Wednesday, September 05, 2007

Stock Market Trends and Volume: Tracking Institutional Participation


I recently posted on the topic of how volume from large traders is disaggregated during execution, making it difficult to identify when institutional participants and large locals are moving the markets. From this post, you'll be able to see one way I'm approaching that issue.

First, the conceptual underpinnings: I track the median 1 minute volume over a five-day lookback period. If the average trading price for the current minute exceeds the average trading price for the prior minute *and* if the average trading volume for the current minute exceeds the median one-minute volume over the prior 5 days of trading, then I call this a "buying bar". We're getting higher prices on elevated volume.

Similarly, when the average trading price of the current minute is below that of the prior minute *and* the average trading volume for the current minute exceeds the five-day median, we get a "selling bar". That represents lower prices on elevated volume.

All other one-minute bars are ignored. The idea here is that, although individual trades are disaggregated, aggregating them on a one-minute basis enables us to see if there is above average participation in the market at that time.

The top chart (click for greater detail) shows the median one-minute volume in the ES futures over a moving five-day lookback period. This is depicted for the period 7/12/07 through 9/4/07. Note how, during the market decline, average volume steadily increased. Note how volume decreased during the market's last leg down and now how it is increasing during the recent market rise.

The idea is to see if higher or lower prices are attracting increased market participation. That certainly was the case Tuesday, as higher prices brought expanded volume through much of the day.

The bottom chart (click for greater detail) illustrates a one-day (400 minute) moving average of buying bars minus selling bars (as defined above) for the period 8/29/07 through 9/4/07. This tells us whether we're getting more periods of significant buying interest vs. selling interest. Note that, in a bull swing, there is a general expansion of buying vs. selling bars and that expansions in the number of selling bars occur at successively higher price levels.

These are merely first approximations of indicators I'll be developing. Notice that the indicators can be constructed for any size bar and for any stock or futures contract with consolidated volume data. By pairing trend indicators with such volume measures, we can infer whether large traders are participating in directional movement and jump aboard such moves early.

RELEVANT POSTS:

Analyzing Market Volume

NYSE TICK Volume
.

3 comments:

T. said...

I enjoyed reading this post very much. Interestingly, I have been working on the same themes lately. I am currently in the process of developing an indicator based on volume correlations; I'll post it to you once I have a better idea of its usefulness.

One question, when you say mean volume over the past 5d, do you mean arithmetic mean (average = (1+2+3)/3)?

Best regards,
T.

Brett Steenbarger, Ph.D. said...

Hi T,

Volume is averaged over 5 days worth of one minute readings (about 2000 values in all). And, yes, keeping a running correlation between price change and volume creates an indicator much like the Power Measure I described earlier.

Brett

Mike said...

He asked if it was an arithmetic mean. Why didn't you answer his question? Clearly he is asking b/c of the contrast b/w the arithmetic mean and the geometric mean. An important one I'll add.