Didier Sornette is a mathematician, but don't hold it against him. He works in the same genre as Mandelbrot in being a gifted polymath considering epiphenomenon. Sornette is one of the most interesting practitioners in the field because he publishes and makes a few predictions along with being a fairly lucid writer with some interesting ideas on many concepts.
Sornette stated on July the 10th, 2009 that the Chinese Equity market would face serious challenges and peak between the 16th and 27th of July 2009. So far his prediction has worked out pretty well. Only time will tell if the prediction was effective on a risk: reward basis. There is a chart below. And yes, I am extremely loath to put up funny pictures with pointless lines and other forms of tripe on them as a way of divining the future for the lottery ticket crowd.
Figure A indicates the point at which the prediction and its associated paper,entitled Chinese Equity Bubble About to Burst was made public. Figure C represents the max risk if you had shorted on the earliest date of the prediction and held. You would just recently have crossed the risk experienced versus reward ratio of 1:1, so there is still time to see if this plays out to some interesting risk reward degree. So far so good, one could have Value Weighted into the position over the prediction period and be showing a paper profit. The question now is, when to exit?
One of his more interesting papers puts forth the statistical concept of Dragon Kings, this is a more empirically robust construct than the Black Swan in my mind and more interesting as it approaches causal agents. Dragon-Kings as near as I can tell are extreme outliers caused by cascading co-incident system failures. They aren't surprising empirically in the right statistical framework, but rather somewhat predictable, which may make them more useful.
The known unknowns are more interesting to study than unknown-unkowns if you will. Dragon-Kings may be more useful than Black Swans.
After all warning people about Black Swans and then telling them to fear what they can't know isn't so helpful. It produces anxiety, but not so much understanding or potential for action.
Dragon-Kings are functionally assumed to be derived or rather driven from multiple discrete similar or self similar processes co-inciding in time, like a perfect storm. A fuller explanation is available in this Sornette Paper: Dragon-Kings, Black Swans and the Prediction of Crises.
Sornette has put these ideas forth before in a book entitled Why Stock Markets Crash and a more academic book entitled Critical Phenomena in Natural Sciences: Chaos, Fractals, Selforganization and Disorder: Concepts and Tools (Springer Series in Synergetics).
The key thesis in Why Stock Markets Crash is that co-incident processes create feedback loops in a network dynamic which manifests itself outwardly in a time-series as a double exponential moving growth curve.
Basically a few things co-incide: cheap credit, securitization, housing policy, 1031 transfers etc. and out pops an unsustainable double exponential growth curve of debt growth for example.
I like the idea because rarely is an interesting system change driven by a single variable. Most complex systems tend toward stability, even in the face of a single vector shift. It is the co-incident shifts that can create in the words of Charlie Munger the Lalapalooza effect.
Buffet and Munger have stated that they look for positive co-incident factors in the firms in which they invest in.
Forget the silver bullet, bring an automatic to the party loaded with silver bullets.
The concept of coupled networks or states exhibiting phase change behaviour is intuitive after one reads a few example in the literature of network dynamics such as Barbasi.
Economically of course the question is begged, if participants in the collective became aware of such a state wouldn't they mitigate it thereby preventing the bubble? Read the book and or papers and see what you think. I have a link on my site entitled books which may provide some tips.
On the prediction front, Sornette's prediction still has to play out. The prediction was made Publicly at point A in the chart above with a range of dates, but no end objective or termination. So far the risk reward looks good, but the value in any prediction is its ability to be utilized on a risk reward basis with a degree of confidence, statistical and otherwise.
In that vein I offer up a critique of a sad blogger's past China prediction for analysis and comparison. This blogger boldly stated in June 2007 that China Shares will Bust by Sept 1 with a 40% decline. Here is a graph of the outcome.
Figure A was his prediction date, Figure B was his end state and Figure C was when a target of 40% from peak was reached. The prediction wasn't bad, but probably exposed the investor to 200-1,500 points of risk depending on the entry point. The blogger in question was me and I was using Sornette's approach.
It was a sloppy and stupid thing to post in retrospect as it was really only more econo pundit junk in the blogosphere without context and proper actionable rigor. In addition I published another piece indicating a 60% likelihood for the potential macro change of Chinese politics entitled the Rice Bowl Cracks. The range for that prediction is running out in the next few weeks, but again it is fairly useless unless someone finds interest in the Gott's Theorem tool. I still personally see political problems ahead for China, but that is an opinion and therefore has limited value to others.
All in all, this blogger has learned his lesson, predictions without proper measurements and rigor are only cheap opinions and as such, I will dial them back in the blog as the opinionsphere offers little in the way of education or insight for others. China may face severe political crises in the next few months, but that doesn't make my earlier prognostications right or useful for others.
I still recommend the Sornette books and papers for their analytical rigour and explanations of complex dynamics, coupled states and phase changes. I think Sornette is on to something interesting in the broader area of complex system dynamics. From a predication perspective it just needs to be phrased in a more actionable context in such a way that probabilities and risk limits are assigned and outcomes can be measured, an empirical falsifiable set of repeatable experiments.
If Sornette is onto something, maybe mega bubbles can be minimized, but I am not sure even perfect knowledge can get in the way of profoundly recurring human economic foibles. Read Kindleberger to get a healthy does of bubble reality history and how the shows always goes on or better yet read Shakespeare, only the actors names have changed in the ongoing human drama.