Taleb Mishandles Fragility

Christmas traditions have gone from stockings and exchanging gifts, to fruitcakes, bad sweaters, NBA games, and now Taleb books, a sign that perhaps the Mayan return isn't so much an apocalypse but rather a mercy killing. Taleb is one of many best-selling authors I don't enjoy (Tom Friedman, Robert Kiyosaki, Snooki), but as he is prolix, pretentious, petulant and clueless, I enjoy commenting on his latest blather (my review of Black Swan here, Bed of Procrustes here).

His latest book Antifragile is driven by his discovery that there is not an English word for the opposite of fragile, which he thinks could not be 'robust' (this neologism is one of the few new ideas presented in this book, not that I think we need more new Taleb ideas). Fragile things lose a lot of value when mishandled, 'anti-fragile' things increase a lot in value when mishandled. He thinks this is very profound and therefore needs a book. The problem is that mishandle implies an adverse effect by definition, which is why there isn't a word for something that goes up in value when you mishandle it.

The concept of things increasing in value with small probabilities is well-known. Words used for this concept include: good luck (when preparation meets opportunity), lottery tickets, a home run, teenie (a low-delta option), eureka moment (scientists), ten-bagger (a stock that can increase in value ten-fold). These are compound nouns, and if English were German, these would all be one word. They are the basis for patent trolls, venture capital, oil drilling, poring over a sheet of financials, and dating (a single prince makes the many other tedious dates worthwhile). Having good luck, winning lottery tickets, is nice, but how to achieve this is not straightforward, and certainly not simply by being long lots of them.

One interviewer's take-away from his anti-fragile thesis was the following:

So what to buy? Taleb chooses investments with small downsides and large upsides: penny stocks, distressed assets, and options. “You want investments that clip the left tail.”

An option has a truncated left-tail: it pays off zero or the stock price different than some strike price, but is not necessarily a bargain because the price is non-zero. In general this is why penny stocks and long option positions have lower-than-average returns (and distressed assets are not better-than-average), as lazy investors chase instant riches. Further, contra Taleb, it is not the quantifiability of lottery tickets, or the fact that they have a maximum payoff, that makes them bad investments: things with lottery-ticket type qualities with uncertain parameters such as internet business opportunities are generally a fraud with a poor expected return, and things like IPOs, or analyst disagreement, which are also riskier than average yet nonquantifiable have lower-than-average returns.

He doesn't identify key attributes of attractive, risky (oops, antifragile!) opportunities, just implies they are the ones that unlike options and lottery tickets, work well. In fact, he's anti-theory, so one supposedly finds them by random sampling (aka 'trial and error'). That's a strategy statistically proven to underperform, catering to the biases most investors have, why both day trading bucket shops thrive and low volatility investing works. As a self-help book it's like someone saying you should eat more sugar, a strategy many will find highly convenient.

The book is really a big spread argument that it's good to be long gamma, bad to be short it. Gamma is the essence of an option, why there's 'time decay' or theta, a predictable expense that anticipates the payoff times the probability. Whether or not this theta is adequate for the gamma is whether an option is priced fairly or not, and generally people pay too much for gamma, why historically the VIX has been about 1% higher than the SP500's actual volatility, and this implied volatility bias has been even higher in the tails. Being long options (positive gamma), especially out-of-the-money options, has been a losing strategy.

One key to understanding Taleb is the Freudian concept of projection: he applies his greatest faults to others. For example, he defines the "Joseph Stiglitz problem" as cherry-picking his prior statements to claim they predicted something when they did not, referring to Stiglitz's ill-fated Fannie-Mae prediction and subsequent recollection of calling the 2008 financial crisis in a later book. Yet Taleb himself did the same thing, as he criticized Fannie Mae for not understanding the embedded interest-rate option in their mortgage portfolio, but then claims he accurately predicted Fannie's failure. Prepayment risk is very different than collateral risk, and Taleb mentioned nothing about collateral risk prior to 2007, and instead alluded to the prepayment option problem. It's like a guy who says corn prices might increase because of risk from floods, and when a collapse in the dollar causes it's price to rise, states, 'I told you so.' Hindsight bias, name dropping, pretentious mathematics, and charlatanism, are all Taleb signatures he sees everywhere in others.

