Join 3,501 readers in helping fund MetaFilter (Hide)


The Department Of War Math
June 6, 2013 8:19 AM   Subscribe

You Are Not So Smart: Survivorship Bias, demonstrated through Abraham Wald's work at the Statistical Research Group in World War 2.

A REPRINT OF "A METHOD OF ESTIMATING PLANE VULNERABILITY BASED ON DAMAGE OF SURVIVORS"
BY ABRAHAM WALD
(PDF)
posted by the man of twists and turns (48 comments total) 55 users marked this as a favorite

 
This reminds me of reading somewhere - Pynchon's "V"? - a discussion of porpoises rescuing drowning shipwreck survivors and carrying them to land. The argument was, they are just as likely to carry you out to sea, but you'll never meet a guy in a port town bar who tells you that story.
posted by thelonius at 8:28 AM on June 6, 2013 [20 favorites]


Reminds me of this article in a similar vein about business writing.

This is also something I struggle with as a fair number of my friends and acquaintances seem to fall into the trap of guru-ing their life, moving their cheese and doing all the other stuff that seems to be in-vogue these days. Great post.
posted by Carillon at 8:35 AM on June 6, 2013 [1 favorite]


Nor could people today know as they open umbrellas and twist heels on cigarettes, that nearby, in an apartment overlooking Morningside Heights, one of those soldiers once effortlessly prevented the United States military from doing something incredibly stupid, something that could have changed the flags now flying in capitals around the world had he not caught it, something you do every day.

OK, I get what the author is saying in this sentence, but that tangle of clauses led me to ask myself if I had changed flags flying in capitals or, possibly, stopped the US military from doing something incredibly stupid today. Let me tell you, it took the weight off my shoulders when I finally worked it out.

Fortunately,successful and unsuccessful published sentences are available for review, so we can become better writers.

Other than that, I enjoyed this essay.
posted by GenjiandProust at 8:38 AM on June 6, 2013 [1 favorite]


Thanks for posting this, tmotat. I love the maths.
posted by Mental Wimp at 8:39 AM on June 6, 2013


Awesome awesome article. I'd watch this boring boring war movie.
posted by Potomac Avenue at 8:47 AM on June 6, 2013


The argument was, they are just as likely to carry you out to sea, but you'll never meet a guy in a port town bar who tells you that story.

I think this should be statistically verifiable. Like, what's the probability of meeting dolphins after a shipwreck vs how many (claim to) get helped by them.
posted by DU at 8:50 AM on June 6, 2013


Also this paragraph (not the focus of the article but still) is also a good description of how being born privileged in many ways affects your life.

"Wiseman speculated that what we call luck is actually a pattern of behaviors that coincide with a style of understanding and interacting with the events and people you encounter throughout life. Unlucky people are narrowly focused, he observed. They crave security and tend to be more anxious, and instead of wading into the sea of random chance open to what may come, they remain fixated on controlling the situation, on seeking a specific goal. As a result, they miss out on the thousands of opportunities that may float by. Lucky people tend to constantly change routines and seek out new experiences. Wiseman saw that the people who considered themselves lucky, and who then did actually demonstrate luck was on their side over the course of a decade, tended to place themselves into situations where anything could happen more often and thus exposed themselves to more random chance than did unlucky people. The lucky try more things, and fail more often, but when they fail they shrug it off and try something else. Occasionally, things work out."
posted by Potomac Avenue at 8:52 AM on June 6, 2013 [11 favorites]


This blog is awesome. Definitely buying his book when it comes out. Soulful statistics are the bomb.
posted by Potomac Avenue at 9:01 AM on June 6, 2013


Interesting stuff. Explains the thinking behind Failfaire and various journals of negative results.
posted by gottabefunky at 9:02 AM on June 6, 2013 [2 favorites]


fall into the trap of guru-ing their life, moving their cheese and doing all the other stuff

"Moving their cheese" is a fantastic metaphor and it frustrates me that I don't know what it means. Can someone spell it out for me?
posted by notyou at 9:06 AM on June 6, 2013


"Moving their cheese" is a fantastic metaphor and it frustrates me that I don't know what it means. Can someone spell it out for me?

