Sunday, October 25, 2009

Taleb Confronts His Critics


The ever amusing Nassim Taleb has penned yet another response to his critics. He simply oozes defensiveness, which combined with his arrogance, strong opinions, and popularity, makes him incredibly fun to write about (if you haven't been accused of WEB VANDALISM by NNT, you are missing out).

So, here's his new summary, which he notes parenthetically, "I have had to repeat continuously" (perhaps his emphatic reassertion is not compelling?). Anyway, he notes that "theories fail most in the tails; some domains are more vulnerable to tail events." I agree. Newton's theories don't work at the Plank length, cosmology has trouble explaining the first minute of the universe, and evolutionary biologists have trouble explaining the Cambrian explosion 500MM years ago. Explaining most of the data is easier, and generally more important.

Now, one may protest, it's not a new point. However, Taleb argues that "nobody has examined this problem in the history of thought", highlighting the common problem of autodidact philosophers, that they tend to be too dismissive of the scientific literature. He notes what he is not talking about includes most everything that is directly related to extreme events and uncertainty: falsification, power laws, Hume's problem of induction, Knightian Uncertainty, Austrian Uncertainty, integrating fat tails into models, etc. He's aware of these arguments, but claims his idea is different

His big twist: that rare events can't be estimated, because they are rare, especially, when they are rare and have large impacts. Well, I would argue this issue is addressed in the literature he notes is unrelated to this point, as long disquisitions on the difficulty in estimating an event like WWI, or standard errors on order statistics, seems like the same subject to me. He can say these earlier discussions are flawed, but they address his key point. The more he explains himself, the more he sounds like some passage from Knight, Keynes, Hume, Minsky, etc. once you translate his neologisms (historia, ludic fallacy). I would argue Taleb is much less clear than these writers, because he's trying so hard to make the old sound new. Saying it's really new doesn't make it so.

He ends with a strong plea to not be wedded to a theory, to look at the facts and avoid fitting them into a preconceived theory. I'm sure the vast 'preconceived, untestable theory' crowd has a lot of soul searching to do. Yet, I think he's the best example of their kind. He has a theory: that he's saying something really 1)new, 2)true, and 3)important. He says many things, often contradictory, but he never manages more than 2 of those attributes in any assertion.

14 comments:

Anonymous said...

First of all a big thanks for your time & ideas on your blog, I really enjoy it. But if I can borrow 2 lines from your previous work: "Good ideas are worth repeating, and they not only make you appear smart, they actually make you smarter!
I don't think it is a complete philosophy of life, but like a flea market I can take what I want, and there's a lot there."
So quite a few smart people are arrogant and self serving, so what. If they can raise an iota of curiosity or stir debate within the stuffy academia-more power to them and more benefit to the common man (who is a bit curious).
Don't nit pick and harp on Taleb, (sounds like sour grapes). Your ideas & original work is a much better read then your criticism. Thanks again, SK

Anonymous said...

You say that there is literature, but NNT is also right that many have operated as if it were not there.

Or maybe you are defending a mainstream with a literature they did not follow.

I've seen, for instance, people use Black and Scholes option price models on oil. Oil? Who having read your literature would think that past volatility predicted future in that absurdly political domain?

Eric Falkenstein said...

I think Taleb's statements are annoying because it's like someone arguing that inflation is primarily caused by increases in the money supply, and then demanding that people recognize the novelty of the statement. It's true, but it's definitely not new, and the insistence that it is new highlights a real bizarre, willful, ignorance.

Anonymous said...

NNT is an ass. You've got it right: if his work was truly unique, then he wouldn't have to spend so much energy trying to convince everyone that it is so. He seems to care more about whether these people recognize his brilliance than the ideas he's trying to put forth.

michael webster said...

Eric;

In the piece you point to, Taleb says that rare events are even less likely than we would think from reviewing past data - they are just more harmful.

This sounds to me in direct contrast to people who believe that the problem with our models is that rare events are not that rare.

If true, it would be important and novel. Yes?

Eric Falkenstein said...

He doesn't say rare events are rarer than expected, that would be a testable statement...he qualifies that by saying 'sometimes'...so it's untestable, pointless.

Anonymous said...

data are (data is =inncorrect) data are plural

Eric Falkenstein said...

" Explaining most of the data is easier"
Here, 'is' pertains to 'explaining', not 'data'.

michael webster said...

@ Eric, ok suppose he says "sometimes", that doesn't make it untestable. We just want to know which rare events he thinks are even less probable than the standard theory would have predicted or retrodicted. If there aren't any, then that would make the theory vacuous. My sense is that you don't think Taleb or any of his supporters are up to this task. Yes?

Eric Falkenstein said...

he does not apply it ex ante, only ex post. If you can find a consistent assertion that certain events(eg, government bonds crashing) are undervalued, I'd like to see it. It would be a testable hypothesis, scientific. Instead, he blathers on with neologisms and claims of originality. He embodies everything he writes against: pompous, unempirical, obtuse.

Walt French said...

Avast, ye swab! Newtonian gravity works only too well at a plank's length — but the Wikipedia entry for Planck Length was interesting in its own right.

Thanks for the concept, even if your word processor unhelpfully miscorrected what you wrote to what you got.

PS @Anon from one pedant to another: on my side of the pond, data is often accepted as a (singular) “mass” noun, much as we all accept agenda despite its roots as a Latin plural.

michael webster said...

@Eric,

The question about ex ante/post is interesting and raises other issues.

I find Taleb, like Soros, interesting, despite both of them being maddeningly unclear and having an undergraduate's cockiness.

(Soros may get better because of his association with Colin McGinn, a very good philosopher.)

We are unlikely to agree about the value of Taleb's work, probably because I don't work in financial world and so his jabs don't bite so much.

Karl K said...

As an options trader, I have read all of Taleb's books.They are all worth reading.

But basically, his whole approach boils down to one idea: descriptive statistics work until they don't.

I don't know whether that is an original idea or not, but it is a very USEFUL idea, and if you believe that market returns will ALWAYS be lognnormally distributed, or that risk can can always be hedged, well, good luck with that.

Eric Falkenstein said...

I don't know anyone who believed that returns are always lognormal. I suppose that's true for someone totally knew to finance.