Kent Osband, author of “Pandora’s Risk: Uncertainty at the Core of Finance,” sent along this paper:
Fluids are turbulent when tiny differences in space or time make for gross differences in behavior. The mathematical signature of turbulence is an endless moment or cumulant hierarchy. Bayesian tracking of continuous-time processes turns out to have a similar mathematical structure. As a result, tiny doubts about regime change or tiny errors in estimation or calculation are prone under stress to balloon into gross differences of opinion. In effect, reasonable people are bound to disagree. This finding has profound implications for our understanding of financial markets and other instruments of social learning. In particular it explains forecast degradation, frequent trading, excess volatility, and GARCH behavior without imputing widespread irrationality. Rational learning makes markets turbulent.
I have not tried to evaluate Osband’s argument but the general idea is very appealing to me, and I hope that people with more knowledge than I will look into this.