I believe it's just a reference to Who Moved My Cheese.
posted by Steely-eyed Missile Man at 9:08 AM on June 6, 2013 [1 favorite]


"Moving their cheese" is a fantastic metaphor and it frustrates me that I don't know what it means. Can someone spell it out for me?

Who moved my cheese?
posted by cosmic.osmo at 9:08 AM on June 6, 2013 [1 favorite]


Thanks for the pointer. I'd be more concerned about who ate my cheese, but I guess it's the same.
posted by notyou at 9:11 AM on June 6, 2013


"Moving their cheese" is a fantastic metaphor and it frustrates me that I don't know what it means. Can someone spell it out for me?

I was recently reading about all sorts of business fables in addition to "Who Moved My Cheese?"
posted by Cool Papa Bell at 9:15 AM on June 6, 2013 [1 favorite]


Here is an easy example. Many people believe old things represent a higher level of craftsmanship than do new things. It’s sort of a “they don’t make them like they used to” kind of assumption. You’ve owned cars that only lasted a few years before you had to start replacing them piece by piece, and, would you look at that, there goes another Volkswagon Beetle buzzing along like it just rolled off an assembly line. It’s survivorship bias at work. The Beetle or the Mustang or the El Camino or the VW Minibus are among a handful of models that survived in large enough numbers to become iconic classics. The hundreds of shitty car designs and millions of automobile corpses in junkyards around the world far outnumber the popular, well-maintained, successful, beloved survivors

Yes! This!
posted by ghharr at 9:16 AM on June 6, 2013 [6 favorites]


what's the probability of meeting dolphins after a shipwreck vs how many (claim to) get helped by them.

I doubt that this actually happens much, it sounds like an urban (oceanic?) legend. I think it was an aside in a novel, not someone seriously advancing an argument.
posted by thelonius at 9:20 AM on June 6, 2013


what's the probability of meeting dolphins after a shipwreck vs how many (claim to) get helped by them.

I doubt that this actually happens much, it sounds like an urban (oceanic?) legend. I think it was an aside in a novel, not someone seriously advancing an argument.


Not to belabor the point, but the argument doesn't really have anything to do with the likelihood of meeting dolphins if you are shipwrecked. What he's saying, is that if some guy at the port says "I saw these dolphins, and they led me back home to safety!" that does not mean that if you see dolphins when you are shipwrecked that you should necessarily follow them. The dolphins (for all we know) could be just as likely to lead you out into the open sea, but these poor people never come back to the port to tell their tale.

Basically, one way of restating the main point is that sometimes bad decisions lead to good results and that by only dissecting the successes, we are likely to incorrectly think that those bad decisions were good., when in fact the person was either lucky or just successful despite the bad decision.
posted by artichoke_enthusiast at 9:27 AM on June 6, 2013 [1 favorite]


Also this paragraph (not the focus of the article but still) is also a good description of how being born privileged in many ways affects your life.
"Wiseman speculated that what we call luck is actually a pattern of behaviors that coincide with a style of understanding and interacting with the events and people you encounter throughout life. Unlucky people [don't wade] into the sea of random chance open to what may come.... The lucky try more things, and fail more often, but when they fail they shrug it off and try something else. Occasionally, things work out."
I don't think so, since the "unlucky" attitude seems like something that anyone could change on their own (with equal ability to anyone else).

It’s survivorship bias at work. The Beetle or the Mustang or the El Camino or the VW Minibus are among a handful of models that survived in large enough numbers to become iconic classics. The hundreds of shitty car designs and millions of automobile corpses in junkyards around the world far outnumber the popular, well-maintained, successful, beloved survivors.

This is also a reason to be biased against new designs, and least until they've proven themselves a little.
posted by cosmic.osmo at 9:33 AM on June 6, 2013


Not to belabor the point, but the argument doesn't really have anything to do with the likelihood of meeting dolphins if you are shipwrecked.

Yes, it completely does.

Let's say the chance of meeting dolphins after a wreck is 20%. And say we know the chance of surviving a wreck (to the point of landing in the water alive) is 40%. So we know that out of 100 wrecks, we'll have about 8 people meeting dolphins.

