Monday, February 26, 2007

Tracking the Dollar Volume Flows of Stocks

Over the weekend, I've been working on some fascinating research to answer a group of questions: How much money is actually flowing into or out of a particular stock on a given session? Are these dollar volume flows increasing or decreasing over time? Is the stock attracting more dollars than other stocks in the same sector? Than stocks in other sectors? And--most important of all--how is the amount of dollars flowing into a stock related to how that stock behaves going forward?

As I indicate in my latest Trader Performance entry, Relative Dollar Volume Flow appears to be a promising measure of underlying demand or supply for a stock. The idea is simple: take every trade that occurs in a stock during the day and multiply the price at which the trade was transacted by the volume of that trade. If the trade occurred on an uptick (or if the last price change was an uptick), you add this value to a running cumulative total for the day. If the trade occurred on a downtick (or if the last price change was a downtick), you subtract this value from the running cumulative total for the day. At the end of the day, the final cumulative total reflects the buying or selling of large market participants (institutions that transact large blocks, large portfolio managers), because their transactions account for most of total (given the large volume of their purchases or sales).

If you have institutions eager to get into a stock, they will buy large positions on upticks. If those institutions are eager to bail out of a stock, they will sell large positions on downticks. The logic is similar to Market Delta, only this is a measure intended for longer-term analysis. (It would be interesting to study it on an intraday basis for signals, but so far I have not done so).

I refer to this as *Relative* Dollar Volume Flow because I divide the cumulative total at the end of the day by that day's trading volume. This provides a better comparison of the money flowing into one stock vs. another: One million dollars flowing into a large cap stock that is an institutional favorite will reflect lower Relative Dollar Volume Flow than one million dollars flowing into a lightly traded small cap issue.

One application of Relative Dollar Volume Flow is to ETFs. Does the amount of money flowing into or out of an ETF tell us anything about the Fund's performance going forward?

I conducted a small investigation of the Dow Jones Industrial Average ETF (DIA) by tracking its daily Relative Dollar Volume Flow from the beginning of 2004 to the present (N = 771 trading days). When the 20 day Flow for DIA was positive (net money flowing into DIA; N = 442), the next 20 days in DIA averaged a gain of only .05% (245 up, 197 down). When the 20 day Flow was negative (net money flowing out of DIA; N = 329), the next 20 days in DIA averaged a very solid gain of 1.11% (227 up, 102 down). This is a pronounced reversal effect: when we see money flowing out of DIA over a 20-day period, returns have been superior. When DIA has attracted dollars over a 20-day period, returns have been subnormal.

Very interestingly, if we calculate the Dollar Volume Flow for the Dow by individually summing up the Flows for the 30 component stocks, this value correlates with the Flow for DIA by only .14. In other words, money flowing into and out of DIA is not especially reflective of money flowing in and out of the individual Dow stocks. Rather, DIA's flows (like those of other index ETFs) may be more skewed by the activity of hedgers and program traders who might not be making long-term portfolio adjustments.

Indeed, the Dollar Volume Flows for the summed 30 Dow stocks shows that money has been continuously flowing into those issues through the bull market and really accelerated after we made a bottom in July of 2006. Although there are individual issues with net dollar outflows (MSFT is notable in that regard), there is little evidence at present that money is systematically flowing out of the Dow. (Inflows have slowed over the last two weeks, but remain positive).

An especially promising strategy is to identify stocks with favorable expectations based on their Relative Dollar Volume Flows and those with subnormal expectations and create a long/short strategy of buying the former and selling the latter. This would create a diversified portfolio that would be relatively market neutral. I'm also beginning to take a look at falling stocks with rising Dollar Volume Flows and rising stocks with weakening flows to see if we can anticipate turns in those issues. Clearly, there's a lot of room for interesting research here. I hope to report results in the near future.


Mark said...

Hi Brett,

First of all, thanks so much for giving us this terrific resource, and giving it for free. You obviously put a lot of thought and work into it and I, for one, appreciate it. I save all of your posts.

I'm betting I speak for a few others when I say that I'd also like to know the nuts and bolts of how you go about getting the information you use for your research. For example, today--what tool did you use to collect the money flow information? Does your data provider provide info on individual trades? It would definitely be helpful to me if you could include how you set up your data collection when you have these great ideas so we don't have to spend time recreating your wheel, and perhaps doing so incorrectly.

Thanks again and keep up the great work.


Brett Steenbarger, Ph.D. said...

Hi Mark,

Yes, my raw data came from my data provider, RealTick. All my analyses are performed in Excel. One of the topics I'm hoping to take up in a future presentation or video is how to identify market patterns with simple Excel analysis. Thanks for the interest--


John said...

Hi Brett,

I have an idea on your relative volume work. I am not sure if you can get this data but if you can consider doing this:

Take the trade size over the last N days and get the median. Get the standard deviation of the trade size. Divide the trade size by the standard deviations. Define large size by a measure of the ratio of trade size to the standard deviation. If trade size is over a certain number of deviations then consider it institutional.

I am not a statistics guru (you may already know that). Perhaps rather then divide by the standard deviation you could just divide by the median.

This would make the various stocks comparable because you are basing what is institutional on how big trades are relative to the past.

I come up recently with a way to tell if a stock is overbought or oversold. I take all the stocks over 30 dollars and over 250k in volume and get their relative strength versus the SP500. I get the median and standard deviations. I look at stocks that are 5 times away from the market on the upside or downside. I then look to enter those stocks when they show weakness or strength.

Brett, I just found your site today. I think it is great.



Scott said...

I like your thoughts on Dollar Volume. I use this metric in my trading systems and find it helps to normalize symbols and see what has the most relative volume for the day.

I would like to take this concept further and try to gauge the flow of money from one index to another. It seems logical that the money doesn't dissapear when it leaves the DIA but would flow into another index. How would you go about measuring this flow of money between indexes?


Scott said...


When this money flows out of the DIA it seems like it could be valuable to develope a way to track where it goes. It wouldn't seem that it would just disappear but would move to a different index. I have been giving this idea some thought but would like to get your input on a way to measure and test where the money goes when it leaves DIA.

Thanks for your work

Brett Steenbarger, Ph.D. said...

Hi Scott,

I think it would be hard to track those flows entirely, since some money may move to other asset classes. I do think that, by tracking flows in and out of individual stocks and sectors, you can gain a sense of money moving in and out of stocks overall.