Another key to understanding Taleb is that he has a French post-modern tendency to write to impress rather than explain. He provides hundreds of loosely related anecdotes, reminding me of the Talmud quote that 'when a debater’s point is not impressive, he brings forth many arguments.' I actually agree with a lot of Taleb, such as the intractability of risk because it is endogenous, and I think he's vaguely libertarian, but he says so many inconsistent things that doesn't mean much.

Then there are the many confused or dubious assertions, such as that the improbable events that underlie his strategy of embracing Black Swans are both impossible to quantify and highly rewarding, or that fragility is like risk in that it is what causes things to fail and has a return premium but unlike risk is quantifiable, that finance professors don't understand 'real options', or that economists don't understand that E[f(mean(x)]E[f(x)], or that the biggest investing problem created by Markowitz is too much optimization. His equation for fragility has a couple of subjective parameters (K, and the density of alpha) that are probably unspecifiable given his definition of Black Swans (its probability can't be estimated!), and equations with unknown parameters are like bad haikus, or rather, very helpful if you want to impress the mathematically challenged. Then there are the ramblings about lots of Greeks and 'street smart' traders. In his confused mind this all fits together like Euclid's Elements, but it's more like a non-humorous version of David Sedaris's Me Talk Pretty One Day.

Taleb often suggests it is good to be long volatility, things that gain from greater uncertainty (see his YouTube on this here). As the VXX has shown, while this has nice covariance properties with the stock market (going up in 2008), it has a horrible long run return even including disasters like 2008 or the 1987 stock market crash, it's cost is much greater than the thing it is supposed to hedge. I bet many of the unfortunate investors who have ridden the VXX to zero over its existence have a copy of The Black Swan on their bookshelf (and you can extrapolate it backwards, and even if it started in 2006 it would be a loser). The 'long vega' bias simply isn't a good one. Another example: mathematician, publishing mogul and Taleb-fan Paul Wilmott's big advice during the recent financial crisis to buy volatility--it gains from uncertainty!--which was like recommending earthquake insurance right after the big one hits. Good trade, wrong sign.

The fund Universa, of which he is affiliated, states that it is no longer merely long volatility or gamma, but timing when to be long volatility or gamma. I'm sure all those investors who jumped in Universa circa 2009 would be surprised to know that's the strategy, but as part of management, he benefits from the gamma resulting from the asset management fees from investors fooled by randomness. He does have an excuse here, as he did write a book on that, so it's not like they weren't warned.

I checked on his book Antifragile back in late October on Amazon, and saw the reviews from those who got the pre-release version. A few reviews where negative, and in the comment section (you can comment on reviews, and comment on comments) Taleb himself was in there angrily responding at length to these negative reviewers, and his cult-like fans piled on. From a guy who writes in Antifragile that criticism should be welcomed, his response to criticism is consistently hysterical. A week later, one of the negative reviews was deleted, the poor sap didn't anticipate the venom from simple Amazon review. I have received many spirited emails over the years from his acolytes, and back around 2005 NNT himself sent my boss emails on two occasions telling him I was saying hurtful things about him on the interweb and that I must stop. He's got the skin of a mudskipper.

For example, a commentator on a negative Amazon review writes:
Please respond to Nassim Taleb's rebuttal and more clearly define your expertise and argument with the message of his book. I bet if you engage Mr. Taleb (once again, a rare honor) you will find that the both of you fall along the same lines of understanding. If you do not respond, it simply means that the review was an after-thought to retain review ratings on Amazon and not an honest intellectual review of the book.

That's the fawning tenor typical of his fans, and that kind of intellectual insulation doesn't encourage reality, let alone clarity, which Taleb notes is a major problem among other people.

Taleb doesn't do himself any favors by responding to one review by noting that This review is grounded in a fundamental error. It falls for the conflation described in the book between medicating and overmedicating, intervening and overintervening. The book NEVER says that mental illnesses should not be diagnosed in children, it says that it should not be OVERdiagnosed and OVERMEDICATED.