If out of 100 wrecks we find ~8 people that say they were saved by dolphins, we know that the hypothesis "the dolphins could be just as likely to lead you out into the open sea" is false. They are in fact much more likely to save you.

But we can only figure that out if we know how often shipwrecked persons meet dolphins.
posted by DU at 9:33 AM on June 6, 2013 [3 favorites]


I have maintained for years that dolphins must be the most fearsome predators in the sea, on the grounds that when they decide to attack they never leave a single witness alive. All they have to do is lead one deliciously squrimy morsel fellow mammal to shore every now and again, and that's proof that they're lovable.
posted by metaBugs at 9:38 AM on June 6, 2013 [6 favorites]


But we can only figure that out if we know how often shipwrecked persons meet dolphins.

Just often enough, I suspect. Just often enough.
posted by It's Raining Florence Henderson at 9:47 AM on June 6, 2013


There’s a lot of snake oil sales going on. And a lot of well meaning people who won the lottery telling everyone to go buy lottery tickets while financial advisors shake their head.

Yeah ... About those financial advisors ...
posted by ZenMasterThis at 10:00 AM on June 6, 2013


When I worked in the medical device industry, marketing and MBA types kept talking about "best practices." As a statistician, I knew they were making exactly this mistake, compounded by regression to the mean. It frustrated me to no end that when I tried to explain it to them, their response was, "How can you go wrong mimicking the best?" I threw up my hands.
posted by Mental Wimp at 10:13 AM on June 6, 2013 [1 favorite]


When I worked in the medical device industry, marketing and MBA types kept talking about "best practices." As a statistician, I knew they were making exactly this mistake, compounded by regression to the mean. It frustrated me to no end that when I tried to explain it to them, their response was, "How can you go wrong mimicking the best?" I threw up my hands.

Clearly, the class on survivorship bias is right after the class on ethics.

The bit about the lucky and the unlucky reminded me of a passage in Nabokov. Maybe it was in The Gift or one of the short stories. Describing an oaf, he says, "Like all pessimists, he was unobservant," or something to that effect. It stuck out at me because it was a generalization, and no one is more vocal in his hatred for generalizations than Nabokov. To see it illustrated in a study surprises me.
posted by Rustic Etruscan at 10:19 AM on June 6, 2013 [1 favorite]


This was a great essay, especially after the style settled down a bit (the writing was slightly overwrought at first).

marketing and MBA types kept talking about "best practices." As a statistician, I knew they were >making exactly this mistake, compounded by regression to the mean. It frustrated me to no end that when I tried to explain it to them, their response was, "How can you go wrong mimicking the best?" I threw up my hands.

I teach MBAs, and, even more challenging, teach them entrepreneurship, and this is a real problem (and not just for MBAs, for everyone!). It is especially the case in entrepreneurship because we have lots of very solid statistical evidence on entrepreneurial success, and almost all of it directs you away from following Jobs or Zuckerberg. For example, founders who give up a lot of control to partners and high-quality hires tend to become a lot richer/more successful than those that hold onto control. Or that founding teams outperform lone individuals. But these things contradict the heroic success stories that founders tell, and thus people are not inclined to believe it.

In business academia, survivor bias is rarely ignored, but in popular business writing it is everywhere. My favorite business book on the topic is The Halo Effect, which attacks shoddy studies of business success. Also, on entrepreneurship specifically, the Founder's Dilemmas does a similarly good job.
posted by blahblahblah at 10:28 AM on June 6, 2013 [10 favorites]



I think this should be statistically verifiable. Like, what's the probability of meeting dolphins after a shipwreck vs how many (claim to) get helped by them.


OKCupid did an infographic on this. It all depends on how you shot your profile picture and whether you mention survival or rescue as an interest in your profile.
posted by srboisvert at 10:35 AM on June 6, 2013 [2 favorites]


Describing an oaf, he says, "Like all pessimists, he was unobservant," or something to that effect. It stuck out at me because it was a generalization, and no one is more vocal in his hatred for generalizations than Nabokov. To see it illustrated in a study surprises me.

