In his book Freakonomics the economist Stephen Levitt describes many methods of cheating in baseball, sumo, minority testing at public schools, and bagel stealing. He is good at coming up with algorithms that find cheaters and this might be highly applicable to market moves. He is excellent at explaining such things as increased cheating on bagel eating before Christmas and Thanksgiving when employees are more anxious, and less cheating around holidays like Labor Day and July 4th when there is no anxiety. A change occurred in the results but he attributes this to post 9-11 stress. But this is the promiscuous hypothesis squared that so many behaviorists at Chicago are rightfully accused of, and alternate explanations are available for all of the regressions that Dr. Levitt likes to report.
I examined whether holidays in the stock markets tend to show this same bias with negativism around Christmas and Thanksgiving, and positivism around the others, and whether there was a 9-11 shift. I looked at the moves in the one and five day periods before and after these holidays for the past eight years to see whether any predictive hypotheses might be confirmed.
To bring the holiday study into some systematic and refutable form I looked at the moves in the prior and subsequent one and five days for each of the holidays from 1995 to 2004 with a view to accessing the consistency of any one, and the difference between enthusiasm for the stress and non-stress holidays.
|Move From N Days Prior||Move to N Days After|
Since the standard deviation of each one day move is approximately 10 points, and the standard deviation of each five day move is about 25 points, it is apparent that not one holiday shows a consistent pattern even approaching one standard deviation.
The results of the individual holidays are even more glaringly consistent with chance. Consider for example the most hopeful: the moves five days after Memorial Day averaging +10 points, and the moves following MLK Day averaging -11 points. The moves the five days after Memorial Day were: 1995 +13; 1996 -11; 1997 -2; 1998 -19; 1999 +35; 2000 +89; 2001 -10; 2002 -44; 2003 +36; 2004 +20.
The period started out with a large negative cumulative sum thru 1998. You might have played it that way until 1999, at which time you would have lost a fast 124 points on the short side. Then when you played it in 2000 for a long you'd have a nice -54 points in 2001 and 2002, and then if you shifted back or called it a day you'd miss or lose 56 points in 2003 and 2004. The results for each of the holidays was completely consistent with randomness. No holiday showed a consistent performance. The anxious holidays were no different from the non-anxious ones. There was no tendency to go up before or after to an inordinate degree. All in all, a very disturbing result casting grave doubt on seasonal sages, and economists given to promiscuity.