Book Review: Why Stock Markets Crash: Critical Events in Complex Financial Systems
In fairness to the local tv 'journalists,' they focused on an aspect of the book that made me skeptical then and, to some extent, makes me skeptical now: the last chapter of the book (chapter 10). But if we (temporarily) ignore the last chapter, Didier Sornette's book is the best book in the still nascent discipline of econophysics and one of the greatest contributions to the scientific understanding of markets that has ever been written.
The best one-line endorsement (although I will give a detailed review below) I can give this book is the following: Replace "stock" with any word you like ... "credit" or "mortgage" or "real estate" or "tulip" ... and you pretty much get a rock solid scientific explanation of why markets crash and why, like the geological faults that criss-cross California, they will inevitably rupture.
In this review, I will try to do what the local tv reporter should have done ... focus on chapters 1 to 9. I will deal with chapter 10 in due course.
A large portion of the early part of the book is spent on describing financial markets and how traditional financial economics models have failed to properly explain and predict market crashes. This may not sound too original since other books, more or less, have done the same thing. What is very original in this book is that -- rather than a lot of handwaving and limiting itself only to historical anecdotes -- it is not afraid to employ maths (sorry, British English) and physics to scientifically analyze the nature of crashes.
I should note at this point that the level of mathematics knowledge required to understand this book is somewhere near a good American high school level of math. Sornette, unlike other physicists who have written 'popular' books, doesn't subscribe to the apocryphal story about Stephen Hawking being told by his publisher that each equation that is added to his book leads to some proportionate drop in book sales (although Sornette's book is nowhere near as popular as A Brief History of Time ... although it should be!). That having been said, one can safely ignore most of the maths and still get a lot of insights from the book.
Having said that, if you happen to be comfortable with mathematics, then all the better because there is a lot of beautiful and interesting mathematics and physics in this book. Sornette is a real scientist rather than a mercenary that happens to use (quasi-) scientific tools in the pursuit of filthy lucre. Reading his book, one gets the sense that Sornette really loves maths and science, because he talks about things like Polya urn models, Benford's law, evolutionary psychology, and computer simulations of 'artificial life' that others writing in the field of quantitative finance either don't care to write about or write about carelessly.
He invests time in introducing the reader to the ideas of complexity theory and econophysics. Fractals, power laws, critical phenomena, etc., are carefully described in this part of the book. The fact that Didier Sornette is a geophysicist by training is perfect for studying market crashes since there are many fruitful analogies that can be drawn between earthquakes -- which can be best understood in the light of complexity theory -- and market tremors.
What I found useful in this part of the book is the critique of how econometricians have tried to deal with what are clearly non-Gaussian aspects of financial market data. Sornette is right to criticize GARCH (Generalized Autoregressive Conditional Heteroskedasticity -- for you nerds out there) because the fat tails that model tries to pick up aren't fat enough. Sornette is also right to point out that most of the linear correlations useful for arbitrage have been arbitraged away (and the only linear correlations useful for arbitrage involve such short time intervals that the costs doesn't justify chasing after them). What are really interesting then are non-linear 'correlations' (really, they're not correlations or covariances, but some sort of inter-relationships that are happening at higher statistical moments ... presuming 'moments' are even meaningful in that context) that manifest themselves when markets exhibit complex, critical behavior (i.e., there is some dynamic 'attractor' where financial agents are either bidding up towards, or -- in the current credit crises -- bidding down towards -- in a non-linear way).
Having laid the intellectual groundwork, in the middle to later parts of the book, Sornette uses these concepts -- along with other tools from econophysics -- to develop models to predict crashes. Let me repeat that: Sornette comes up with models that almost precisely predict crashes! And, it seems to work pretty well (so far). What Sornette does is to fit and calibrate financial market data to a highly non-linear model. These predictive models seem to have done a decent job of predicting market downturns in a timely manner.
What's more, these models are not described in a hand-waving, 'black box' manner in the book. He prints them (or at least the non-proprietary version) in the book. That alone should make people run out, digging into their wallets, to buy this and other books by Sornette.
But this is where the skeptic in me awakens. In Nassim Nicholas Taleb's The Black Swan, the careful reader may have noticed an 'inside joke' when Taleb -- who heartily endorses Sornette's book(s) (as I am doing as well) -- suggests that people not tell Sornette about Taleb's skepticism about predictive models so that Sornette won't stop creating his interesting models.
Let me throw in my 2 pence into that discussion among intellectual comrades ... Yes, I think Sornette and his models are brilliant and they are probably right and even useful, but almost any non-Gaussian model would do a better job than Gaussian methods of calibrating predictive models based on financial data. Sornette's models are probably a lot better than the other non-Gaussian or pseudo-non-Gaussian models out there (like GARCH, stochastic volatility, etc.) but -- like most models -- there is probably some model risk with Sornette's models as well and their overuse will -- as Fischer Black noted with the Black-Scholes option pricing model -- lead to noise trading and model breakdowns.
Which leads me to chapter 10. In chapter 10 -- which, if you will recall, is what the local tv reporter exclusively focused in on -- is where Sornette 'predicts' that the year 2050 is when the era of economic growth will end. This seems vaguely reminiscent of Nostradamus or the Book of Revelations in the Bible. Unlike Nostradamus or St. John, Sornette uses equations, time series data, and ideas from self-organized criticality to make his bold prediction.
Is he right? I don't know. Maybe he's right. Maybe he's wrong. That was what bothered me 5 years ago. Honestly, it still bothers me ... but much less so, and certainly not enough to give this book anything less than my highest endorsement and praise.
Why? Because this book and its author goes against the grain.
Look at all of the so-called 'quants' that were terribly wrong about credit derivative models. They used rubbish pseudo-science in a mercenary way, dressed it up with things like AAA ratings that weren't worth the paper they were printed on, and then went off and bought luxury yachts while the rest of us are left holding the bag. When confronted with their failures, they inevitably turn defensive. There is no genuine humility there. The same folks (or people who will be virtually indistinguishable from them) will lead us into another round of financial catastrophes. It may not be mortgages next time. Maybe it will be inflation linked derivatives. Maybe it will be commodity linked derivatives. Whatever. It will happen. To borrow a phrase from a Jarvis Cocker song, they will kill again.
But Sornette is different. He is a real scientist. Every page of his book is embued with a sense that this man loves science. He is applying genuine science -- physics, math, etc. -- to scientifically examine a phenomena that is often more earth-shaking and causes more damage than the earthquakes that a geophysicist is usually concerned with ... market crashes. Maybe he will be wrong. He is willing to risk it and be honest about it. He has the genuine humility of a genuine scientist. If his models don't work out, then he will acknowledge that and work to improve it rather than cynically 'putting lipstick on a pig.'
I own all of his books and I believe that they will be useful to me in my attempts to understand uncertainty. But if his models don't work out ... well, that's science. Sornette, as a real scientist, is willing to risk failure to know ... to understand ... to learn how the world works. When I finally put my initial skepticism aside and read the book, I got that feeling as well. Not only do I take solace in that, I am inspired by it.