- Taleb, Nassim Nicholas.
The Black Swan.
New York: Random House, 2007.
ISBN 978-1-4000-6351-2.
-
If you are interested in financial markets, investing,
the philosophy of science, modelling of socioeconomic
systems, theories of history and historicism, or the
rôle of randomness and contingency in the unfolding
of events, this is a must-read book. The author largely
avoids mathematics (except in the end notes) and makes
his case in quirky and often acerbic prose (there's something
about the French that really gets his goat) which works
effectively.
The essential message of the book, explained by example in
a wide variety of contexts is (and I'll be rather more
mathematical here in the interest of concision) is that while
many (but certainly not all) natural phenomena can be well
modelled by a Gaussian (“bell curve”) distribution,
phenomena in human society (for example, the distribution of
wealth, population of cities, book sales by authors, casualties in
wars, performance of stocks, profitability of companies,
frequency of words in language, etc.) are best described
by scale-invariant power law distributions. While Gaussian
processes converge rapidly upon a mean and standard deviation
and rare outliers have little impact upon these measures, in
a power law distribution the outliers dominate.
Consider this example. Suppose you wish to determine the mean height
of adult males in the United States. If you go out and pick 1000
men at random and measure their height, then compute the average,
absent sampling bias (for example, picking them from among college
basketball players), you'll obtain a figure which is very close to
that you'd get if you included the entire male population of the
country. If you replaced one of your sample of 1000 with the
tallest man in the country, or with the shortest, his inclusion
would have a negligible effect upon the average, as the difference
from the mean of the other 999 would be divided by 1000 when computing
the average. Now repeat the experiment, but try instead to compute mean
net worth. Once again, pick 1000 men at random, compute the net
worth of each, and average the numbers. Then, replace one of the
1000 by Bill Gates. Suddenly Bill Gates's net worth dwarfs that
of the other 999 (unless one of them randomly happened to be
Warren Buffett, say)—the one single outlier dominates the
result of the entire sample.
Power laws are everywhere in the human experience (heck, I even
found one in
AOL search queries),
and yet so-called “social scientists” (Thomas Sowell
once observed that almost any word is devalued by
preceding it with “social”) blithely assume that
the Gaussian distribution can be used to model the variability
of the things they measure, and that extrapolations from
past experience are predictive of the future. The entry
of many people trained in physics and mathematics into the field
of financial analysis has swelled the ranks of those who naïvely
assume human action behaves like inanimate physical systems.
The problem with a power law is that as long as you haven't yet seen the
very rare yet stupendously significant outlier, it looks pretty much like
a Gaussian, and so your model based upon that (false) assumption
works pretty well—until it doesn't. The author calls these
unimagined and unmodelled rare events “Black Swans”—you
can see a hundred, a thousand, a million white swans and consider
each as confirmation of your model that “all swans are white”,
but it only takes a single black swan to falsify your model, regardless
of how much data you've amassed and how long it has correctly predicted
things before it utterly failed.
Moving from ornithology to finance, one of the most common causes
of financial calamities in the last few decades has been the appearance
of Black Swans, wrecking finely crafted systems built on the
assumption of Gaussian behaviour and extrapolation from the past.
Much of the current calamity in hedge funds and financial derivatives
comes directly from strategies for “making pennies by
risking dollars” which never took into account the possibility
of the outlier which would wipe out the capital at risk (not to mention
that of the lenders to these highly leveraged players who thought
they'd quantified and thus tamed the dire risks they were taking).
The Black Swan need not be a destructive bird: for those who
truly understand it, it can point the way to investment success.
The
original business concept
of Autodesk was a bet on a Black
Swan: I didn't have any confidence in our ability to predict
which product would be a success in the early PC market, but I
was pretty sure that if we fielded five products or so, one
of them would be a hit on which we could concentrate after the
market told us which was the winner. A venture capital fund
does the same thing: because the upside of a success can be vastly
larger than what you lose on a dud, you can win, and win big, while
writing off 90% of all of the ventures you back. Investors can
fashion a similar strategy using options and option-equivalent
investments (for example, resource stocks with a high cost of
production), diversifying a small part of their portfolio across
a number of extremely high risk investments with unbounded upside
while keeping the bulk in instruments (for example sovereign debt) as
immune as possible to Black Swans.
There is much more to this book than the matters upon which I have
chosen to expound here. What you need to do is lay your hands on this
book, read it cover to cover, think it over for a while, then read it
again—it is so well written and entertaining that this will be a
joy, not a chore. I find it beyond charming that this book was
published by Random House.
January 2009 