Customer Reviews

Why Stock Markets Crash: Critical Events in Complex Financial Systems

5 star:		(20)
4 star:		(3)
3 star:		(3)
2 star:		(2)
1 star:		(3)

 

 

If you love to read works on economics, math and physics and love to assemble models of the world, I cannot recommend this book highly enough. Indeed, if economic models were this much fun when I was an undergraduate, I might have become an economist.

Funny thing though, this was not written by an economist, but by a geophysicist.
It seems physicists and psychologists in particular are writing more interesting economics books these days than economists themselves.

The core focus of the book is a derivation of a market model that includes value investors, momentum investors and the herding effect of individual economic agents acting in a world of partial information. The final model is stunning.

Sornette points out the main problem with predicting bubbles: even if all the signs say "yes," there is still a pretty good chance that the bubble will be self-correcting. Turns out chasing market bubbles is a little like chasing soap bubbles - they may simply disappear at any moment. Thus, the book and the model are of limited use in any type of market timing. Indeed, the model suggests that the market should now be in the tank, and yet it continues to hover on the higher side of its expected range.

As much as I loved the book, there was a slight aftertaste that this was all nothing but a very mathematical and high-minded type of technical analysis. That at base, when all was said and done, this was not all that different from the various "tools" in the chartist's handbook, e.g. MACD, RSI and OBV, etc., etc., etc. The difference may be solely that Sornette knows his statistics and would easily and readily dismiss any model which did not perform significantly different from chance.

Finally, this book will have you trotting out your old high school calculus book. It brought back memories of just how much fun mathematics can be.

All in all - I give it 5 stars.

Ever since I bought gold at $800 an ounce (the very top) 20 years or so ago, I have been fascinated by financial markets and their tendencies to produce bubbles that fool the majority. I know that complex systems and positive feedback and other phenomenons are at play and I wanted to find a book that covered the topic with enough depth. I thought Sornette's book was the one and some other reviews might make you think it is. Not for me though. Granted, it is extremely well researched with more than 460 references. It covers all the possible theories for stock markets price fluctuations and crashes. But its merit for me stops here.

The author warns the reader at the outset that mathematical explanations in smaller characters could be skipped in a first reading. The problem is that 90% of the book should be in smaller characters. The main text will be as hermetic to most readers than the small characters sections. Unless you have a graduate degree in a mathematics or physics and an extended experience in the disciplines that Sornette covers you'll be lost (BTW I do have one and I was lost). Here is an example of the kind of explanations you will find:

"The novel insight is that the arbitrary bubble component X, of an asset price plays a role analogous to the so-called 'Golstoine mode' in nuclear particle and condensed physics. Goldstone modes are the zero energy infinite-wavelength mode fluctuations that attempt to restore broken symetry."

Did you get it? I didn't.

This book might be of interest to researchers and acdemics in the field but it is way beyond the level of the educated general public. It is regreatble that Mr. Sornette
has chosen such a complex and esoteric way to treat the topic and has not made the slighest attempt to make it understandable to a wider public.

So I will keep looking for the book that will explain the fractal nature of stock markets in a documented but simple and interesting way.

Why Stock Markets Crash by Didier Sornette is an interesting book. He is a geophysicist who specializes in predicting failures in complex systems.

The book contains some rigorous mathematical proofs for a 'popular' book which means it probably won't be popular. But even if one merely glances the math, which again is mostly for proofs and for those with an analytical inclination, the overall text and thoughts and analysis are extremely thought provoking and insightful.

Its really good. You should read it if you have a background in stats or finance and are interested in the theory of efficient markets and the occurence of 'secular' events.

I must confess that I read
Didier Sornette's book with much pleasure; and I also
read the reviews posted on this site in particular those who
contend that, unless you are Stephen Hawking, you will be unable
to grasp the message the book conveys. Well, as I'm not
Hawking this opens a debate which is
worth some moments of reflexion.

