The Blackened Swan

by Gene Callahan

I am now on my second review of Nicholas Nassim Taleb’s The Black Swan, the first of which will appear in The Review of Austrian Economics soon, and the second of which, an essay length review, is due out in a forthcoming issue of Critical Review. I have very mixed feelings about this book — Taleb is very good in his areas of expertise, but apparently feels that his skill in those domains makes him emminently qualified to propound on any damned thing that comes into his head. I wanted to share with ThinkMarkets readers a little excerpt from my upcoming CR review:

Unaware that he is illegitimately importing explanatory terms from domains in which they make perfect sense into a territory in which their use is banned for incoherence, Taleb writes, “But it is hard to look at a computer or a car and consider them the result of aimless process. Yet they are” (170). Certainly, if we restrict ourselves to considering the human past in terms of matter and energy mechanically interacting to produce unwilled motion, or even in terms of the statistical aggregates of a social science that analyzes society as a sort of mechanism that is controlled by the influence of such aggregates, then we have pre-determined that all we will discover are aimless processes. That only means that we have found what the rules of our search restricted us to finding. That does not render such findings useless or incoherent, but only demonstrates their conditional nature. Other than an unfounded prejudice against any alternative guidelines for exploring the human past, there is no reason we shouldn’t conduct other explorations using other investigative principles, such as attempting to understand how events of the past arose from purposeful human action. From that perspective it is ludicrous to propose that computers and cars were the outcome of “aimless process,” since it is obvious that their inventors set out intending to create useful machines, and were not merely throwing together random materials only to be shocked that the result could, for instance, be used to get around town more rapidly or run a spreadsheet program.

19 thoughts on “The Blackened Swan

  1. Are you being fair to Taleb, Gene? There is a sense in which the modern automobile is like the eye in biology. You have this huge network of activities to support the modern car from the inputs to the factory, to the distribution network, to the infrastructure supporting the extraction, refinement, and distribution of gasoline, to the knowledge and experience of the driving public that lets them understand modern cars well enough to drive them. It’s an evolved system, not a designed system. Each individual car was planned in the very sense you indicate, as were all — well, most I guess — of the individual innovations on the way. But there is that other sense, surely Taleb’s sense, in which the car or the computer is like the mammalian eye: a complex structure the emerges marvelously from an undirected evolutionary process. Henry Ford did not envision monster trucks.

  2. “Are you being fair to Taleb, Gene?”

    Yes, yes I am.

    One of my criticisms is that Taleb actually boasts about how little he cares about the quality of his writing, and it shows! I did not suspect that this was what he was talking about, and if it was, he could have clarified it in a brief sentence — but he couldn’t be bothered.

  3. I dunno, Gene; I’m not feeling it. It seems to me that Taleb’s basic lesson is the world is characterized by 1) radical uncertainty, 2) fat tails, 3) regime change while humans are characterized by a kind of hubris whereby they imagine they have knowledge they do not in fact have. People toss out outliers and then say, “see the data are smooth and we know just what the future will bring.” These are quite reasonable and true arguments IMHO. Taleb’s on to something, n’est–ce pas?

    I saw him on TV recently talking about the crisis. He did seem too alarmist to me. But in comparison to his basic argument, such excess pessimism is a pretty trivial infirmity.

  4. Yes, Roger, he’s good with his basic argument — which I already noted. He’s just very sloppy when he steps outside it.

  5. Economists often got criticised for ‘economics imperialism’ when they commented on political systems, the family, evolution and all else. So maybe we now are seeing ‘statistician imperialism’??

  6. I have been struggling with his application of fractal geometry to human action. I guess I don’t understand it at all. He criticizes the quants for their lack of understanding of statistical analysis (rightly so), but he then applies something which appears to be a mechanistic analysis of human behavior.

    I believe he’s on to something in his argument against those who have failed to understand risk, but I don’t understand the math.

  7. J Harding:

    For the fractals of finance, I think you kind of have to go to the source, namely, Mandelbrot. He has a semi-popular book on the topic, though I forget the title just now. I guess it is “mechanistic,” but we do observe, e.g., “fat tails” in asset-market time series. Right tool for the right job and all that. I’ve used Mandelbrot’s work in a perfectly “Austrian” way in my “Big Players” work, although my technical level is obviously waaaaay below that of Mandelbrot.

  8. I understand that Mandelbrot has discovered patterns in nature that allow us to explain nature mathematically. And I guess I can understand that the power laws may apply to market patterns; I just don’t understand why. More work to do.

  9. Taleb’s use of Mandelbrot is a off base. Mandelbrot has won few converts for the simple reason is that his models are no better than the current ones. Taleb is disingenuous on this topic. No one in Economics or Finance disagrees with the fat tail issue. The problem is that to date there has been no distribution offered by critics that truly outperforms current models.