Going from "people who would describe themselves as very unlucky" to "this describes all pessimists" is a bit of a stretch.
posted by Steely-eyed Missile Man at 10:41 AM on June 6, 2013


Yeah, it is. It was an associative jump more than the thought "Aha! He got it right!"
posted by Rustic Etruscan at 10:46 AM on June 6, 2013


For clarification, I don't think pessimists are unobservant.
posted by Rustic Etruscan at 10:50 AM on June 6, 2013


"I think this should be statistically verifiable. Like, what's the probability of meeting dolphins after a shipwreck vs how many (claim to) get helped by them."

Probably strongly correlates with the ability to see the forest for the trees.
posted by Eideteker at 10:50 AM on June 6, 2013 [4 favorites]


Great read... always love this blog, and I guess, like the great sage Donald Rumsfeld, always fascinated by the idea that we don't know that we don't know. Thanks for posting this!
posted by ph00dz at 10:54 AM on June 6, 2013


It’s survivorship bias at work. The Beetle or the Mustang or the El Camino or the VW Minibus are among a handful of models that survived in large enough numbers to become iconic classics.
This. I own a "classic" 1983 VW T3 Westfalia. Other than the chassis and electrical system, it's a fragile maintainance/money pit. (They're also quite safe for the front passengers in most accidents. Too bad about anyone behind them.) However, millions of them were made from 1979 until 2002 when production ended in South Africa and Brazil. They survive as classics because they're ubiquitous, parts are cheap and readily available, and the campers (especially the 4WD syncro campers) are the only vehicle that can do what they do at a reasonable price point with worldwide parts availability. IF one does the frequent, endless maintainance and replacement of everything not welded on. The unlucky T3s are parts in the running ones.
Survivorship bias, indeed.
posted by Dreidl at 11:08 AM on June 6, 2013


My brother and I had a lesson in the survivor effect as young teens. We were big fans of all the British comedy and comedians we saw, living in California. Monty Python, The Goon Show, Peter Sellers. We knew British comedy as sharp, and so much smarter than almost all the junk Americans made. Then we visited the U.K. and eagerly switched on the TV in our hotel room. Result? Dad's Army. Now Dad's Army isn't horrible, but it isn't Monty Python. It took us a while to figure it out: People only bother to export the good stuff.
posted by benito.strauss at 11:16 AM on June 6, 2013 [5 favorites]


The part about lucky/unlucky people really hit home.
posted by tommasz at 11:35 AM on June 6, 2013 [2 favorites]


That blogs whole schtick drives me crazy. Please stop telling me what I see or feel. Statistical likelyhoods ignore individual differences and basing the entire blog on second person statements feels manipulative and disrespectful.

Also the few times he has gone over things that I had studied, I found his understanding shallow to the point of being misleading and his examples were mostly cherry picked and alternate viewpoints excluded.
posted by psycho-alchemy at 12:24 PM on June 6, 2013


Great read, thanks for sharing it.

I too work in an industry where "best practices" is an oft-repeated phrase along with "just do it like $bigsuccessfulcompany does with $successfulproduct," and it's taken me forever to realize there's often -- but not always! -- more to be gained by looking at what didn't work and what customers absolutely hated.
posted by lord_wolf at 12:25 PM on June 6, 2013 [2 favorites]


This whole MBA-can't-see-the-bias thing reminds me of perhaps the most intense moment of clarity I've ever experienced as a human being.

After smoking many bowls with a friend of mine, I asked him whether his cat ever got stoned just from the contact high.

"We'll never know," he said, before proceeding to blow my fucking mind. "We'll never know because, you see, we're stoned, too. We won't be able to really know if he's stoned because our own perceptions are affected."

I recall thinking that was some real Jean-Paul Satre shit right there.
posted by Cool Papa Bell at 1:13 PM on June 6, 2013


Here is an easy example. Many people believe old things represent a higher level of craftsmanship than do new things. It’s sort of a “they don’t make them like they used to” kind of assumption.

I think this misses an important point, though: For the last several decades, even at the design stage, products have often been deliberately designed to fail in ways that prevent maintenance and that ensure consumers will have to make another purchase of the same product within a shorter window of time.