Before seventeenth century physicists unraveled the mysteries
of vacuum and atmospheric pressure, the accepted explanation,
we are told, was that "nature is averse to vacuum". Obviously,
such an anthropomorphic explanation is both easy to grasp and
intuitively appealing. Although based on a number of nice
experiments, the scientific framework which replaced it did
not have the simplicity and intuitive attractiveness of the
former statement.
Why do stock markets crash? Well, the answer is very simple.
Because investors are averse to uncertainty, because
of an abrupt change in their mood and overall
utility function. OK. But unless, we can assess and
measure in some objective and quantitative way
either market uncertainty or the
investors' global utility function, we will not go
very far with such kind of explanations.

Now, let us come back to Didier Sornette's book. The author
proposes a new framework which is based both on a set
of new ideas and new scientific tools. As to the ideas
one can mention the two following.
* Stock market boom-crashes are a recurrent process,
which implies that they can be studied from a comparative
perspective. Specifically this means that it makes sense
to compare the crash of the Paris stock market in 1881
and the crash of the NASDAQ market in March 2000.
* What happens on stock markets is the result of the
COLLECTIVE BEHAVIOR of agents which means that the
decision which I take today depends upon what
my friends or colleagues David, John
and Stanley have done (and vice versa).

Simple as they may seem, these two ideas are quite
innovative. Even in the circle of behavioral economists
few (if any) researchers would probably accept them
altogether.

Naturally, in order to translate these ideas into
workable models we need some mathematical tools. Is
it at that point that I need to be Hawking? I don't
think so. Discrete scale invariance or
log-periodicity techniques are
not more complicated than modern theories of
equity arbitrage or Ito's lemma (which is used in
continuous time arbitrage theory), but these techniques
are less familiar to us and we therefore need some time
to get acquainted with them. Through numerous
figures and graphs this book provides an
introduction to these techniques
which is aimed for newcomers and pedestrians.

Of course, you don't need to take my view as
gospel truth, just experiment by yourself.

Didier Sornette has written an elegant and penetrating study of the complex elements which contribute to financial booms and their associated busts. Its most important conclusion is that potential crashes are proceeded by statistical fingerprints, largely independent of the particular markets involved, which permit their timing to be estimated within narrow limits by mathematical modeling, as demonstrated by numerous examples.

The book attempts two difficult challenges: first to model the potential timings of instabilities conducive to crashes in financial markets, and second to describe both the resulting models and their underlying phenomena intelligibly to the lay reader unfamiliar with much, or even all, of the mathematics involved. I found the author remarkably successful on both counts. The book reads uncommonly well provided one does not get distracted by the inevitable unfamiliarity of some of the mathematical terms, and in support of its main argument presents a wealth of interesting and uncommon information. Importantly, it also reflects a familiarity with the realities of financial markets, typically lacking in academic studies of market phenomena.

This appraisal will not be shared by all readers. If you are a fan of Kramer and Kudlow or prefer information about financial markets in sound bites from CNBC, or if you are looking for specific guidance on how to make money in markets, this book is not written for you. Furthermore, the book contains references to a considerable amount of serious mathematics which is likely to annoy some fraction of its readership. This can be circumvented, as suggested, by simply skimming whatever is unfamiliar. What is missed will have been addressed to a different audience, and not much of relevance will be lost. However, if you are frustrated by (or hostile to) unfamiliar mathematical terms and references, however inessential to the gist of the argument, best give it a pass.

For the rest, this is a deep study of engaging interest which repays more than one reading.

This is a fascinating book, both on finance and many other complex phenomena. I have now read it three times.

It appears he has made a genuine advance in understanding financial bubbles and crashes.

However reading the book does require that you are prepared to think. If you are after someone telling you what to buy or sell, this is not the book for you.

Yes there is some math but you really can skip it without losing too much. The quotes that people have included in their reviews are minor asides that merely point the interested to further related material.

