Jonathan Katz (of “in defense of homophobia” infamy) had some interesting things to say about this issue:

    Why Terrorism Matters

    The terrorists of al Qaeda killed 2998 people on September 11, 2001. This is fewer than the number of Americans killed every month in highway accidents and about the same as the estimated number who die every ten days as a result of medical mistakes. Yet we consider terrorism to be an urgent national problem and an important threat, justifying a pervasive range of security measures of dubious value ranging from the inconvenient (long lines at airport security checkpoints) to the humiliating (public pat-down searches of people who could not conceivably be terrorists, such as old people in wheelchairs and small children) to the frankly ridiculous (it’s now illegal to take photographs in the New York subway). In contrast, automobile accidents and medical errors are accepted as inevitable features of everyday life, justifying only modest efferts to reduce their frequency.

    He goes on to hypothesize that terrorism matters because its an assault on national honor. I do not totally agree with him, but of course I think he is right that there is a lot more at issue than just the body count.

    • Squirrel of Doom says:

      To me the difference is that terrorism is done by an intelligent force that aims to kill us all if we don’t stop them.

      Falling furniture, even if completely ignored, will not methodically escalate like that, since it is a mindless force.

      • sabril says:

        To me the difference is that terrorism is done by an intelligent force that aims to kill us all if we don’t stop them.

        I agree that’s a big part of it, but I think there are other factors too.

        Here’s a thought experiment: Suppose there was some weird foreign religious cult which had been around for a hundred years and it had a practice of killing one and only one American per year. The cult has no aim to kill all Americans or even many Americans. Just one a year.

        I think under those circumstances, most Americans would still agree that it’s worth spending a lot of resources to stop this cult. Far more than would be spent just preventing an additional fatal car accident. Why is that? For one thing, most of us would agree that the world needs to know that there are serious negative consequences when you target Americans for killing. It’s not something that we will just ignore. For another thing, honor. But those are two very similar things.

  72. Ben says:

    Thin tails v fat tails

  73. Jason says:

    There is an unspoken element to all this, which is “what problems can we sensibly and efficiently protect against?”

    AI people, terrorism people, and earthquake people believe they know, if not the wingspan and beak size of the coming black swan, at least the direction it’ll fly in from.

    Furniture and flu people are more implicitly circumspect. Their assumption is of an infinite number of potential black swans against which we cannot reasonably protect ourselves. Against this background of meta risk they propose we may as well do what we can about what we can. It would be silly for a person who lives next to a volcano to drive round with their seatbelt undone, for example.

  74. Squirrel of Doom says:

    In defense of the Haitians, the last major quake in Port-au-Prince was in 1770.

    So I don’t know how much you can – without hindsight – blame them for not spending a lot of effort preparing for such an unlikely disaster, when they’re always dealing with tons of actual disasters.

  75. Krzys says:

    The difference is very simple:
    The chair fatality distribution is known. The terror distribution is unknown with high possible peak. That’s what makes all those comparisons idiotic.

    • Murphy says:

      For terrorist attack n is 156,772

      How big a sample do you think you need to be reasonably confident about the distribution?


      I’m wondering whether there’s even 156 thousand recorded chair deaths. Due to better record keeping it seems likely we might have better data about the distribution of terrorist attacks.

      • Protagoras says:

        If, as Scott reports, 50 people a year die from falling from chairs in Britain alone, that puts it in the ballpark of 1 per million people per year. Assuming British people aren’t unusually susceptible to fatally falling off chairs, extrapolating that to the world as a whole easily gets you several times more chair deaths over the years than terrorism deaths (also assuming that chair deaths haven’t gone up over time). Admittedly, the former assumption is perhaps questionable, and the latter assumption definitely is (population of older people has definitely gone up over time, and amount of time people spend in chairs also seems to have), but neither seems likely to be so far off as to leave chair deaths below terrorism deaths.

