Investors may underreact when information arrives in small, continuous bits
Every field has its mysteries to solve. Physics has dark matter and parallel universes, history has Jack the Ripper and Amelia Earhart, the art world has Banksy and NFT valuations, and behavioral finance has momentum investing, among others. Momentum investing, the tendency of a stock’s relative performance to be predictable from its relative performance over the previous three to 12 months, is an anomaly to the efficient market hypothesis. Despite many theories put forward by academics, there is still no consensus as to why momentum continues to exist and work.
A working paper by University of Lausanne’s Amit Goyal, Emory University’s Narasimhan Jegadeesh and UCLA Anderson’s Avanidhar Subrahmanyam evaluates current theories on momentum. Following the growing trend of meta-research in which researchers check other researchers’ findings for reproducibility and whether the results hold up in out-of-sample data (data other than the data originally used to train their model), they use international equity market data to retest the theories. The researchers suggest that one explanation of momentum, in particular, performs strongly across all their tests. Additionally, they identify the types of markets in which momentum is the strongest within their dataset.
The team of researchers has extensive experience in momentum studies. UCLA’s Subrahmanyam’s past research on the topic focuses on better understanding why momentum profits continue to exist. Jegadeesh uncovered the return pattern of momentum in a seminal paper with Sheridan Titman in 1993 (both at UCLA at the time). Goyal has conducted research on several finance anomalies, as well as similar meta-research testing of variables for predicting equity premium.
An Abundance of Theories
Goyal, Jegadeesh and Subrahmanyam chose the MSCI Developed (ex-U.S.) and the MSCI Emerging market indices for their international dataset and used each index separately and jointly to test existing theories on momentum. Past research has already shown that momentum strategies are profitable in non-U.S. markets as well as U.S. markets. Since most of the existing explanations were already tested on U.S. market data, using an international dataset provided out-of-sample data and tested the strategies on equal footing.
The theories or explanations for momentum come in two categories. In one category, a variable is created as a proxy for the explanation, such as book-to-market for valuation and as an indicator of investor overconfidence. The change in the variable’s value is a predictor for the state of momentum in the market. The second category examines the relationship between momentum and the market’s direction or volatility. The researchers test both categories of theories.
The table below describes the theories and, where relevant, provides the variable used in the model. These theories were tested, using the international data, with regression models and portfolio-based analyses.
Winner, Winner, Frog Leg Dinner
The results of the researchers’ tests cannot even be described as close. Many of the theories did not appear to hold up at all in the international dataset while others had mixed results. Only one theory had robust results across all tests and was the clear winner: This was Da, Gurun and Warachka’s theory known as the frog-in-the-pan hypothesis. The name of the theory comes from the premise that a frog dropped in boiling water will jump back out of the water, but a frog put into room temperature water that is only very slowly brought to a boil will underreact to the rising temperature and be boiled to death. Similarly, FIP pins the cause of momentum on investors underreacting to small bits of information released continuously over time while reacting appropriately to large chunks of information released all at once.
The variable designed for this theory aims to measure if information is likely arriving in large discrete chunks that would cause big moves in the stock on just a few days or in small continuous bits more likely to cause many small positive (or negative) returns over time. To determine this, it counts only the sign (positive or negative) of the daily returns rather than the size of the move. A high percentage of positive daily returns relative to negative daily returns implies that a stock, with let’s say a 10% return over the last six months, can attribute that gain to many small positive returns and a flow of information that is more continuous than if the same 10% gain had a low percentage of positive daily returns to negative gains. In such a case, it’s more likely information came in larger, more discrete chunks causing a few large daily returns.
Within the international dataset used, momentum profits were larger during upward-trending markets and during less volatile markets. These results appear to confirm findings in previous research studies. Interestingly, FIP’s variable measure is lower in upward-trending markets and during low volatility. This suggests that information might reach investors in small bits at a slower, more continuous pace in these types of markets.
While many theories attempt to explain the mystery of momentum, the frog-in-the-pan hypothesis warrants further attention given its strong performance on the international data. Additionally, the researchers’ results suggest that overweighting momentum strategies as markets trend upward or while volatility is low may cause their returns to leap even further.
Distinguished Professor of Finance; Goldyne and Irwin Hearsh Chair in Money and Banking
About the Research
Goyal, A., Jegadeesh, N., & Subrahmanyam, A. (2022). What Explains Momentum? A Perspective From International Data.