Older things aren't necessarily seen as representing a higher level of craftsmanship than old things based on magical thinking or cognitive bias alone: It's an actual documented reality that built-in obsolescence has taken hold as a deliberate design strategy when it comes to the design of consumer goods in more recent decades.

People think things were crafted "better" (where "better" means to be more durable, reliable and maintainable) because they actually were deliberately crafted to be more durable, reliable and maintainable due to cultural differences in the past--not just because of the effects of survivorship bias.
posted by saulgoodman at 1:35 PM on June 6, 2013 [5 favorites]


People think things were crafted "better" (where "better" means to be more durable, reliable and maintainable) because they actually were deliberately crafted to be more durable, reliable and maintainable due to cultural differences in the past-

Were they more durable, reliable and maintainable? From your linked Wikipedia article "By the late 1950s, planned obsolescence had become a commonly used term for products designed to break easily or to quickly go out of style. In fact, the concept was so widely recognized that in 1959 Volkswagen mocked it in an advertising campaign. While acknowledging the widespread use of planned obsolescence among automobile manufacturers, Volkswagen pitched itself as an alternative. "We do not believe in planned obsolescence", the ads suggested. "We don't change a car for the sake of change."

If the automobile industry was doing it in the 50's, other industries undoubtedly were as well.
posted by nooneyouknow at 1:43 PM on June 6, 2013


Well, a lot of more modern products--everything from cars to digital alarm clocks--are intentionally designed to be either disposable in the event of failure (when they could be designed with user replaceable/servicable components instead) or to be more difficult for users to service themselves (as part of business models that involve collecting revenue from service contracts).

So I'd say those products are not designed to be "better" from the point of view of consumers, but only "better" from the point of view of the vendors. And a lot of people are very keenly aware of these changes in how products are designed and crafted. So it's not so much a magical belief in a Golden Age of Craft that makes people wary as it is noticing the differences in quality and occasionally catching glimpses of discussions among industry leaders in the trade journals and assimilating that information into the product evaluation process.
posted by saulgoodman at 1:53 PM on June 6, 2013


I don't know if this is really true or just observation bias, but I think a lot of products used to be vastly over-engineered when first built, and as successive generations of product take their place, efficiencies can be found and the bulk of items reduced, or molded plastic substituted for machined metal.

I guarantee it's been a long time, if ever, since car manufacturers cared about the longevity of their products; any car made today is almost light-years more durable than one made years ago. It used to be 100,000 miles was the Mark of Death on any car; now it's just a used-car mileage which isn't that big a deal.

One thing that creates a bias towards older things is that they're often repairable by laymen. The materials involved are often things a talented craftsman can recreate in their own workshop. You cannot do that (yet) with a complex injection-molded item.

Geoff Healey, who with his father designed and manufactured cars, including the Austin Healey, once expressed surprise at a car show (many years after the last Healey rolled off the lines) that any survived at all. They engineered them to last nine years, he said.
posted by maxwelton at 1:57 PM on June 6, 2013


Cars may live longer (despite being less user servicable), but consumer electronics are definitely not "better crafted" if you include reliability, maintainability and longevity in your criteria for craftsmanship.

Can't find a cite right now, but there was a round of news only recently about how industrial leaders in the manufacturing sector are only just now beginning to ask whether the contemporary approach to design would be better served by designing component systems to be reusable--not, of course, so consumers can service their own products more easily, but so components of discarded electronics devices could be more cost-effectively recycled back into the market for additional revenue capturing opportunities.
posted by saulgoodman at 2:12 PM on June 6, 2013 [1 favorite]


Your point is a good one, saulgoodman. But I think the author's original contention in that paragraph is still valid, especially if it's generalized in a way that relies less on specific examples.

One could say that Sturgeon's Law was just as true in the past as it is today: ninety percent of everything was and is crap. But people commonly rationalize their feeling of nostalgia by resorting to survivorship bias. They already have an affinity for older things, and since the objects, works of art and so on that survive to the present day tend to be the more valuable/ better made ones, they use that as proof that standards were better "back in the day."
posted by Kevin Street at 2:15 PM on June 6, 2013


So it's not so much a magical belief in a Golden Age of Craft that makes people wary as it is noticing the differences in quality and occasionally catching glimpses of discussions among industry leaders in the trade journals and assimilating that information into the product evaluation process.

