Some of the best thinking occurs when we approach old topics in fresh ways. Consider the topic of sentiment. Through our normal lenses, we parse the world into bulls and bears. Suppose, however, we look at sentiment differently and measure it as the degree to which traders behave in more vs. less differentiated ways. If traders respond to markets in a non-differentiated way, they move as a herd, and we would expect transacted volume to be lopsided toward advancing or declining stocks. In a differentiated mode, market participants discriminate between better and worse investments and apportion volume to advancing and declining issues accordingly.
My measure of Volume Concentration takes the difference between advancing and declining volume on the NYSE and divides that by the sum of advancing and declining volume. This gives us an index that ranges from +100 (all directional volume is concentrated in advancing stocks) to -100 (all directional volume is concentrated in declining stocks), with the zero level denoting a market in which directional volume is evenly divided between advancing and declining issues.
I went back to 1965 and found that the average level of daily Volume Concentration is just a bit above zero, with a standard deviation of 38. For practical purposes, I set a two standard deviation threshold to represent herding behavior. Whenever the Volume Concentration Index (VCI) exceeds +75 or falls below -75, we'll call that herding sentiment.
What we find historically is that, since 1965 (N = 10,754 trading days), there have been relatively few days of herding sentiment. Specifically, we've had 160 days of herding in which volume has been highly concentrated in advancing issues and 223 days of herding in which volume has been highly concentrated in declining issues. Thus, a bit under 4% of all trading days meet our strict criterion of herding.
A look at the distribution of herding sentiment finds that it tends to occur in clusters. I examined the number of herding days over a moving 250 day period. The average for the entire sample was 8.6. We saw repeated readings over 15 during the market bottom and recovery around October, 1966; May, 1970; December, 1974; October, 1978; March, 1980; August, 1982; October, 1987; October, 1990; and March, 2003. The pattern is that we see herding behavior as traders panic on market declines toward the end of bear markets and then herding behavior as investors pour into markets to grab value at the start of bull markets. Indeed, one might say that it is precisely this transition of herding to the downside and herding to the upside--short-term panicky participants disgorging falling stocks and longer-term participants grabbing them as they rise--that makes for major market bottoms.
Interestingly, we're seeing just such a clustering of herding days presently in the market. Out of the past 250 trading sessions, fully 28 have qualified as herding days--far more than the average of 8.6. Indeed, going back to 1965, the only times we've had more than 20 days of herding in a 250-day period have been late 1974 into 1975 and late 1987 into 1988. Those were very important market lows.
One possible implication of the data is that the current market weakness, which has extended episodically from March through the present but is clearest in the housing and financial sectors, represents a decline of historic proportions despite its mildness when measured by price action in the large cap indexes. If the historical pattern holds in the present case, we should see an explosion of buying (herding to the upside) once investors perceive it's safe to return to the water. To be sure, this could occur at lower price levels. Nonetheless, the dates of high herding noted above--late 1966, mid 1970, late 1974, late 1978, early 1980, late 1982, late 1987, late 1990, and early 2003--were times to be thinking about owning stocks for the long run, not disgorging them with the crowd.
At periods of comparable extreme herding--late 1974 and late 1987--there were precious few reasons to think about owning shares. The current period offers similarly gloomy prospects. History has favored contrarian bulls at such times.
There is one other possible implication of the data that we will only know in the fullness of time. We have had readings continuously above 20 since August of this year. It may well be that, owing to the concentration of capital among hedge funds, pension funds, sovereign wealth funds, and the like--combined with the competitive need to contain risk and manage it daily--that we are seeing a secular rise in herding behavior. If this is true--and it is only a hypothesis--it has important implications for volatility and money management challenges for daytraders and portfolio managers alike.
Herding-as-sentiment in general, and the VCI as a specific measure of herding, provide us with an interesting framework for analyzing historical price patterns in the stock market. In a future post, I will take a more granular look at what happens historically after days of extreme herding.
What Drives Investor Sentiment?