    Fractals are simply a nice approximation of nature. Nature is not really a fractal structure. It may appear so at a quick glance. The number of levels that remain are quite finite. Optimization models have shown to be a better tool.

    Another issue that Taleb ignores is the difference between risk and uncertainty. I think that Knight offers a clearer picture of the issues that confront decision maker than the attempt to reduce everything to a probability. The assumption suggests a simple function generates values rather than the probability as an emergent function of the system.

  10. Hi, Simon.

    Of course any model simplifies and is, therefore, an approximation. Anyway, it could be that Mandelbrot’s crazy functions buy you nothing, but I tend to think the finance community is just not giving him his due. In my modest forays into asset pricing, I was driven to use his tools and concepts. He came to the empirically. He calls them “phenomenological” though obviously not in some Husserlian sense.

    I would spot you the Knightian point, which is important IMHO.

  11. koppl-

    Mandelbrot’s conjectures on the fractals is challenged by Adrian Bejan who is a MIT professor of engineering. He does an excellent job showing that while fractals are a nice approximations they fail to provide explanatory power. This is an important because of the need to clearly de-mark descriptive versus causal models.

    I remain skeptical Mandelbrot’s offers economics and finance a worthwhile path to pursue for the very reason that Adrian challenge refutes a simplistic fractal description. Mandelbrot is simply approximating a characteristic of the financial system not its underlying structural nature.

    I am not saying that Mandelbrot is not a great mathematician, I am simply saying he is not modeling the relevant phenomena in question.

  12. Simon,

    May I have a cite on the Bejan critique? I came up empty when I did a quick google. Fractals are self-similar geometric forms. But no one even said geometry is causal! Thus, I’m just not getting any intuitions about how Bejan might be criticizing Mandelbrot’s work in financial markets.

  13. koppl-

    Adrian Bejan work is not related to Finance. He is an engineer. His critique of fractals is limited to optimization in engineering realms. The particular book I am referencing is Shape and Structure, from Engineering to Nature: From Engineering to Nature. It is an outstanding and thought provoking book.

  14. I’m following this conversation with interest. It seems to me that these arguments go to the fundamentals of Taleb’s conclusions. I have a couple of questions:

    Simon Says:”The problem is that to date there has been no distribution offered by critics that truly outperforms current models.”

    True? I thought that was Taleb’s basic argument. He says the current risk models are crap. Or are you talking about something else?

    Simon Says: “Another issue that Taleb ignores is the difference between risk and uncertainty.”

    I thought that was one of Taleb’s major points. Perhaps I don’t understand the argument. but isn’t uncertainty the most important factor in risk assessment? I understand that the bigger the bet, the greater the risk. But, because of uncertainty, the amount of the bet increases the risk.

    I’m not an economist or an academic, so I’m asking, not arguing.

  15. Simon:
    Some guy has some sort of criticism of fractals in engineering, therefore Mandelbrot’s work in finance in no good. Is that really what you mean to argue? Forgive me if I’m off base, Simon, but that’s what you *seem* to be saying.

    J Harding:
    Right. Fat tails is big point with Taleb and Mandelbrot. Big events are more likely than you think if you’re modeling the world as Gaussian.

    The risk vs. uncertainty is more of a technical point. Risk, however, large or small, can be expressed in the language of probability and statistics. Uncertainty in its technical Knightian sense cannot. There’s a huge debate on whether non-quantifiable uncertainty really exists. I say yes, but some deeply serious people deny it.

  16. Koppl-

    Bejan’s observation on the explanatory power, or should I say their descriptive nature, of fractals is material regardless of where they are being used. Fractals are a very interesting modeling abstract. The fact that they exhibit particular properties that are useful at first blush does not mean they are actually reflecting the underlying phenomena in question. Bejan outlines this issue in a number of areas where fractals have been used to describe flows and branching problems. They offer little guidance in optimal design. This is s strong indicator that approach is not causal in structure. My point remains that I am highly skeptical that fractals will offer a path this anything more than descriptive. In addition, I see Mandelbrot’s approach as descriptive in nature rather than causal. That is not to say that I could be wrong.

  17. Hm! So that *is* what you’re arguing, simon, and you’re standing by it.

    I repeat: Mandelbrot describes his finance fractals as “phenomenology.” Fractals are geometric forms and thus not causal. How are your points a criticism?

    Standard finance is not causal either. It’s equilibrium theory based on some version of a no-profit assumption. You can indeed do lots with that framework and it completely trumps what it replaced, namely, loose rules of thumb. But in a naive Popperian context, it’s been falsified by the persistant, seemingly universal, fact of fat tails in finance time series.

    I co-authored a paper showing that “X-skewing” may pass a standard GARCH test, though it is not GARCH. ( That result suggest that current empirical standards in finance might be self-referencing, but inappropriate to the underlying process — the very point Mandelbrot makes.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s