In the instance of cars, anyway, the article rather contends correctly that "cars from the last two decades are far more reliable and safer than the cars of the 1950s and ‘60s, but plenty of people believe otherwise because of a few high-profile survivors." I don't think this clashes with planned obsolescence, where present, but there's a strongly colored bias to have our nostalgia confirmed.
Your sense of a past era tends to be informed by paintings and literature and drama that are not crap, even though at any given moment pop culture is filled with more crap than masterpieces. Why? It isn’t because people were better artists back in the day. It is because the good stuff survives, and the bad stuff is forgotten. So over time, you end up with skewed ideas of past eras. You think the artists of antiquity were amazing in the same way you associate the music of past decades with the songs that survived long enough to get into your ears. The movies about Vietnam never seem include in their soundtracks the songs that sucked.
posted by Marisa Stole the Precious Thing at 5:53 PM on June 6, 2013


I wonder how often survivorship bias occurs in athletics. Is it possible that people study a high-achiever's technique and emulate it, then that technique becomes the norm?
posted by ogooglebar at 8:26 PM on June 6, 2013


Then we visited the U.K. and eagerly switched on the TV in our hotel room. Result? Dad's Army. Now Dad's Army isn't horrible, but it isn't Monty Python. It took us a while to figure it out: People only bother to export the good stuff.

You're comparing apples and oranges here. Dad's Army wasn't trying to be anything like Python; it belonged to a different comedic tradition. And of its kind - a particular branch of British sitcom - it's considered one of the best. It also travelled perfectly well (it was on all the time when I was growing up in Australia). There are many, many worse examples out there, and they're the ones that never went anywhere beyond a single run on UK television. (We've just seen a prime example on BBC One.)
posted by rory at 2:19 AM on June 7, 2013


Hattiesburg, Mississippi represent! Love my smart hometown friends!
posted by thebrokedown at 7:48 PM on June 7, 2013


One of the interesting things about currently available vintage hand tools is they mostly fall into two camps. One group is user planes which you'll find lots of but they pretty well all show lots of wear. It's not hard to find say a Stanley 78 because most finish carpenters had one but it is fairly difficult to find one whose iron isn't worn down and one that still has it's fence. Because these planes were used and the little fiddly bits tended to go missing. The other group is weird and rare planes that no one bought because they were more marketing than useful. These planes, if you can find one, are almost all in beautiful shape and often have all their fiddly bits because they lived on a shelf or in a box for their whole life.

Looking at those two groups you'd think the weird and useless planes were the best designed because they all seem to be in excellent shape but that simply isn't what is actually happening.

On the serviceability front and specifically car serviceability manufacturers have moved strongly towards non-serviceable parts because a) people tend to let maintenance slide and b) they get pilloried when their product breaks down. The math comes down to: do we put in a bearing that needs greasing every 3000 miles and will last forever but will fail after 10,000 miles if is goes dry or do we install a sealed bearing that doesn't need to be, and in fact can't be, lubed that will last for 100,000 miles but then need complete replacement. Not too shockingly they go with the latter option because it won't get them writ up in Consumer Reports because the car owner didn't lube the bearing. Even though the bearing requiring maintenance would last forever.

the fustration comes when they combine a bunch of different and cheap bits into one monlithic unit. Take for example the power window switch on my Tiburon. The stupid thing is starting to not roll the window up. A two pole center off momentary switch is less than $10 but because the switch is paired with the switch for the other side of the car, the power mirror controller, the window lock button and the trim bezel the whole assembly is $300. Crazy but it allowed them to save assembly costs when making the car shaving a buck or two of the retail price and the people buying new cars don't care about the serviceability of a unit after they plan to sell it.
posted by Mitheral at 12:44 AM on June 8, 2013


« Older In 1993, while browsing in a junk shop, artist Oli...  |  Sometimes you have to deal wit... Newer »


This thread has been archived and is closed to new comments