Some others have commented that his predictions have not all worked out. This is all discussed at length in the book; in such a field predictions are not infallible. About 40% of market crashes are caused by external events and so are not predictable. However he seems to have the S&P500 worked out. Last years he predicted a choppy rally in 1Q2003, then from 2Q2003 a major fall ending in 1h2004. So far so good.

The author aside from the problem of crashes presents an insightful exposition of tipping points. I don't know why his approach makes it clearer and deeper than those of Watts and Barabasi --is it due to his using financial markets as a base? or his being an expert at fat-tailed dynamics?
His work builds on the "abyssus abyssum invocat" (panic begets panics) and the dynamics of compounding disequilibria. In addition the notion of "CRITICAL POINT" is made very clear.
Honestly I don't care for the idea of crashes; the same concepts can apply to sudden and unexpected euphorias.
I learned more from this book than any other on disequilibrium.

This book is a very comprehensive effort to model (and hopefully) predict market behavior. It is a work in progress that will require more time for definitive testing; thus, the "record" is bound to change as the model evolve.

The basic thesis of the book is that market crashes are "outliers", that is, events that fall outside the realm of "normal" market behavior and obey a different set of rules.

Why is this work important to all of us? Say granpa invested 10$ for you in 1928 and you took your profits in december 2002. Let's assume 4 scenarios:

1) Buy and forget. You'd have $10,957.
2) Missed the 30 best months you'd have $154.
3) Missed the 30 best months and the 30 worst months, your $10
would be worth $18,558
4) Missed the 30 worst months, your $10 would be $1,317,803.

We can safely conclude that preservation of capital has its place in any investment strategy. :-)

Caution: The book is replete with math and statistics, so, caveat emptor. However, the text does a very good job at explaining the equations and figures presented.

This book is not easy. Nor is investing for that matter. But the effort is rewarding.

Buy low and sell high: In trading stocks this cliché is obvious, but its ramifications can be extremely painful for all those involved, as well as many who are not, especially when a large collection of traders act in concert and engage in massive sell-offs. Financial bubbles, stock market crashes, and out-of-control speculation have been the subject of countless books and research papers and dozens of Hollywood movies, and mathematicians, economists, systems analysts, and financial engineers have spent thousands of hours of time attempting to understand and predict financial meltdowns, all with varying degrees of success. Some of these researchers have argued that it is the large-scale, collective properties of the financial markets that must be understood if stock market crashes are to be predicted or at least anticipated quantitatively.

The author of this book, a geophysicist by training, is one of these and has taken the "buy low-sell high" strategy and some straightforward mathematical constructions to give an interesting and plausible explanation of why stock markets crash. Intuitively, traders synchronously try to buy low and sell high, and a "herding effect" results in a rapid sell-off that results in a market crash. This behavior is analogous to what one observes in physical systems that are close to a `critical point', where they can undergo `phase transitions' and their behavior as they near the critical point has been shown to be "universal" in the sense that quantities called `scaling exponents' can be calculated explicitly and measure the "universality" of the systems near the critical point.

Naturally the stock market crashes of 1929 and 1987 are ones that that will stand out in every reader's mind and ones that must be test examples for the author's assertions. He discusses these two examples and many others in enough detail that a convincing case is made. The techniques he uses however are out of the mainstream of econometrics, and no doubt some in this mainstream will there be highly skeptical of his conclusions. The physicist-turned-financial engineer however will be delighted, as there will be many familiar concepts and constructions throughout the book. There may be a few new ones to such a reader however, such as `algorithmic complexity' and `computationally irreducible'. The clarity of the author's writing and the many real-world examples that he employs makes the assimilation of these concepts and others much more palatable than would be the case in a standard mathematical monograph on the subject. In addition, the techniques he uses can be applied to areas of finance that are not discussed in the book, such as the mortgage "antibubble" in net credit losses for second-lien (HELOC) loans that began in the second quarter of 2006. And with respect to applications, the author is honest enough to note that he refrains from discussing many of them in detail since his involvement in them is strictly proprietary. Most refreshingly, charts, graphs, and tables appear throughout the book, as would be necessary in the validation of any algorithm or predictor in financial theory.

