[From Bruce Gregory (2004.0828.1511)]
"Suppose you can simulate on your computer an artificial stock market.
Based on your ideas about how one piece of the economy connects to
another, you build an elaborate econometric model. It inputs data you
give it about the weather, inflation, economic growth, industry
specialization, and the companies being traded; and it calculates what
its algorithms tell it is the optimal "fundamental" value of the
company's stock. To that value, it adds millions of small random
changes--perhaps reflecting fictitious news events, or fickle investor
preferences, some taken separately, some added together. So, what kind
of variability generator will you use? If mild, the resulting price
charts will vary within a certain well-defined range; their trace will
be the product of many small computer-generated events. Very different
is wild variability, even though it can be "tuned down" to be less
extreme than Cauchy's. Wild price charts will be a hair-raising record,
mixing small movements with very, very large dislocations, many
computer-generated news items with a few cataclysmic bulletins, many
small transactions with large institutional block trades--all, a mix of
the small and routine with the large and rare. In such a wild world, an
imaginary investor participating in this economic simulation could be
wiped out overnight."
_The (Mis)Behavior of Markets: A Fractal View of Risk, Ruin, and
Reward_ by Benoit Mandelbrot and Richard L. Hudson.
"Great Doubt: great awakening. Little Doubt: little awakening. No
Doubt: no awakening."