Saturday, May 19, 2007

Goldman Sachs (GS): Does Money Flow Matter?

I'm building out models of individual stock performance based on historical patterns of money flow. Let's take GS, which made a new high on Friday. On a five-day basis, money flow in GS has been outright negative: investors have pulled dollars out of the stock. It turns out that the correlation between five-day price change in GS and five-day money flow is only .08. This suggests that flow represents a different variable (sentiment) from simple price change.

But does money flow matter?

Going back to 2004 (N = 845), when GS has been up more than 1% over a five-day period and money flow during that time has been negative (N = 158), the next five days in GS have averaged a gain of .73% (87 up, 71 down). When GS has been up more than 1% and money flow has been positive over the five day period, the average gain in GS over the next five sessions has been only .06% (113 up, 105 down). It thus appears that five-day returns have been subnormal when GS has been up strongly over a five-day period on positive money flow. When GS has been up strongly on negative flow, returns have not been subnormal; indeed, they're slightly stronger than the average (.54%) from 2004 to the present.

How about when GS is down over a five-day period? When GS loses ground over five sessions and money flow is negative during that time (N = 164), the next five days have returned an average of a whopping 1.60% (115 up, 49 down). That is much stronger than the average gain when GS is down over five days on positive money flow (.13%; 106 up, 85 down).

What this suggests is that sentiment, as measured by money flows, makes a difference for short-term returns in GS. When five-day money flow has been positive, the next five days in GS have returned an average of only .08% (264 up, 224 down). When five-day flows in GS have been negative, the next five days have returned an average of 1.17% (226 up, 131 down).

If you envision separate models for different stocks--some of which will show positive sentiment and some negative--you can see how a worthwhile long/short strategy might evolve. By being long the stocks with the most bullish flow patterns and short the stocks with the most bearish patterns, you can trade off sentiment and eliminate a large measure of general market (directional) risk.

RELATED POSTS:

Overview of Relative Dollar Volume Flow

Going With the Money Flow

Money Flow and Returns in the Dow Jones Industrials

5 comments:

James said...

Hello,

Looking at the past week in GS, there was a fairly well-defined pennant pattern. The volume and price action gradually decreased over the first four days, with Wednesday and Thursday both being "inside" days to the prior day's range. The implied option volatility also shaved off 2.5% during this time.

It looks like money flow is actually a very good sentiment indicator during these periods. Negative MF during a consolidation pattern suggests the smart money is gladly (and patiently) accumulating shares at the bid price.

The inside days strike me as something to look into. Does an inside day (or at least a consolidation) occur in many of those N=158 bearish periods?

Jeff said...
This comment has been removed by the author.
Jeff said...

Hi Brett,

Why do you suppose this high price gain/ high money flow observation represents exhaustion for the individual stock, yet represents tradable momentum for the Dow index as a whole (a prior post)?

Perhaps, on balance (no pun intended), there is positive feedback at the index level among its constituent stocks? At some level, however, I would think one would observe the same exhaustion effect as with the individual stock?

Perhaps one way to reconcile the discrepancy would be to measure flow vs price effect for each stock in the index to generate a weighted accumulation vs exhaustion index?

Thanks for all your good work, Jeff

Brett Steenbarger, Ph.D. said...

Hi James,

That's a great idea for a research project: tracking money flow during flat market periods and seeing if there's any predictive value. Thanks for the idea; I'll see if I can post something of value in the near future--

Brett

Brett Steenbarger, Ph.D. said...

Hi Jeff,

I'd have to look more carefully into the interesting question you raise. Note that I'm looking at raw money flow data with GS in my post, not relative dollar volume flows as in my past posts.

Brett