No Value for Moving Average? From Kim Zussman
One question is whether stock price relationship to some moving average has any predictive value, in that stocks show negative correlations in shorter time periods, tested trending and reversal as a function of moving average.
SPY weekly closes (since 1993, with dividends) were categorized as above or below 20 week moving average (as of that week). Then, looked at mean week returns following up or down weeks, for cases when stocks were above or below 20w MA:
weeks
after up after dn
above MA -0.0006 0.0053
below MA 0.003 0.005
The relationship between the mean returns of these weeks shows better in ANOVA 95% CI's:
Individual 95% CIs For Mean Based on Pooled St Deviation
Level N Mean StDev
+———+———+———
all wk ret724 0.00221 0.02129 (—-*—–)
dnwk ov 187 0.00534 0.01720 (———*———)
dnwk und 129 0.00447 0.03023 (———–*———–)
upwk ov 331 -0.00063 0.01738 (——-*——-)
upwk und 72 0.00273 0.02656 (—————*—————)
+———+———+———+———
-0.0030 0.0000
0.0030 0.0060
Note that weeks after down weeks are slightly higher than all weeks. There is little difference between weeks following down weeks, irrespective of current 20w MA relationship (dnwk ov, dnwk und). But interestingly, during periods when stocks were above 20w MA, weeks following down weeks were higher than weeks coming after advances:
Two-sample T for upwk ov vs dnwk ov
N Mean StDev SE Mean
upwk ov 331 -0.0006 0.0174 0.001 T=-3.8
dnwk ov 187 0.0053 0.0172 0.0013
This suggests that when stocks are above (some) moving average, rather than trending up after advances, there is stronger reversal of declines. Of course the choice of moving average and return period will likely influence outcomes.