There are also many interesting "toy models" in the book that enhance the didactic quality. One of these concerns the dependences that can occur in successive price variations. The author constructs a model where there is zero correlation but where one can predict the current price variation with an accuracy of over 50% by only knowing the variations in the past two days. This example serves as a good counter to the idea that the frequency distribution and two-point correlation function must always be non-zero in order to obtain successful prediction. The author goes on to use this example to show how "drawdowns", rather than the correlation structure, play the predominant role in measuring price changes. This example, among others, also illustrates the author's belief that the "standard" models in the financial industry have great difficulty in dealing with large financial crashes.

Although the author uses somewhat elementary mathematics in the book, its implications are profound. One will find for example discussions of log-periodic oscillations in hierarchical systems, the renormalization group, and complex fractal dimension. Central to the author's case is the concept of discrete scale invariance, which is a specialization of the continuous case, and which is manifested in real data by log-periodic oscillations. These should be viewed as corrections to simple power law scaling. The author discusses a few natural systems that exhibit log-periodicity, such as bats and dolphins, and in evolutionary biology. He also discusses other examples in mathematics such as the Newcomb-Benford law. Pre-crash stock market data is fitted to expressions that have log-periodic corrections to a pure power law and the validity of the fits discussed in great detail. The author's arguments are powerful and convincing, and the formalism that he outlines needs to be part of every financial analyst's toolbox.

There are many books on many topics. The one that is inspiring and teaches you something that will resonate for a long time is a rare find indeed.

Sornette's work is one such book. Complex systems, critical phenomena, nonlinear interactions are buzzwords of modern physics thrown all over. Yet tying them together in a consistent analysis, presented as simply as possible (but no simpler!), as an application to a real system while remaining engaging is a non-trivial feat. The book is extremely well researched, cites over 400+ references and could be recommended to any beginning physics graduate student to learn something about the scientific method, critical phenomena, non-linear dynamics, statistics, and probability theory. It is so well rounded in that nothing is taken for granted, everything is introduced by plausible argument, evidence, historical analysis or derivation, with references to back up every statement. Still his work is no ordinary dreary text book. Sornette's authorship is spirited and well orgnized, spiked with historical facts, anecdotes and technical asides that enrich but do not distract from the main premise of the book, namely to provide an answer to the question in the title.

The main thrust of the book is as follows: Bubbles (and subsequent crashes) are systemic instabilities present in markets as an innate characteristic. When the market is in a bubble regime, it displays super-exponential growth that can be fitted to a power-law. Power-law price divergence is akin to the behavior of physical systems when approaching critical points (for exmple when undergoing a phase transition). Asset prices in critical regime, in addition, display log-periodic osciallations which are a feature of "discrete scale invariance" (systems that are self-similar only under discrete magnification ratios). The log-periodicity is a saving grace, as it is nearly impossible to distinguish power laws from exponential laws in noisy data. Locking onto oscillations improves discrimination substantially. By inverse conclusion, markets that can be fitted to log-periodic super-exponential growth are unstable and subject to a crash. His treatment is probabilisitic in nature, by the very simple argument: If a looming crash were known with certainty, arbitrage traders would take advantage of this and thus remove the instability.

After developing this core tenet, several chapters are devoted to analyzing historical stock market crashes under this lens. He proceeds with a discussion on prediction and its perils, and at last closes with a wider look at economic and population dynamics. Notable among those, for example, is the conclusion that Moore's Law is actually incorrect: The growth of transistor densities is not exponential but super-exponential, predicting a critical point around 2030.

Sornette chose a non-academic format to publish his ideas. And while arduous mathematical derivations have been skipped, the concepts presented still borrow from physics and mathematics to no small extent. Without an upper division background in those disciplines, the reader may be hard pressed to follow the argument.

This book is not a trading system manual either. This book gives no predictions. This book is about a theory, an analysis and models. And as such it is fascinating.