Daily
Speculations
Spurious Serial
Correlation often
enters into calculations of momentum. The first such was discovered by working
where Kendall used the average of monthly prices and found that series based on
averages showed positive serial correlation. But of course if the close today
is higher than close one month ago, the average is likely to be higher than
last month's average and the close is likely to be higher than the current
average. Thus, if no change in the current month from close to close, the
average will still be higher. Similar defects with using high and lows to
compute averages. Also similar with respect to using the levels of an index
rather than the future since the Fisher effect of untraded stocks adds to the
spurious correlation.
The Wiz and I
discovered an interesting one in a recent study of durations. Given that you're
up at close from 4 days ago, i.e., Friday's close is above Monday's close, what
is average number of days until today's close goes below the close four days
ago close? I.e., until next Tuesday's close below last Wednesday's close or
next Monday's close below last Tuesday's close, etc. Answer is 4 days. But
given that today's close is above the close 4 days ago, for four consecutive
days, the average number of days until the next close is below the close 4 days
ago is 8 days. It's not a anomaly but merely a statistical property of random
numbers. --Vic (10/7/3)