Volume, from Jim Sogi
Interesting drop in volume on today's up move in ES:
2010-02-01, 1896011
2010-01-29, 3068079
2010-01-28, 3052088
2010-01-27, 2634603
2010-01-26, 2449263
2010-01-25, 2080292
The old TA theory is that an up move on low volume is weak, and a down move on increasing volume is strong, but it doesn't prove out scientifically.
Sushil Kedia comments:
A homegrown theory I developed borrowing on the Chair's applications of the concept of struggle to markets interprets volume as a struggle for price discovery. Extending this with memetics, a higher volume indicating a higher struggle for price discovery meme implies an ongoing persistence of the meme. So, within any time frames or patterns or noise, if you perceive a meme then interpreting lower volume as lesser struggle tilting towards consonance and thus implying a fading meme comes by. This way of looking at price and volume relationships does lead to several testable ideas getting the gut feel closer to science.
Bill Rafter writes:
The only use we have found for volume is as a surrogate metric for volatility. Indeed S&P volume and VIX have very interesting relationships. However the standard TA mantras that (a) volume “confirms” price and (b) volume-weighting indicators makes them better, have not been confirmable.
Here is an excellent graphic I prepared relevant to volume.
Dr. Rafter is President of Mathematical Investment Decisions, a quantitative research consultancy
Kim Zussman writes:
Here's another check. SPY daily returns (c-c), with volume traded, 1993-present, were used to check for big up days = >+1% (0.01). Then calculated relative volume (RV) as:
(today's volume) / (average volume prior 5 days)
Then compared next day's return for two cases; if it followed a big up day which had low RV (RV<0.8) or a big up with high RV (RV>1.3):
t-Test: Two-Sample Assuming Equal Variances
low RV hi RV
Mean 0.0010 -0.0015
Variance 0.0002 0.0002
Observations 151 146
Pooled Variance 0.0002
Hypothesized Mean Difference 0.0000
df 295
t Stat 1.5263
P(T<=t) one-tail 0.0640
t Critical one-tail 1.6500
P(T<=t) two-tail 0.1280
t Critical two-tail 1.9680