        • Jiro says:

          If someone dies in falling off a chair, but they also have a medical condition (or just old age, which is a combination of medical conditions) which makes them unusually susceptible to death from falling off chairs, it isn’t fair to count that as 100% a “death caused by falling off of a chair” and 0% a “death caused by a medical condition”. Your figure of 50 per year is, because of this, drastically overstated.

          Also, deaths which happen in small quantities can have high percentages of misreporting. How many of those deaths caused by falling off of chairs are deaths caused by something embarassing (death by auto-asphyxiation), or even suicides or mob hits, and they’re just officially reported in vague terms as “death caused by falling off of chair”?

          (And if those two factors have an efect, it might not generalize to non-developed countries or countries with different cultures.)

  76. Yehoshua K says:

    Of course there will be a successful mass-casualty attack that dwarfs 9-11, sooner or later. There are a lot of would-be terrorists plotting, plenty of them are plenty smart, and they have a lot of resources backing them up. It is not plausible at all that America and similar countries will defeat every single attempt at a mass-casualty attack forever; therefore, it is not plausible that there will never again be a mass-casualty attack. Sooner or later enemy luck + skill + resources will beat friendly luck + skill + resources. Isn’t this obvious?

  77. Troutwaxer says:

    These numbers only apply to the U.S. I’ll line ’em up and get back to furniture and terror in a minute.

    We lose about 2.5 million people every year to all causes of death. We lose approximately 600,000 each year to heart attack and another 600,000 each year to cancer. Lower respiratory diseases kill about 150,000/year, with similar numbers for accidents and stroke. We lose about 70,000 people each year to diseases they caught while in a hospital, 45,000/year to car accidents specifically (as opposed to all accidents above) and 30,000 each year to deaths by gun (2/3 of those are suicides.) We lose about 7,500 each year because someone took an over-the-counter pain reliever to which they were allergic or otherwise incompatible. (I’ve never heard of anyone dying this way.)

    The first thing we can determine when we discuss terror and furniture deaths in the light of this data is that deaths from both causes are minuscule as a percentage of the whole. The very worst set of terror deaths in any given year, 9/11, killed approximately 1/3 of the people who would die that year due to purchasing the wrong over-the-counter pain reliever…

    Taking an average over time, say from 1996 to 2016, (leaving in 9/11 just to be fair,) we usually see 0-10 deaths in the U.S. every year which evaluate to terrorism. These are usually domestic and done by right-wingers. We have two huge outliers to consider in the U.S., which would be 2001 and this year’s slaughter of 49 people at a nightclub in Florida. Most years we don’t see more than 10 deaths due to terror, and frequently we see less than 5 deaths due to terror. Consider the following website, particularly Table 2. (Note that in 1999 I think the author transposed U.S. deaths and Israeli deaths by terror. I think I’d remember an incident where 218 people died.)


    3161 people were killed by terror attacks between 1996 and 2016, which means we lose an average of 158/year due to terror attacks, even if we leave in the outlier. If we don’t leave in the outlier – the 2998 people killed on 9/11 – we end up with an average of 8.1 people killed by terror each year.

    Which means that furniture wins when considering the average number of deaths each year.

    If we take it year-by-year instead of considering an average, and we assume that furniture ONLY kills 10 people a year, chairs still killed more people than terrorists in 1996 – 1998, 2000, 2003-2008, and 2010-2014, giving furniture a 15-5 margin of victory during the last 20 years.

    Someone call Donald Trump. We need to deport all the Furniture. And we’ll make Ikea pay for it!

    • Glen Raphael says:

      we usually see 0-10 deaths in the U.S. every year which evaluate to terrorism. These are usually domestic and done by right-wingers.

      Got a source on that last bit?

      • Troutwaxer says:

        There’s a decent summary here:

        It lists terror attacks by ideology at the bottom. When I talk about right-wing attacks I’m lumping together the Right-Wing, Anti-Abortion, KKK, and Anti-Semitism as being generally right-wing/conservative stances.

