Okay, Sornette. You have our attention. Now show us the Money!

The Physics arXiv blog today recaps the prediction of Didier Sornette and crew that the Shanghai Composite was a bubble that would pop between 17 July and 27 July. It didn't. It popped on 4 August instead:

Thanks to the Physics arXiv blog for putting together the graphic. I was bemused when I first saw Sornette's paper. See earlier post on the prediction here. The Shanghai Composite indeed looked a little overheated. But nobody can see into the future, and (more importantly) Sornette gave us almost no information about how and why he was predicted a crash within this window. Nevertheless I was impressed with his courage to make a prediction that is so easily falsifiable. Of course, this high-risk, high-reward strategy is also a favorite of cranks.

That being said, the general idea that super-exponential growth is unsustainable is one that I have explored myself on this very blog. See excerpt from this earlier post on e: The Fundamental Speed Limit to Startup Growth.:

the very fastest that you can grow anything is by continuously compounding -- by feeding whatever new revenue you generate back into generating more revenue instantly. If you plot revenue as a function of time, and look at the base of the logarithm, the largest base results when you instaneously begin compounding every new bit of revenue as soon as it comes in. That base is e. There is no way to grow revenue faster than e.

Although these statements are unequivocal, I will be the first to admit that it's very difficult to verify any of them empirically by looking at actual firm growth rates -- much less the Shanghai Composite.

So here we are, with a falsified prediction and no information about how it was produced -- only tantalizing hints. Okay, Sornette. For better or worse, you have our attention. Let's see how you came up with your prediction so that we can understand whether this was simply a lucky guess.

Comments

[Andrew McCauley]

A recent paper titled, Dragon-Kings, Black Swans & the Prediction of Crises, by Sornette, asks whether power laws are really the whole story & suggests that in a significant number of complex systems, extreme events are even more exaggerated than predicted by the extrapolation of the power law distributions in their tail.

What is of interest from a practical point of view is that Sornette believes that these Dragon-Kings are identifiable in real time. This means they have predictive power. In terms of financial markets Sornette looks for qualifying signatures that identify a potential end to financial bubbles. Basically these signatures are statistical properties of time series price returns that are significantly different from the rest of the population. This process provides clues in a diagnostic sense that highlights the maturation of a system towards crisis.