Finally, after almost two years, I’ve gotten around to reading the much talked about book The Black Swan, by Nassim Nicholas Taleb. The term Black Swan has entered financial vocabulary, so it’s good to evaluate the source. Overall, it’s a good term; a Black Swan event is a rare event, hard to predict, that has a large impact. For instance, until the 17th century Europeans thought all swans were white. They are, in Europe. In Australia, they’re not.
Overall, it’s a pretty good book. I’ll award it a 3 out of 5. However, it can be tough to read at times. Dr. Taleb is pompous. Very pompous. He seems to be basically alone in this world. Describing someone, he wants to “put a rat down his shirt.” While I don’t appreciate false humility, either, or too much self depreciation (as in the saddest hoopster), this book would have more power if it was more targeted. A statistician’s review says it well, so I’ll quote Robert Lund in the American Statistician: “reckless at times and subject to grandiose overstatements; the professional statistician will find the book ubiquitously naive.” As a professional statistician, I agree wholeheartedly.
Before I come back to why I agree, let me list what Dr. Taleb gets correct. He is right that many things in this world are not Gaussian or normal, and not subject to the bell shaped 95% cutpoints like the linked applet. There are many things in his Extremistan, particularly those that follow Mandlebrot power laws, and putting them into his Mediocristan is dangerous. Note that by choosing the word Mediocre for the other group, against the currently positive word Extreme, he induces stylistic bias. Who would ever want to be in Mediocristan? Well, except nature with height and growth processes and stuff. So he’s got a correct point, but overreaches.
Another clear point is about the Ludic Fallacy, which notes that real life does not have the same structured randomness of games. Though he had to invent a term for it, the author is correct. Too many introductory examples involve games of chance, cards and dice and roulette wheels. Equal probabilities and independent events are much rarer in reality than Moore and McCabe’s introductory text. In my classes, I make a point about independence, with several reminders of its importance, including the most important one, the test question.
In other words, I do my job as a professional statistician. That’s the big deal here. I know Dr. Taleb has had dealings with the professional statisticians since the publication of this book, since I quoted Dr. Lund from the American Statistician special August 2007 issue on the book. I suspect that he had very few dealings with my kind before. Unfortunately, what passes for “statistics” nowadays mostly comes from economists. Economists, well, are pretty nasty. A friend of mine once said that “Economists are just sociologists with Asperger’s Syndrome.” I replied that I’ve always thought of them as megalomaniac statisticians, or just plain bad ones. Even the so called revolution, this Freakonomics thing, is really just regression analysis, at the second course level. In some places, it’s not even a good second course level, as this blog post acknowledges a mistake and then tries to defend an abortion model with serious interaction problems. Do they not understand saturation or degrees of freedom? I recommend this critique; even though I might not like all the articles from Steve Sailer, he’s demolished Dr. Levitt on this one. It’s too bad he’s not cool enough to get a fancy book. And who knows – maybe I’ll venture into that perilous topic myself, soon.
On page 239 of my hardback edition, Dr. Taleb thinks that in comparison to what he calls dull statistics classes, “Clearly it would have been more beneficial, and certainly more entertaining, to have taken classes in the neurobiology of aesthetics or postcolonial African dance, and this is easy to see empirically.” I should be careful here. Particularly before computers became readily available, statistics courses focused heavily on computation. I’ve seen those books from before roughly 1990; they weren’t as good as the courses of today. Given his age, he likely had one of those older courses. Postcolonial African dance might well be better. I also wish that standard deviation had a name that didn’t imply standards, like how I prefer credible interval to confidence interval. That doesn’t affect the main point, that the true professionals know the problems, and that at least now we’re trying to get those across.
Precision, research, and greater balance could have taken this book from decent to great, so I look in a little frustration at Dr. Taleb’s work. I do have his other book, Fooled by Randomness, which might not show the problems of this one. It won’t be random, or a Black Swan, if I get around to reading that.