        That being said, it depends very deeply on what time period we’re talking about. In the past twenty years, even excluding 9/11, the Islamists have committed far more murders and more attacks generally than the right wingers, though the number of killings changes in the right-winger’s favor if we go back one more year and include the Oklahoma City bombings.

        So I’ve had to change my position a little…

        • I went through the Wiki list for 1990-1999. About two thirds were arguably right wing, mostly anti-abortion. For the previous decade it looked as though only a minority were, but I didn’t do as careful a count.

          For the 2000-2009 decade, I only counted one right wing incident (anti-abortion) out of about twenty. Several were by environmentalists, others by pro-Palestinian/anti-Israeli. In current politics, that’s a left wing position not a right wing position.

          So I don’t think the evidence supports the claim that terrorist deaths are usually done by right wingers. To be fair, I was counting terrorist incidents listed by Wiki, not all of which led to deaths.

      • hlynkacg says:

        The standard MO is to use count Muslims and school shooters in the Klebold mold as “right-wingers” for these purposes. So yes, most of the deaths from terrorism in this century have been caused by “right-wingers”.

        Edit: seems I’ve been ninja’d though I don’t see how adding another 168 deaths to the right’s total “tips the balance”.

        • Troutwaxer says:

          When I’ve written Right-Wing I’ve meant Right-Wing Domestic terrorists, not Islamic Terrorists. If we discard our outlier (9/11) and add the further 168 deaths to the column for Domestic Right-Wing terror, we get a lot more deaths in that column. The problem is that this is not strictly fair.

          My personal take on Islamic terror is that in the last twenty years or so we’re mainly seeing terror undertaken in response to Wahabi propaganda from Saudi Arabia, so I tend to see the whole thing as an undeclared war with the Saudis than some standard notion of “Terrorism.”

          • Titanium Dragon says:

            That’s a flawed analysis. See also: Israel.

            The reality is that Islamist terrorism has been a problem for a long time. It just hasn’t been *our* problem.

  78. Ben says:

    A big criticism of the war on terror is that it does little to actually address black swans instead of wasting everyone’s time with shoes and bombing the middle East back to the stone age.

  79. hawkice says:

    I don’t mean to be insensitive, but the second chart seems to indicate there were no deaths from falling furniture on Sept. 11th, 2001. Considering how much fell and how (being an office building) some of that was furniture, and that as many as 1,000 of the casualties inside the building died AFTER the towers fell, I find this a less-than-obvious “fact”. Also, depending on your definition of ‘falling’, some of the deaths were likely caused by blunt force of the internals of the aircraft, which, notably, contain furniture.

    This is all by means of saying that these are not mutually exclusive categories. Using categories to describe these events kind of sucks, because we don’t care about categories, we care about which interventions will help the most. But it’s harder to create a chart that says, number of lives that could have been saved by intervention, because it’s more obvious that (1) they definitely ARE NOT mutually exclusive, and (2) no list of interventions is complete. It’s not X vs. Y. Which calls into question why on earth you’d use a bar chart for this kind of data anyway.

  80. Five Daarstens says:

    It’s not just dying of terrorism – Tens of thousands (maybe more?) were involved in 9/11 who survived in the surrounding blocks.
    Tens of thousands were also involved in the first WTC attack in 1993, even though only six died.

  81. Voronoff says:

    Isn’t this the exact form of argument you used in the Reverse Vox post? You looked at the number of deaths caused by chairs and compared it against the historical data for something that has a known black swan structure.

  82. ASK1 says:

    I don’t think the “exclude 9/11” motivation is as cherry-picked as you suggest. The reason for excluding black swans is not that they are not important; it’s that they are empirically un-analyzable, thus lending themselves to use by bigoted know-nothings who instinctively use Terror Management Theory to manipulate the populace (not that I’m thinking of anyone in particular 🙂 ).

    Even experts can stay up late arguing about whether 9/11 should be treated as currently possible, and if so what does it indicate for the likelihood and magnitude of another massive attack. The point (to me at least) is that if you are arguing from historical experience, and leaving out one Hail Mary event makes your argument look stupid, then your argument is probably stupid.