CHAPTER IV. EMPIRICAL FINDINGS
IV.I D ATA
IV.4 T WO - DAY FILTER
information will subdue the noise, or the noise begins to dissipate soon. Also, it seems when lagged return is more extreme, the magnitude of future return tends to be greater.
Although the percentage of significant filters increases to 31.67%, and the picture of table seems to be much clearer and less random, again, Wang’s theory (1994) cannot be realized. It is hard to say what happened in the fourth quadrant- positive return follows the lagged price and the lagged volume change increase simultaneously- can be the evidence for the theory, because the results in the other three quadrants is with no relation to the theory. The result corresponds more to that found in Gervais, Kaniel, Mingelgrin (2001), even though we cannot see future price increase with lagged volume increase, but we can see the counterpart.
IV.4 Two-day filter
There is another possibility: the market would respond fully to the information at very short time. Thus, I shorten the calculation period for the lagged return and the lagged volume change to only two days. Still, the logic of methodology is the same. Measure two days’ return and volume change to fit in a filter, and then buy the stock at open price on the day just next to the strategy-formed-day. But I modify the grid width of price filter to 1.5%, and 20% for high-volume filter, 12% for low-volume filter. I also change the observation period from 1-4 weeks to two, four, and six days. It is reasonable
‧
to say that the market would respond fully and faster under this time frame.
Return filter
A better result was found this time. Table IX presents results for the case where profits are calculated two days after the strategy-formed-day; Table X four days, and Table XI six days.
As can be seen from Table IX to XI, future returns are all significantly positive in negative lagged return filter. Although the return in small lagged return filter is not economically significant (0.04% in filter of lagged return larger or equal to -1.5% and smaller than 0%), the return grows larger when the absolute value of filter becomes larger, and the return also becomes larger with time elapsing within 6 days. In positive lagged return filter, the future returns are not all significant, but they still show good outcome. When there is an extreme increase in price, the future returns do not tend to be reversed; but when price does not increase so extremely, the future returns also tends to be reversed. The magnitude of future return in positive lagged return filter is not as large as in negative return filter, but we still can see the return grows larger when the absolute value of filter becomes larger, and the return also becomes larger with time elapsing within 6 days (excluding those returns in extreme filter).
Return and volume filter
If we examine through Table IX to Table XI, we can see that no matter
‧
lagged volume. We can see that there are more significant filters in negative lagged volume filter than in positive lagged volume filter in Table IX. This phenomenon seems to disappear in the case where profits are calculated four days after the strategy-formed-day in Table X, yet does not disappear completely.Nevertheless, if we just focus on Table IX, we can see that the result has some similarity with Cooper’s (1999). In Cooper’s result, he compared the returns in the combination filter of return and volume change with the returns in the filter of return. He set the return in the return filter as standard. He found that the return in the “loser, high-volume” filter is positive but smaller than the return in the “loser” filter, i.e. the return is reversed rather than different from zero, so we can say that the returns in third quadrant is smaller than in return filter and the returns in fourth quadrant is larger than in return filter. It can also be described that winners and losers experiencing high growth in volume tend to experience reduced reversals, so was shown in
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Cooper’s study.
Figure IV presents the returns of three two-day filter strategies. Line 1 indicate the average return of filters whose lagged returns are smaller than -4.5% and lagged volume changes are smaller than -36%. Line 2 indicates the average return of filters whose lagged returns are smaller than -4.5% and lagged volume changes are between -12% and -36%. From line 1 and 2, we can see that with moderate decrease in lagged volume change, the following return is greater than that with lots of decrease in lagged volume change. Line 3 indicates the average return of filters whose lagged returns are smaller than -4.5% and lagged volume changes are larger than or equal to 60%. We can see that with high growth in lagged volume change, the following return tends to experience reduced reversals.
Figure IV. Return of two-day filter strategy
‧
At the beginning, I hypothesize, with time, the significant return, positive or negative, will lose its return slowly. However, as duration extends from two days to six days, the magnitude of return does not decrease but increase. It may imply that information needs some time to be fully utilized.