I read Why Stock Markets Crash by Didier Sornette last year, and I thought it was a pretty good book. Sornette builds on the quantitative work of Benoit Mandelbrot to make models of market bubbles using non-linear dynamics. The basic idea is that bubbles are created when the expectations of people in a market become spontaneously synchronised. It has some heavy-duty math to back up this idea, which makes it a less readable book than one might like, but to me Sornette’s models ring true.
One of the features of Sornette’s models is that he believes they include information that can predict the time when a market reaches an inflection point, which may in turn cause a crash. When I first read this, I was fairly skeptical. However, he is doing something that very few orthodox economist are willing to try – he is making public predictions of market crashes – including dates! In collaboration with several other authors, Sornette predicted date of the US housing market crash (roughly), and the date the 2008 oil bubble would burst. Most recently, they predicted a stock market crash in the Shanghai stock market, with the most likely dates being sometime between 17 and 27 August 2009. It didn’t. The market waited to crash until August 4th.
This is actually pretty remarkable. I’ve read several different blog posts discussing this issue, a lot of people are arguing whether or not a 20% drop in value in 2 weeks is a crash or not, and lots of other things. Many others talk about whether or not these are self-fulfilling prophecies. As much as I like econophysics, I’m pretty certain that it’s pretty close to impossible to credibly argue that articles in Physica A are moving the US housing market. Personally, I’m still not convinced that I buy Sornette’s underlying model, but I still think that the general approach has merit. The lesson that I take away from it all is that there is enormous potential value in modeling complex economic systems using non-linear methods. We need to do more of it since when we’re innovating, we are trying to introduce ideas into a complex, non-linear system.
Here’s the article with the heavy-duty math behind the prediction…