Shocks, from Ken Smith

March 2, 2007 |

From Wikipedia's entry on the Dust Bowl:

 On November 11, 1933, a very strong dust storm stripped topsoil from desiccated South Dakota farmlands in just one of a series of bad dust storms that year. Then on May 11, 1934, a strong two-day dust storm removed massive amounts of Great Plains topsoil in one of the worst such storms of the Dust Bowl. The dust clouds blew all the way to Chicago where filth fell like snow, dumping the equivalent of four pounds of debris per person on the city. Several days later, the same storm reached cities in the east, such as Buffalo, Boston, New York City, and Washington, D.C. That winter, red snow fell on New England.

Abrupt changes occur as natural consequences of invisible butterfly activity in far corners of the planet. Temperature inversions occur in a minute. Ice age abrupt changes are historical facts. Spontaneous combustion occurs when least expected. Methane gas explosions occur unexpectedly. Tornadoes come in the middle of night, wiping out whole villages. Lightening strikes out of nowhere. Typhoons take lives and property when people least expect disaster.

Financial chaos occurs too. Abrupt drops in market prices are as historical as other natural disasters. We can say such drops are natural, are integral to the nature of things. The record is clear. Risk is inherent. Risk that is unaccountable, appears out of nowhere, cannot be foreseen.

Or can it be foreseen?

Nothing comes to mind that is more important to trading than finding hypotheses that provide insights for predicting the unforeseen. The Holy Grail is out there, somewhere, and as long as my mind is working I'll be looking for it.

Kim Zussman adds:

In comparison to recently a volatile stock market, Tuesday's 4% decline was quite unusual. In further tests, SP500 index daily returns (1980-07) were checked for instances of declines more than 3%. The absolute move was compared to the prior 20-day standard deviation, as a ratio, to rank big declines relative to recent market climate.

Variable = (abs(decline))/(standard deviation prior 20-day).

Turns out, Feb 27 was the biggest relative decline since 1989 and ranked 3/38, behind only such instances in 1989 and 1987 (and even greater than 10/97).

In regressions of the next 10 or 20-day returns as a function of the ratio, the following returns were positive and were not significantly related to the ratio.

Here are the dates and decline ratios (2/27 was 8.9):

Date |ret|/sdev
03/24/03 2.22
09/27/02 1.85
09/19/02 1.89
09/03/02 2.02
08/05/02 1.24
07/22/02 1.84
07/19/02 2.30
07/10/02 2.06
09/20/01 2.15
09/17/01 4.57
04/03/01 1.84
03/12/01 3.78
12/20/00 2.03
04/14/00 3.77
02/18/00 2.25
01/04/00 5.21
10/01/98 1.36
09/30/98 1.17
08/31/98 4.50
08/27/98 2.86
08/04/98 3.56
10/27/97 7.31
03/08/96 3.78
11/15/91 6.72
08/06/90 3.42
10/13/89 11.85
04/14/88 4.10
01/08/88 4.04
12/03/87 1.99
11/30/87 2.06
10/26/87 1.54
10/22/87 0.73
10/19/87 11.47
10/16/87 3.61
09/11/86 5.16
07/07/86 3.60
10/25/82 2.83
03/17/80 2.89






Speak your mind


Resources & Links