From: John Conover <john@email.johncon.com>
Subject: Re: Mahathir's Phobia of Speculators
Date: 19 Apr 1999 18:52:02 -0000
Burkhard C. Schipper writes:
> Well, if anybody has better technology then I expect that this technology
> will diffuse into the market especially through such movements at financial
> markets. Nobody can really own such superior technology. But what kind of
> better technologies? To forecast? Well, I think exact forecasting is
> impossible through emergent properties in financial markets. An innovation
> can not be forecasted because it is new. If something is new then nothing or
> little is known about it. Nonlinear dynamic systems are difficult to
> forecast.
>
> conover@netcom.com srieb in Nachricht <7fdtjk$qp7@sjx-ixn8.ix.netcom.com>...
> >
> >Hi Burkhard. But what if a few of the large speculators' economic
> >methodologies and/or mathematical techniques were better than the
> >rest?
> >
> > John
> >
Hi Burkhard. This is true, non-linear dynamical systems are difficult
to forecast in a deterministic static sense. But as Mandelbrot pointed
out in the early 70's, prices of financial instruments that exhibit
high kurtosis can give "essentially riskless profits,"[1] and
concludes that for Hurst exponents outside the range of 0.4 to 0.6,
less than 300 transactions would be required to get the riskless
profits.
As an unrelated side note, in the reference he also argues that
fractal Brownian motion is an inadequate model for financial
instrument prices unless the market is very inefficient, which is
contradictory with the paradigm of the Efficient Market Hypothesis,
(EMH,) which assumes that all information available in the market is
already reflected in a financial instrument's price-ie., the Hurst
exponent is exactly 0.5, meaning a fractal Brownian motion model,
which is efficient.
John
[1] Mandelbrot, B. (1971), "When can price be arbitraged efficiently?"
A limit to the validity of the random walk and martingale models,"
Review of Economics and Statistics 53, pp 543-553.
--
John Conover, john@email.johncon.com, http://www.johncon.com/