The Black Swan: a book review of sorts
Tuesday, July 31, 2007
Over the last 2 days I read The Black Swan: The Impact of the Highly Improbable
, by Nassim Nicholas Taleb
(partly research, party self interest).
Taleb's premise for the book is the effect outliers have in everything around us: these outliers are what he calls Black Swans. For example, a system that appears to be fully predictable all of the sudden displays an unexpected event dramatically changing the whole system.
In the book there are two worlds: Mediocristan and Extremistan. According to Taleb, we live in Extremistan and the Black Swans are what make our world go round. There are many examples of extreme events, for example, Bill Gates' net worth: he is worth around $60 billion, the rest of us are not, and are more or less aligned around a more realistic average. This is one of the books main points, as building averages with this type of outliers is what makes Taleb's argument valid. In a world with Bill Gates and John Does around, the average measure of wealth in our society is meaningless. Imagine the average of a population of 2, with Bill on one end and John in the other: ($60,000,000,000 + $10,000)/2 looks good on paper--the average would suggest that the whole population is doing very well.
Hence, our world is Extremistan, where not everything can be averaged or meaningfully represented by well defined statistical distributions such as the Normal distribution.
To explain how we are blindsided by Black Swans, Taleb presents the example of a turkey that is fatten up for 1000 days, and then on day 1001 (it's thanksgiving) there is no more food, as the turkey is cruelly decapitated and scrumptiously cooked.
To us, the feeders, this is not a Black Swan, as we know what's coming to the poor turkey. To the big ugly bird, though, this is a Black Swan: it was fed for 1000 straight days and there was no indication the 1001th day would be any different.
It is this lack of predictability of events that drives Black Swans. Taleb argues that our inductive methods of proof fail to predict Black Swans because we are only looking at well behaved data: our methods can't look or predict outliers. More specifically, if we were the turkey of the story (and we are in so many ways, e.g., we can't predict how many people will die in a war), we rely on historical data to live our simple eating, fat lives.
I don't think anyone will disagree with Taleb. He admits, though, he is not the first one to point this out. His beef, however, is with our reliance of mathematical theories and models to try to put order in a random world. For example, he really doesn't like CAPM, or the French, or well paid CEOs, or Economists, or guys who wear suits and $250 ties, or tenured faculty, or Nobel price winners. In fact, his insistence on being sarcastic and jabbing the French every fifth page is distracting. And he doesn't tell us why he doesn't like the French. On a personal note, I don't know "The French," but I don't dislike them.
Overall, I found his writing interesting, otherwise I wouldn't have read the book in two days. Nevertheless, I could have read Chapter 15, 16, and 17, and still have gotten the point; though, he does tell the "technical" reader to skip these very chapters. To him, those who have been trained in the sleazy arts of mathematics or finance are already tainted with bias.
He calls everything we do a Gaussian fraud, because of our over-reliance on the Normal distribution. But, to be fair, not everything has a normal distribution and he knows it but he doesn't clarify it throughout. I'm also puzzled why he doesn't mention the Central Limit Theorem. Perhaps there is no need to make his point, but I didn't like the omission.
Furthermore, Taleb calls himself an empiricist: he says he tries things and then tries to apply models. The rest of us do the opposite: we are top-down users, as we rely on mathematical models to play with nature. He has been a derivatives trader, and perhaps made a fortune while doing it (he's retired), so maybe he knows what he is talking about. Personally, though, I haven't gone through a finance course where all these mathematical methods he shuns don't come with a big if
attached to them. There is always the disclaimer that "sure, these methods work on paper, but in the real world they don't so well." No professor worth his tenure would claim that the mathematics work in every scenario.
So why do we study all these useless concepts? That's also a question he asks, and perhaps his source of disgruntle (if you read the book, you will see what I mean): I think he wants his theories to become mainstream. The problem is that, as he was told, we can't just change everything in one day, and, most important, "...you cannot throw the baby out with the bath water."
He does present an alternative, based on the work he is doing with Benoit Malderbrot (right, the fractals
guy). He vaguely explains how these outliers can be incorporated into a new model. The problem I see, however, is that Black Swans are not really predictable: they are just outliers, which are highly improbable events. For example, I can't turn invisible but it doesn't mean that it is impossible--just highly improbable. And this is the point of the statistical models we do study: should you be basing your future on something that is not likely to happen in the history of the whole universe (that's way more than 6000 years, as some believe). Probably not. Nevertheless, being able to predict the next famine or the location of the birth of the next suicidal dictator is probably a good thing, though unlikely to be predictable.
He does write early on in the book that if a Black Swan is not observed, it probably doesn't mean that it wasn't going to happen; but it could mean that the outlier was prevented by some random act of heroism. In the world we live in, our heroes are not the steroid pumping athletes we see on TV, but the ones who kill the butterflies before flapping their wings on the other side of the world, i.e., the ones who prevent the storms without knowing.
I do believe there should be a more technical book, a book that will not make the New York Times bestsellers list, but a book for the rest of us: the ones who don't skip chapters 15, 16, and 17. (I've looked for the mathematics, but I've only found introductory articles
As far a recommending the book goes, I'm a bit conflicted. It is an easy read and is the popular book of the times (similar to Wikinomics or The Tipping Point), but it left me wanting more (perhaps it's the main goal).
Nonetheless, I think you should contribute to his retirement fund: buy the book, read it, and impress your peers with your knowledge of fat turkeys and their limited intellect that doesn't allow them to predict that they will be eaten in important holidays (we would know, right?). Also, by buying this book the chances of a next book increase: I'm expecting his next tome to have an explanation of why he dislikes The French
, and Nobel laureates, and fund managers, and...(Well, he kind of alludes to the fact in the book, but maybe I skipped over that part.)