Why Stock Markets Crash

Sornette, a geophysicist who applies statistical physics to financial markets, argues that large market crashes are not caused by the news that appears to trigger them but are endogenous outcomes of a cooperative process among traders, and that they are preceded by an identifiable signature.

His model treats a bubble as a system in which traders imitate one another with increasing strength, producing positive feedback. As imitation increases the system approaches a critical point, and the price path in that regime follows a specific mathematical form: faster-than-exponential growth decorated with accelerating oscillations, which he calls log-periodic power law behaviour. The oscillations arise from a discrete scale invariance in the underlying interaction structure. He argues this signature is detectable in advance and fits it to the 1929, 1987 and 2000 episodes and to several others including the 1997 Asian crisis.

The conceptual claim, that a crash is the culmination of a slow build-up of internal instability rather than a response to an external shock, is well supported by the observation that no adequate news event has ever been identified for the largest crashes.

The predictive claim is where the dispute lies. Critics have argued the log-periodic fits have too many free parameters, that the pattern can be fitted to price series with no subsequent crash, and that out-of-sample performance has been poor and selectively reported. Sornette has made public predictions with mixed results. The book requires mathematics and is written for readers comfortable with statistical physics.