Weekends in October, from Kim Zussman
In honor of Friday's perilous drop, checked October Friday (Ths cls-Fri cls) and weekend (Fri cls-Mon cls) returns in SP500 index, both before and after Oct 1987. Its being 20 years since that fateful month, the post-sample is 1988-2007 and the pre-sample 1967-1986.
Here is regression of weekend vs Friday pre-1987:
The regression equation is
F-M pre = 0.00071 + 0.462 Fri pre
Predictor Coef SE Coef T P
Constant 0.0007 0.0011 0.65 0.520
Fri pre 0.4617 0.1313 3.52 0.001
S = 0.0103498 R-Sq = 12.4% R-Sq(adj) = 11.4%
Weekend strongly correlated with Fridays.
What about Octobers after 1987?
The regression equation is
F-M post = 0.00145 + 0.0333 Fri post
Predictor Coef SE Coef T P
Constant 0.0014 0.0012 1.13 0.261
Fri post 0.0333 0.0969 0.34 0.732
S = 0.0116829 R-Sq = 0.1% R-Sq(adj) = 0.0%
Poof! No more weekend effect.
As an extension on this extension, here is a check on symmetry of pre-87 weekend effect. Re-ran regression with 2 independent variables:
Fri preUP = Friday return if up, otherwise 0
Fri preDN = Friday return if down, otherwise 0
The regression equation is
F-M pre = - 0.00147 + 0.732 Fri preUP + 0.019 Fri preDN
Predictor Coef SE Coef T P
Constant -0.0015 0.0017 -0.87 0.387
Fri preUP 0.7324 0.2061 3.55 0.001
Fri preDN 0.0190 0.2922 0.06 0.948
S = 0.0102408 R-Sq = 15.3% R-Sq(adj) = 13.3%
Prior to 1987, positive weekend correlation with Friday returns arose from continuation of up Fridays (as opposed to continuation of downs).