At 18:05 07/01/2003 +0000, Dr Allan White wrote:
>Colleagues,
>
>Statisticians may be amused (or upset) by a recent article in the
>Financial Times last weekend. In the `Money' section, six `experts' from
>the asset management industry were providing forecasts for the stock
>market for the coming year. Two of these `experts' appeared to be saying
>that the market was unlikely to fall far this year because it has already
>suffered three bad years in succession, which made a fourth bad year
>statistically unlikely.
The markets have actually fallen for four or more years in a row *more*
often than would be expected from a random walk.
However, the runs test shows that annual returns do show some significant
dependence.
For example, four or five consecutive bull or bear years happens more
frequently than one would expect, whereas (bear, bull, bear) and (bull,
bear, bull) both happen much less frequently than would be expected if the
market followed a random walk.
See:
http://www.cs.ucl.ac.uk/staff/M.Sewell/runs_test/
Cheers
Martin
PS Many thanks to all those who suggested improvements to my significance test.
http://cgi.mvsewell.plus.com/cgi-bin/significance_test.php
|