Nov

29

A correction in SP 500 index (1950-07) was defined as:

As of Friday, this week's low was more than 10% below the high of any of the preceding 10 weeks, AND this was the first occurrence of such a 10% drop in 10 weeks.

There were 42 instances of such corrections since 1950, and the return for the following 5 and 10 weeks was flat: 

One-Sample T: nxt5, nxt10
Test of mu = 0 vs not = 0
Variable N      Mean     StDev  SE Mean          95% CI         T      P
nxt5     42 0.00079  0.06162  0.00950 (-0.01840, 0.02000)  0.08  0.934
nxt10    42 0.00395  0.08031  0.01239 (-0.02107, 0.02897)  0.32  0.751

21/42 times the next 5 weeks were positive
23/42 times the next 10 weeks were positive

Gibbons Burke adds:

Here is a graph of these results for all time frames up to t+50 (10 weeks). The peak of the bullishness occurs at the t+43 holding period relative to the event date on the 26th. The occurrence return lines are colored red or green relative to whether their return is positive or negative as of the (post-dictive) peak edge date. The thicker red and green lines are the most recent three occurrences and the most recent is the purple line, which is tracking the upper StdDev line almost exactly so far.

Kim Zussman replies:

Gibbons' interesting path diagram illustrates how traders with stops get penalized. I..e, if you stop out at -5% from initial position, you wipe out all the paths which cross -5 but go on to good profits later.

What about traders who decide to stop trading late in a good year (or greatly reduce size), to preserve year-based incentives? This is another kind of stop. Ostensibly they too are stopped out of potentially greater gains - with commensurately lower mean returns over time. And it would be interesting to speculate what effect on December is exerted by "up YTD" folks opting out and "down YTD" fellows trading furiously.

 


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search