While the final months of the year have just begun, the scoreboard for Mondays is not at all good so far. Here we look at the daily index return for the past several weeks, since September 20th, which is when investors pivoted from the September 18th FOMC meeting, to the October debt ceiling contention. There have been 21 trading days since September 20, through today. And while the results may vary slightly for later weekdays, as time moves forward, the results here are still considerably sticky.
The S&P 500 index is up 2%, during this time. But this return ignores the risk from the 4% drop, from September 18, through October 8. Most of that drop was during the past 21 trading days. Now Mondays were down 60% of the time (or 80% if we consider today an actual loss on a risk-calibrated basis, such as accounting for the record high margin utility). While other weekdays had a lower risk rate that was less than half the time, at 44%.
We show, in the illustration below, that the average Monday performance (in red) was poor, at -0.3%, with a relatively tight standard deviation of just less than 0.5%. As an overall contrast, the other days (in green) certainly collectively performed better, at +0.2%. The breakout of results by weekday will be shown in a table at the bottom.
Now neither of the portion differences noted earlier, nor the performance illustration above, are able to be considered statistically significant in isolation. However they fit the directional theme of Mondays being a more risky trading day (on either a risk neutral or absolute measurement basis) in recent weeks, and the results above taken together show to be more statistically significant.
We also show the ANOVA output below, in order to test the analysis of means. The results here, and even an advanced test for the difference of medians, both validate the description and significance of the results we just described. Notice the typical variation between weekdays is just slightly greater than the typical variation within weekdays. Also note that these raw units can be converted to percent by multiplying by 100, so for example, -0.003 would equal -0.3%.
| Summary of DailyChange
Weekday | Mean Std. Dev. Freq.
------------+------------------------------------
Mon. (1) | -.00303257 .00506192 5
Tue. (2) | -.00353137 .00865051 4
Wed. (3) | .00272263 .00746272 4
Thu. (4) | .00569407 .01258733 4
Fri. (5) | .00393512 .00535415 4
------------+------------------------------------
Total | .00095805 .00822574 21
Analysis of Variance
Source SS df MS F Prob > F
------------------------------------------------------------------------
Between groups .000297871 4 .000074468 1.13 0.3778
Within groups .001055387 16 .000065962
------------------------------------------------------------------------
Total .001353258 20 .000067663
Bartlett's test for equal variances: chi2(4) = 3.3571 Prob>chi2 = 0.500
The S&P 500 index is up 2%, during this time. But this return ignores the risk from the 4% drop, from September 18, through October 8. Most of that drop was during the past 21 trading days. Now Mondays were down 60% of the time (or 80% if we consider today an actual loss on a risk-calibrated basis, such as accounting for the record high margin utility). While other weekdays had a lower risk rate that was less than half the time, at 44%.
We show, in the illustration below, that the average Monday performance (in red) was poor, at -0.3%, with a relatively tight standard deviation of just less than 0.5%. As an overall contrast, the other days (in green) certainly collectively performed better, at +0.2%. The breakout of results by weekday will be shown in a table at the bottom.
Now neither of the portion differences noted earlier, nor the performance illustration above, are able to be considered statistically significant in isolation. However they fit the directional theme of Mondays being a more risky trading day (on either a risk neutral or absolute measurement basis) in recent weeks, and the results above taken together show to be more statistically significant.
We also show the ANOVA output below, in order to test the analysis of means. The results here, and even an advanced test for the difference of medians, both validate the description and significance of the results we just described. Notice the typical variation between weekdays is just slightly greater than the typical variation within weekdays. Also note that these raw units can be converted to percent by multiplying by 100, so for example, -0.003 would equal -0.3%.
| Summary of DailyChange
Weekday | Mean Std. Dev. Freq.
------------+------------------------------------
Mon. (1) | -.00303257 .00506192 5
Tue. (2) | -.00353137 .00865051 4
Wed. (3) | .00272263 .00746272 4
Thu. (4) | .00569407 .01258733 4
Fri. (5) | .00393512 .00535415 4
------------+------------------------------------
Total | .00095805 .00822574 21
Analysis of Variance
Source SS df MS F Prob > F
------------------------------------------------------------------------
Between groups .000297871 4 .000074468 1.13 0.3778
Within groups .001055387 16 .000065962
------------------------------------------------------------------------
Total .001353258 20 .000067663
Bartlett's test for equal variances: chi2(4) = 3.3571 Prob>chi2 = 0.500
No comments:
Post a Comment