First, note the deranged use of CAPS, highlighting that he at least follows his own advice to not take Prozac. Then, note that his big idea on mental illnesses is that people should not over diagnose or overtreat them. True enough, given the meaning of the prefix "over", but if that's his point it's tautological. Given Taleb's fixation with word cognates, it's odd that he repeatedly makes these kind of errors. This kind of vapidity is why I think he's a blowhard.

One theme of the book is hormesis, the finding that things that are clearly bad for you at extreme doses, are good for you in small doses; a glass of wine a day, radiation, germs, etc. If you have zero exposure to germs, you won't develop a healthy immune system. Arthur Robinson has been a lead of hormesis with his work in the 1970's, and there's a fascinating tale about how he discovered this in the context of the assertions about radiation extrapolation by Robinson's mentor, the famous chemist Linus Pauling, and a nasty legal battle that ensued. The fact that micro-instability is necessary for greater macro-stability is a very good, and very Austrian point. If he was a serious scholar he would fit his ideas into these threads and highlight his novelty, but as he has no novelty, he avoids this route. Though Taleb is trying to outflank academics he derides, his writings highlight one of the main benefits of academia where scholars usually fit their ideas into the literature so you can better assess their innovation and the state of the art. Autodidacts are often rambling, repetitive, and most importantly, wrong.

He notes there's a sweet spot for most medicines, and that some exercise is good for you, not in spite of its stresses, but because of them. I guess a lot of people find this really shocking advice (Moderation in all things! Who knew!?), so perhaps this is why his audience is so large: he's focusing on people without any common sense. But if the key to benefiting from the 'right' amount of medication is dosage, how does one find this dosage? Trial and error? That's how most animals learn, but it's pretty inefficient in general, I certainly don't want my kids figuring out most of their life lessons that way, because its very time consuming and costly. Surely, a 'moderate amount' of trial and error is essential in everything, but this kind of qualification highlights he's missing the key point.

He still thinks Portfolio Theory, and most Economic Nobel Prize winning research, is predicated on distributions with fixed parameters. It isn't. Financial academic standard bearer Eugene Fama spent half his dissertation in the 1960's on Mandelbrot's observation about fat tails, and like everyone else in the profession, left this thread because it isn't that interesting: the static parameter assumption gives qualitatively similar implications to a more realistic distribution where means have standard deviations ad infinitum, yet gains a great deal in transparency.

Transparency and simplicity, in fact, are key features of models, always a tradeoff with realism, but that's a nuance too subtle for Taleb. The effect of adding fat tails through stochastic parameters is isomorphic to assuming more risk aversion or a higher volatility, so it's trivial to fit inside the box, and the CAPM and other theories are basically the same, just messier. The same is true for Black-Scholes-Merton and the Miller-Modigliani theorem.

As per correlations being stochastic and so 'uninformative', he is wrong again: they are highly predictable, as high beta portfolios formed using past data create portfolios with higher future betas. The same is true for low volatility investing. The problem with betas (ie, correlations), is not that they change so much as to be irrelevant, but that they aren't correlated with returns over long periods as theory suggests (the subject of my book, The Missing Risk Premium, that there are no omnipresent correlations between covariances and average returns). So, I agree modern academic finance is highly flawed, but not for reasons Taleb suggests.

A good amount of gamma, like having just the right amount of medication or specialization, is a good thing. Yet the right amount can be positive, negative, or zero, in various contexts. Many good things have negative gamma, such as the strategy of being nice to strangers: it has great downside, such as when you naively interact with an aggressive stranger, yet being nice is a good default strategy. Then there are things with no gamma, such as brushing your teeth everyday or simply being polite, which generally doesn't have a lot of effect either way in your life any time you do it, but over time is quite salubrious. Noting gamma per se, especially large gamma, doesn't tell you if something is good or bad, rather, just that it could be really good or really bad.

You can price gamma and it's not free, so the question is always whether this price is too high or too low. Indeed, Universa's new emphasis on timing volatility trading begs the question: how do you time these things? How do you identify things that respond hydra-like to having its head cut off? Contra Antifragile I would say: don't bias your portfolio towards lottery ticket investments, even if only 10%. Find something you are good at, become excellent at it, and invest your time and speculative wealth there.