It is sometimes funny when traditional such as Macquarie look at herding stock market behavior and act as if they were Christopher Columbus discovering new concepts. Just like Columbus might not have been the first to “discover” America — that point has been in dispute, particularly among native Americans — practitioners in managed futures CTA strategies might look at a recent presentation delivered by Macquarie Research on the impact of herding on stock prices and think this is like a course in Momentum 101.

Get The Full Ray Dalio Series in PDF

Get the entire 10-part series on Ray Dalio in PDF. Save it to your desktop, read it on your tablet, or email to your colleagues.

We respect your email privacy


herding photoBoth human and technical factors surrounding "herding" effect

It is sometimes difficult, if not impossible, to determine the tone, tenor and context of an investment presentation by looking at nothing but slides that don't contain deep contextual support. Given this constraint the commonality between what appears like a traditional equity investing approach to “herding” – a large number of people buying a stock at the same time – is interesting to consider from the standpoint of algorithmic market formation concepts. When trend exhaustion occurs, herding can deliver pain to investors. But early in the price trend gain can occur if the herding delivers price persistence.

Herding, "crowded trades" in some lingo, is to a degree technical in nature.  It can be measured to based on demonstrable facts such as the numbers of investors entering a position, the extent of potential total portfolio in a position and numbers of course what matters most, the stock price. Fundamental investors look at herding from the standpoint of what drives the discretionary thought process. Some algorithmically-driven analysis might look at it from the standpoint of a supply and demand math equation without necessarily considering the fundamental reason why a market move occurred.

To lead off their August presentation titled “Stalking the herd: Why herding and crowding matter for equity investors,” Macquarie’s London-based analysts Giuliano De Rossi and Jakub Kolodziej pick a stock to demonstrate herding that was perhaps most influenced by discretionary human behavior: Herbalife.

Insider Herding

Macquarie considers "co-momentum" concept of correlated stock picking

The individual actions of activist hedge fund managers in the Herbilife stock story could be attributed as primary causation for the stock market sell-off. In December of 2012, Herbalife stock, trading near $40 per share, experienced extreme selling following news that Pershing Square’s William Ackman had placed a target on the corporation. Claiming the stock and its business model was a fraud, Ackman, then among a small elite cadre of activists, immediately garnered attention and the herding began. The stock quickly moved into the mid-$20 range as short-term traders were looking for blood.

From Macquarie’s perspective, the measurement was not emotional and the herding behavior somewhat common. Citing the “quant meltdown” of 2007 – a title in dispute in certain algorithmic circles – the report points to specific measurements of crowding.  For Macquirie this is defined as changes in institutional ownership and looks at “stock positions, average active share from HF pairs, short interest.” In particular, the report pointed to a “co-momentum idea” that looks at stocks that move together.

Separately, while it is common in algorithmic trading to actively consider correlated markets that "move together," supporting the "co-momentum idea," certain algorithmic market analysts use other tools as a primary indicator of momentum validity. In fact, using the co-momentum concept with Herbalife might not be appropriate. The first rule of determining correlated market validity is to recognize potential core performance drivers. Determining if the core performance driver has a corollary in other markets or stocks is a meaningful consideration in algorithmic market analysis. In the case of Ackman’s Herbalife thesis, the primary performance drivers could be considered allegations of fraud and the Pershing Square team’s ability to generate media coverage and influence potential regulatory policy outcomes. For a correlation to be valid, the circumstantial performance drivers must be consistent.

In other words, if one was to consider a corollary “sell” to Herbalife, it would not be another firm in the same industry sector unless they, too, were engaged in potentially fraudulent business practices. The corollary selling concept typically better extends to stocks where a takeover performance driver is more due to the potential for industry strength or a positive government ruling changing the market dynamic.

Macquarie points to individual signals of a market with price persistence or mean reversion potential

The Macquarie report interestingly moves into the academic backdrop, pointing to general rules that support the phenomena existing. Much like academic studies point to the validity of price persistence, or market price trends, having always been either absent or present on a cyclical basis, the concept behind stock market herding can be academically documented.

Using a backtest of herding behavior without lagging data, they discovered a time horizon variable that extends from one month to 12. Perhaps among the more interesting points is the formation period points to trend success.

Pointing to academic studies, De Rossi and Kolodziej highlight that short formation and holding periods produce what is known in algorithmic circles as trend continuation. “Stocks that are bought by the herd tend to outperform.”

Separate algorithmic market analysis notes that volatility can be a catalyst for quick formation patterns. This is because volatility often precedes price persistence, a correlation that increases with severity. The strength of the price trend is often determined by the point in time market players pick up a buy or sell signal. Different systems, both discretionary and systematic, recognize trends at different points in time and add to the “herding” behavior, which creates sustained price momentum.

There comes a time in all herding behavior history when momentum exhaustion occurs. Analysis of how to determine this point has meet with variable results. The Macquarie pointed to longer formation and holding periods ultimately resulting in reversals. Separate analysis might point out another way to express this concept is to say that momentum absent volatility that comes from a price reaction to fundamental news are often lower in win percentage when the “herding” is not generally recognized.